chore(release): v0.11.3

refactor: simplify daily quota error messages
fix(core): ensure loop detection respects session disable flag (#12347 )
2026-07-23 16:20:57 -07:00 · 2025-10-31 20:24:24 +00:00 · 2025-10-31 11:11:19 -07:00 · 2025-10-31 10:31:04 -07:00 · 2025-10-31 10:31:03 -07:00 · 2025-10-31 10:31:02 -07:00
40 changed files with 2271 additions and 1681 deletions
@@ -473,21 +473,6 @@ a few things you can try in order of recommendation:
    "loadMemoryFromIncludeDirectories": true
    ```

- **`chatCompression`** (object):
-  - **Description:** Controls the settings for chat history compression, both
-    automatic and when manually invoked through the /compress command.
-  - **Properties:**
-    - **`contextPercentageThreshold`** (number): A value between 0 and 1 that
-      specifies the token threshold for compression as a percentage of the
-      model's total token limit. For example, a value of `0.6` will trigger
-      compression when the chat history exceeds 60% of the token limit.
-  - **Example:**
-    ```json
-    "chatCompression": {
-      "contextPercentageThreshold": 0.6
-    }
-    ```
-
 - **`showLineNumbers`** (boolean):
  - **Description:** Controls whether line numbers are displayed in code blocks
    in the CLI output.
@@ -244,13 +244,13 @@ their corresponding top-level category object in your `settings.json` file.
    example `{"run_shell_command": {"tokenBudget": 2000}}`
  - **Default:** `undefined`

- **`model.chatCompression.contextPercentageThreshold`** (number):
+- **`model.compressionThreshold`** (number):
  - **Description:** Sets the threshold for chat history compression as a
-    percentage of the model's total token limit. This is a value between 0 and 1
+    fraction of the model's total token limit. This is a value between 0 and 1
    that applies to both automatic compression and the manual `/compress`
    command. For example, a value of `0.6` will trigger compression when the
    chat history exceeds 60% of the token limit.
-  - **Default:** `0.7`
+  - **Default:** `0.2`

 - **`model.skipNextSpeakerCheck`** (boolean):
  - **Description:** Skip the next speaker check.
@@ -1,12 +1,12 @@
 {
  "name": "@google/gemini-cli",
-  "version": "0.11.0-nightly.20251021.e72c00cf",
+  "version": "0.11.3",
  "lockfileVersion": 3,
  "requires": true,
  "packages": {
    "": {
      "name": "@google/gemini-cli",
-      "version": "0.11.0-nightly.20251021.e72c00cf",
+      "version": "0.11.3",
      "workspaces": [
        "packages/*"
      ],
@@ -596,7 +596,6 @@
        }
      ],
      "license": "MIT",
-      "peer": true,
      "engines": {
        "node": ">=18"
      },
@@ -620,7 +619,6 @@
        }
      ],
      "license": "MIT",
-      "peer": true,
      "engines": {
        "node": ">=18"
      }
@@ -2425,7 +2423,6 @@
      "integrity": "sha512-t54CUOsFMappY1Jbzb7fetWeO0n6K0k/4+/ZpkS+3Joz8I4VcvY9OiEBFRYISqaI2fq5sCiPtAjRDOzVYG8m+Q==",
      "dev": true,
      "license": "MIT",
-      "peer": true,
      "dependencies": {
        "@octokit/auth-token": "^6.0.0",
        "@octokit/graphql": "^9.0.2",
@@ -2606,7 +2603,6 @@
      "resolved": "https://registry.npmjs.org/@opentelemetry/api/-/api-1.9.0.tgz",
      "integrity": "sha512-3giAOQvZiH5F9bMlMiv8+GSPMeqg0dbaeo58/0SlA9sxSqZhnUtxzX9/2FzyhS9sWQf5S0GJE0AKBrFqjpeYcg==",
      "license": "Apache-2.0",
-      "peer": true,
      "engines": {
        "node": ">=8.0.0"
      }
@@ -2640,7 +2636,6 @@
      "resolved": "https://registry.npmjs.org/@opentelemetry/core/-/core-2.0.1.tgz",
      "integrity": "sha512-MaZk9SJIDgo1peKevlbhP6+IwIiNPNmswNL4AF0WaQJLbHXjr9SrZMgS12+iqr9ToV4ZVosCcc0f8Rg67LXjxw==",
      "license": "Apache-2.0",
-      "peer": true,
      "dependencies": {
        "@opentelemetry/semantic-conventions": "^1.29.0"
      },
@@ -3009,7 +3004,6 @@
      "resolved": "https://registry.npmjs.org/@opentelemetry/resources/-/resources-2.0.1.tgz",
      "integrity": "sha512-dZOB3R6zvBwDKnHDTB4X1xtMArB/d324VsbiPkX/Yu0Q8T2xceRthoIVFhJdvgVM2QhGVUyX9tzwiNxGtoBJUw==",
      "license": "Apache-2.0",
-      "peer": true,
      "dependencies": {
        "@opentelemetry/core": "2.0.1",
        "@opentelemetry/semantic-conventions": "^1.29.0"
@@ -3043,7 +3037,6 @@
      "resolved": "https://registry.npmjs.org/@opentelemetry/sdk-metrics/-/sdk-metrics-2.0.1.tgz",
      "integrity": "sha512-wf8OaJoSnujMAHWR3g+/hGvNcsC16rf9s1So4JlMiFaFHiE4HpIA3oUh+uWZQ7CNuK8gVW/pQSkgoa5HkkOl0g==",
      "license": "Apache-2.0",
-      "peer": true,
      "dependencies": {
        "@opentelemetry/core": "2.0.1",
        "@opentelemetry/resources": "2.0.1"
@@ -3096,7 +3089,6 @@
      "resolved": "https://registry.npmjs.org/@opentelemetry/sdk-trace-base/-/sdk-trace-base-2.0.1.tgz",
      "integrity": "sha512-xYLlvk/xdScGx1aEqvxLwf6sXQLXCjk3/1SQT9X9AoN5rXRhkdvIFShuNNmtTEPRBqcsMbS4p/gJLNI2wXaDuQ==",
      "license": "Apache-2.0",
-      "peer": true,
      "dependencies": {
        "@opentelemetry/core": "2.0.1",
        "@opentelemetry/resources": "2.0.1",
@@ -3812,7 +3804,6 @@
      "resolved": "https://registry.npmjs.org/@testing-library/dom/-/dom-10.4.1.tgz",
      "integrity": "sha512-o4PXJQidqJl82ckFaXUeoAW+XysPLauYI43Abki5hABd853iMhitooc6znOnczgbTYmEP6U6/y1ZyKAIsvMKGg==",
      "license": "MIT",
-      "peer": true,
      "dependencies": {
        "@babel/code-frame": "^7.10.4",
        "@babel/runtime": "^7.12.5",
@@ -4324,7 +4315,6 @@
      "integrity": "sha512-AwAfQ2Wa5bCx9WP8nZL2uMZWod7J7/JSplxbTmBQ5ms6QpqNYm672H0Vu9ZVKVngQ+ii4R/byguVEUZQyeg44g==",
      "devOptional": true,
      "license": "MIT",
-      "peer": true,
      "dependencies": {
        "csstype": "^3.0.2"
      }
@@ -4335,7 +4325,6 @@
      "integrity": "sha512-4hOiT/dwO8Ko0gV1m/TJZYk3y0KBnY9vzDh7W+DH17b2HFSOGgdj33dhihPeuy3l0q23+4e+hoXHV6hCC4dCXw==",
      "dev": true,
      "license": "MIT",
-      "peer": true,
      "peerDependencies": {
        "@types/react": "^19.0.0"
      }
@@ -4613,7 +4602,6 @@
      "integrity": "sha512-6sMvZePQrnZH2/cJkwRpkT7DxoAWh+g6+GFRK6bV3YQo7ogi3SX5rgF6099r5Q53Ma5qeT7LGmOmuIutF4t3lA==",
      "dev": true,
      "license": "MIT",
-      "peer": true,
      "dependencies": {
        "@typescript-eslint/scope-manager": "8.35.0",
        "@typescript-eslint/types": "8.35.0",
@@ -5381,7 +5369,6 @@
      "resolved": "https://registry.npmjs.org/acorn/-/acorn-8.15.0.tgz",
      "integrity": "sha512-NZyJarBfL7nWwIq+FDL6Zp/yHEhePMNnnJ0y3qfieCrmNvYct8uvtiV41UvlSe6apAfk0fY1FbWx+NwfmpvtTg==",
      "license": "MIT",
-      "peer": true,
      "bin": {
        "acorn": "bin/acorn"
      },
@@ -5736,7 +5723,8 @@
      "version": "1.1.1",
      "resolved": "https://registry.npmjs.org/array-flatten/-/array-flatten-1.1.1.tgz",
      "integrity": "sha512-PCVAQswWemu6UdxsDFFX/+gVeYqKAod3D3UVm91jHwynguOwAvYPhx8nNlM++NqRcK6CxxpUafjmhIdKiHibqg==",
-      "license": "MIT"
+      "license": "MIT",
+      "peer": true
    },
    "node_modules/array-includes": {
      "version": "3.1.9",
@@ -7015,6 +7003,7 @@
      "resolved": "https://registry.npmjs.org/content-disposition/-/content-disposition-0.5.4.tgz",
      "integrity": "sha512-FveZTNuGw04cxlAiWbzi6zTAL/lhehaWbTtgluJh4/E95DqMwTmha3KZN1aAWA8cFIhHzMZUvLevkw5Rqk+tSQ==",
      "license": "MIT",
+      "peer": true,
      "dependencies": {
        "safe-buffer": "5.2.1"
      },
@@ -8101,7 +8090,6 @@
      "integrity": "sha512-GsGizj2Y1rCWDu6XoEekL3RLilp0voSePurjZIkxL3wlm5o5EC9VpgaP7lrCvjnkuLvzFBQWB3vWB3K5KQTveQ==",
      "dev": true,
      "license": "MIT",
-      "peer": true,
      "dependencies": {
        "@eslint-community/eslint-utils": "^4.2.0",
        "@eslint-community/regexpp": "^4.12.1",
@@ -8691,6 +8679,7 @@
      "resolved": "https://registry.npmjs.org/cookie/-/cookie-0.7.1.tgz",
      "integrity": "sha512-6DnInpx7SJ2AK3+CTUE/ZM0vWTUboZCegxhC2xiIydHR9jNuTAASBrfEpHhiGOZw/nX51bHt6YQl8jsGo4y/0w==",
      "license": "MIT",
+      "peer": true,
      "engines": {
        "node": ">= 0.6"
      }
@@ -8700,6 +8689,7 @@
      "resolved": "https://registry.npmjs.org/debug/-/debug-2.6.9.tgz",
      "integrity": "sha512-bC7ElrdJaJnPbAP+1EotYvqZsb3ecl5wi6Bfi6BJTUcNowp6cvspg0jXznRTKDjm/E7AdgFBVeAPVMNcKGsHMA==",
      "license": "MIT",
+      "peer": true,
      "dependencies": {
        "ms": "2.0.0"
      }
@@ -8709,6 +8699,7 @@
      "resolved": "https://registry.npmjs.org/statuses/-/statuses-2.0.1.tgz",
      "integrity": "sha512-RwNA9Z/7PrK06rYLIzFMlaF+l73iwpzsqRIFgbMLbTcLD6cOao82TaWefPXQvB2fOC4AjuYSEndS7N/mTCbkdQ==",
      "license": "MIT",
+      "peer": true,
      "engines": {
        "node": ">= 0.8"
      }
@@ -8938,6 +8929,7 @@
      "resolved": "https://registry.npmjs.org/finalhandler/-/finalhandler-1.3.1.tgz",
      "integrity": "sha512-6BN9trH7bp3qvnrRyzsBz+g3lZxTNZTbVO2EV1CS0WIcDbawYVdYvGflME/9QP0h0pYlCDBCTjYa9nZzMDpyxQ==",
      "license": "MIT",
+      "peer": true,
      "dependencies": {
        "debug": "2.6.9",
        "encodeurl": "~2.0.0",
@@ -8956,6 +8948,7 @@
      "resolved": "https://registry.npmjs.org/debug/-/debug-2.6.9.tgz",
      "integrity": "sha512-bC7ElrdJaJnPbAP+1EotYvqZsb3ecl5wi6Bfi6BJTUcNowp6cvspg0jXznRTKDjm/E7AdgFBVeAPVMNcKGsHMA==",
      "license": "MIT",
+      "peer": true,
      "dependencies": {
        "ms": "2.0.0"
      }
@@ -8964,13 +8957,15 @@
      "version": "2.0.0",
      "resolved": "https://registry.npmjs.org/ms/-/ms-2.0.0.tgz",
      "integrity": "sha512-Tpp60P6IUJDTuOq/5Z8cdskzJujfwqfOTkrwIwj7IRISpnkJnT6SyJ4PCPnGMoFjC9ddhal5KVIYtAt97ix05A==",
-      "license": "MIT"
+      "license": "MIT",
+      "peer": true
    },
    "node_modules/finalhandler/node_modules/statuses": {
      "version": "2.0.1",
      "resolved": "https://registry.npmjs.org/statuses/-/statuses-2.0.1.tgz",
      "integrity": "sha512-RwNA9Z/7PrK06rYLIzFMlaF+l73iwpzsqRIFgbMLbTcLD6cOao82TaWefPXQvB2fOC4AjuYSEndS7N/mTCbkdQ==",
      "license": "MIT",
+      "peer": true,
      "engines": {
        "node": ">= 0.8"
      }
@@ -10205,7 +10200,6 @@
      "resolved": "https://registry.npmjs.org/ink/-/ink-6.2.3.tgz",
      "integrity": "sha512-fQkfEJjKbLXIcVWEE3MvpYSnwtbbmRsmeNDNz1pIuOFlwE+UF2gsy228J36OXKZGWJWZJKUigphBSqCNMcARtg==",
      "license": "MIT",
-      "peer": true,
      "dependencies": {
        "@alcalzone/ansi-tokenize": "^0.2.0",
        "ansi-escapes": "^7.0.0",
@@ -13388,7 +13382,8 @@
      "version": "0.1.12",
      "resolved": "https://registry.npmjs.org/path-to-regexp/-/path-to-regexp-0.1.12.tgz",
      "integrity": "sha512-RA1GjUVMnvYFxuqovrEqZoxxW5NUZqbwKtYz/Tt7nXerk0LbLblQmrsgdeOxV5SFHf0UDggjS/bSeOZwt1pmEQ==",
-      "license": "MIT"
+      "license": "MIT",
+      "peer": true
    },
    "node_modules/path-type": {
      "version": "3.0.0",
@@ -13930,7 +13925,6 @@
      "resolved": "https://registry.npmjs.org/react/-/react-19.1.0.tgz",
      "integrity": "sha512-FS+XFBNvn3GTAWq26joslQgWNoFu08F4kl0J4CgdNKADkdSGXQyTCnKteIAJy96Br6YbpEU1LSzV5dYtjMkMDg==",
      "license": "MIT",
-      "peer": true,
      "engines": {
        "node": ">=0.10.0"
      }
@@ -13941,7 +13935,6 @@
      "integrity": "sha512-cq/o30z9W2Wb4rzBefjv5fBalHU0rJGZCHAkf/RHSBWSSYwh8PlQTqqOJmgIIbBtpj27T6FIPXeomIjZtCNVqA==",
      "devOptional": true,
      "license": "MIT",
-      "peer": true,
      "dependencies": {
        "shell-quote": "^1.6.1",
        "ws": "^7"
@@ -13975,7 +13968,6 @@
      "integrity": "sha512-Xs1hdnE+DyKgeHJeJznQmYMIBG3TKIHJJT95Q58nHLSrElKlGQqDTR2HQ9fx5CN/Gk6Vh/kupBTDLU11/nDk/g==",
      "dev": true,
      "license": "MIT",
-      "peer": true,
      "dependencies": {
        "scheduler": "^0.26.0"
      },
@@ -16037,7 +16029,6 @@
      "integrity": "sha512-5gTmgEY/sqK6gFXLIsQNH19lWb4ebPDLA4SdLP7dsWkIXHWlG66oPuVvXSGFPppYZz8ZDZq0dYYrbHfBCVUb1Q==",
      "dev": true,
      "license": "MIT",
-      "peer": true,
      "engines": {
        "node": ">=12"
      },
@@ -16248,8 +16239,7 @@
      "resolved": "https://registry.npmjs.org/tslib/-/tslib-2.8.1.tgz",
      "integrity": "sha512-oJFu94HQb+KVduSUQL7wnpmqnfmLsOA/nAh6b6EH0wCEoK0/mPeXU6c3wKDV83MkOuHPRHtSXKKU99IBazS/2w==",
      "dev": true,
-      "license": "0BSD",
-      "peer": true
+      "license": "0BSD"
    },
    "node_modules/tsx": {
      "version": "4.20.3",
@@ -16257,7 +16247,6 @@
      "integrity": "sha512-qjbnuR9Tr+FJOMBqJCW5ehvIo/buZq7vH7qD7JziU98h6l3qGy0a/yPFjwO+y0/T7GFpNgNAvEcPPVfyT8rrPQ==",
      "dev": true,
      "license": "MIT",
-      "peer": true,
      "dependencies": {
        "esbuild": "~0.25.0",
        "get-tsconfig": "^4.7.5"
@@ -16442,7 +16431,6 @@
      "integrity": "sha512-p1diW6TqL9L07nNxvRMM7hMMw4c5XOo/1ibL4aAIGmSAt9slTE1Xgw5KWuof2uTOvCg9BY7ZRi+GaF+7sfgPeQ==",
      "dev": true,
      "license": "Apache-2.0",
-      "peer": true,
      "bin": {
        "tsc": "bin/tsc",
        "tsserver": "bin/tsserver"
@@ -16724,6 +16712,7 @@
      "resolved": "https://registry.npmjs.org/utils-merge/-/utils-merge-1.0.1.tgz",
      "integrity": "sha512-pMZTvIkT1d+TFGvDOqodOclx0QWkkgi6Tdoa8gC8ffGAAqz9pzPTZWAybbsHHoED/ztMtkv/VoYTYyShUn81hA==",
      "license": "MIT",
+      "peer": true,
      "engines": {
        "node": ">= 0.4.0"
      }
@@ -16779,7 +16768,6 @@
      "integrity": "sha512-4nVGliEpxmhCL8DslSAUdxlB6+SMrhB0a1v5ijlh1xB1nEPuy1mxaHxysVucLHuWryAxLWg6a5ei+U4TLn/rFg==",
      "dev": true,
      "license": "MIT",
-      "peer": true,
      "dependencies": {
        "esbuild": "^0.25.0",
        "fdir": "^6.5.0",
@@ -16896,7 +16884,6 @@
      "integrity": "sha512-5gTmgEY/sqK6gFXLIsQNH19lWb4ebPDLA4SdLP7dsWkIXHWlG66oPuVvXSGFPppYZz8ZDZq0dYYrbHfBCVUb1Q==",
      "dev": true,
      "license": "MIT",
-      "peer": true,
      "engines": {
        "node": ">=12"
      },
@@ -16910,7 +16897,6 @@
      "integrity": "sha512-LUCP5ev3GURDysTWiP47wRRUpLKMOfPh+yKTx3kVIEiu5KOMeqzpnYNsKyOoVrULivR8tLcks4+lga33Whn90A==",
      "dev": true,
      "license": "MIT",
-      "peer": true,
      "dependencies": {
        "@types/chai": "^5.2.2",
        "@vitest/expect": "3.2.4",
@@ -17668,7 +17654,6 @@
      "resolved": "https://registry.npmjs.org/zod/-/zod-3.25.76.tgz",
      "integrity": "sha512-gzUt/qt81nXsFGKIFcC3YnfEAx5NkunCfnDlvuBSSFS02bcXu4Lmea0AFIUwbLWxWPx3d9p8S5QoaujKcNQxcQ==",
      "license": "MIT",
-      "peer": true,
      "funding": {
        "url": "https://github.com/sponsors/colinhacks"
      }
@@ -17684,7 +17669,7 @@
    },
    "packages/a2a-server": {
      "name": "@google/gemini-cli-a2a-server",
-      "version": "0.11.0-nightly.20251021.e72c00cf",
+      "version": "0.11.3",
      "dependencies": {
        "@a2a-js/sdk": "^0.3.2",
        "@google-cloud/storage": "^7.16.0",
@@ -17958,7 +17943,7 @@
    },
    "packages/cli": {
      "name": "@google/gemini-cli",
-      "version": "0.11.0-nightly.20251021.e72c00cf",
+      "version": "0.11.3",
      "dependencies": {
        "@google/gemini-cli-core": "file:../core",
        "@google/genai": "1.16.0",
@@ -18071,7 +18056,7 @@
    },
    "packages/core": {
      "name": "@google/gemini-cli-core",
-      "version": "0.11.0-nightly.20251021.e72c00cf",
+      "version": "0.11.3",
      "dependencies": {
        "@google-cloud/logging": "^11.2.1",
        "@google-cloud/opentelemetry-cloud-monitoring-exporter": "^0.21.0",
@@ -18203,7 +18188,6 @@
      "resolved": "https://registry.npmjs.org/picomatch/-/picomatch-4.0.3.tgz",
      "integrity": "sha512-5gTmgEY/sqK6gFXLIsQNH19lWb4ebPDLA4SdLP7dsWkIXHWlG66oPuVvXSGFPppYZz8ZDZq0dYYrbHfBCVUb1Q==",
      "license": "MIT",
-      "peer": true,
      "engines": {
        "node": ">=12"
      },
@@ -18213,7 +18197,7 @@
    },
    "packages/test-utils": {
      "name": "@google/gemini-cli-test-utils",
-      "version": "0.11.0-nightly.20251021.e72c00cf",
+      "version": "0.11.3",
      "license": "Apache-2.0",
      "devDependencies": {
        "typescript": "^5.3.3"
@@ -18224,7 +18208,7 @@
    },
    "packages/vscode-ide-companion": {
      "name": "gemini-cli-vscode-ide-companion",
-      "version": "0.11.0-nightly.20251021.e72c00cf",
+      "version": "0.11.3",
      "license": "LICENSE",
      "dependencies": {
        "@modelcontextprotocol/sdk": "^1.15.1",
@@ -1,6 +1,6 @@
 {
  "name": "@google/gemini-cli",
-  "version": "0.11.0-nightly.20251021.e72c00cf",
+  "version": "0.11.3",
  "engines": {
    "node": ">=20.0.0"
  },
@@ -14,7 +14,7 @@
    "url": "git+https://github.com/google-gemini/gemini-cli.git"
  },
  "config": {
-    "sandboxImageUri": "us-docker.pkg.dev/gemini-code-dev/gemini-cli/sandbox:0.11.0-nightly.20251021.e72c00cf"
+    "sandboxImageUri": "us-docker.pkg.dev/gemini-code-dev/gemini-cli/sandbox:0.11.3"
  },
  "scripts": {
    "start": "cross-env NODE_ENV=development node scripts/start.js",
@@ -1,6 +1,6 @@
 {
  "name": "@google/gemini-cli-a2a-server",
-  "version": "0.11.0-nightly.20251021.e72c00cf",
+  "version": "0.11.3",
  "description": "Gemini CLI A2A Server",
  "repository": {
    "type": "git",
@@ -1,6 +1,6 @@
 {
  "name": "@google/gemini-cli",
-  "version": "0.11.0-nightly.20251021.e72c00cf",
+  "version": "0.11.3",
  "description": "Gemini CLI",
  "repository": {
    "type": "git",
@@ -25,7 +25,7 @@
    "dist"
  ],
  "config": {
-    "sandboxImageUri": "us-docker.pkg.dev/gemini-code-dev/gemini-cli/sandbox:0.11.0-nightly.20251021.e72c00cf"
+    "sandboxImageUri": "us-docker.pkg.dev/gemini-code-dev/gemini-cli/sandbox:0.11.3"
  },
  "dependencies": {
    "@google/gemini-cli-core": "file:../core",
@@ -1572,7 +1572,7 @@ describe('loadCliConfig with includeDirectories', () => {
  });
 });

-describe('loadCliConfig chatCompression', () => {
+describe('loadCliConfig compressionThreshold', () => {
  const originalArgv = process.argv;

  beforeEach(() => {
@@ -1587,28 +1587,24 @@ describe('loadCliConfig chatCompression', () => {
    vi.restoreAllMocks();
  });

-  it('should pass chatCompression settings to the core config', async () => {
+  it('should pass settings to the core config', async () => {
    process.argv = ['node', 'script.js'];
    const argv = await parseArguments({} as Settings);
    const settings: Settings = {
      model: {
-        chatCompression: {
-          contextPercentageThreshold: 0.5,
-        },
+        compressionThreshold: 0.5,
      },
    };
    const config = await loadCliConfig(settings, [], 'test-session', argv);
-    expect(config.getChatCompression()).toEqual({
-      contextPercentageThreshold: 0.5,
-    });
+    expect(config.getCompressionThreshold()).toEqual(0.5);
  });

-  it('should have undefined chatCompression if not in settings', async () => {
+  it('should have undefined compressionThreshold if not in settings', async () => {
    process.argv = ['node', 'script.js'];
    const argv = await parseArguments({} as Settings);
    const settings: Settings = {};
    const config = await loadCliConfig(settings, [], 'test-session', argv);
-    expect(config.getChatCompression()).toBeUndefined();
+    expect(config.getCompressionThreshold()).toBeUndefined();
  });
 });

@@ -610,7 +610,7 @@ export async function loadCliConfig(
    noBrowser: !!process.env['NO_BROWSER'],
    summarizeToolOutput: settings.model?.summarizeToolOutput,
    ideMode,
-    chatCompression: settings.model?.chatCompression,
+    compressionThreshold: settings.model?.compressionThreshold,
    folderTrust,
    interactive,
    trustedFolder,
@@ -1051,15 +1051,15 @@ describe('Settings Loading and Merging', () => {
      });
    });

-    it('should merge chatCompression settings, with workspace taking precedence', () => {
+    it('should merge compressionThreshold settings, with workspace taking precedence', () => {
      (mockFsExistsSync as Mock).mockReturnValue(true);
      const userSettingsContent = {
        general: {},
-        model: { chatCompression: { contextPercentageThreshold: 0.5 } },
+        model: { compressionThreshold: 0.5 },
      };
      const workspaceSettingsContent = {
        general: {},
-        model: { chatCompression: { contextPercentageThreshold: 0.8 } },
+        model: { compressionThreshold: 0.8 },
      };

      (fs.readFileSync as Mock).mockImplementation(
@@ -1074,15 +1074,11 @@ describe('Settings Loading and Merging', () => {

      const settings = loadSettings(MOCK_WORKSPACE_DIR);

-      expect(settings.user.settings.model?.chatCompression).toEqual({
-        contextPercentageThreshold: 0.5,
-      });
-      expect(settings.workspace.settings.model?.chatCompression).toEqual({
-        contextPercentageThreshold: 0.8,
-      });
-      expect(settings.merged.model?.chatCompression).toEqual({
-        contextPercentageThreshold: 0.8,
-      });
+      expect(settings.user.settings.model?.compressionThreshold).toEqual(0.5);
+      expect(settings.workspace.settings.model?.compressionThreshold).toEqual(
+        0.8,
+      );
+      expect(settings.merged.model?.compressionThreshold).toEqual(0.8);
    });

    it('should merge output format settings, with workspace taking precedence', () => {
@@ -1109,13 +1105,13 @@ describe('Settings Loading and Merging', () => {
      expect(settings.merged.output?.format).toBe('json');
    });

-    it('should handle chatCompression when only in user settings', () => {
+    it('should handle compressionThreshold when only in user settings', () => {
      (mockFsExistsSync as Mock).mockImplementation(
        (p: fs.PathLike) => p === USER_SETTINGS_PATH,
      );
      const userSettingsContent = {
        general: {},
-        model: { chatCompression: { contextPercentageThreshold: 0.5 } },
+        model: { compressionThreshold: 0.5 },
      };
      (fs.readFileSync as Mock).mockImplementation(
        (p: fs.PathOrFileDescriptor) => {
@@ -1126,9 +1122,7 @@ describe('Settings Loading and Merging', () => {
      );

      const settings = loadSettings(MOCK_WORKSPACE_DIR);
-      expect(settings.merged.model?.chatCompression).toEqual({
-        contextPercentageThreshold: 0.5,
-      });
+      expect(settings.merged.model?.compressionThreshold).toEqual(0.5);
    });

    it('should have model as undefined if not in any settings file', () => {
@@ -1138,39 +1132,15 @@ describe('Settings Loading and Merging', () => {
      expect(settings.merged.model).toBeUndefined();
    });

-    it('should ignore chatCompression if contextPercentageThreshold is invalid', () => {
-      const warnSpy = vi.spyOn(console, 'warn').mockImplementation(() => {});
-      (mockFsExistsSync as Mock).mockImplementation(
-        (p: fs.PathLike) => p === USER_SETTINGS_PATH,
-      );
-      const userSettingsContent = {
-        general: {},
-        model: { chatCompression: { contextPercentageThreshold: 1.5 } },
-      };
-      (fs.readFileSync as Mock).mockImplementation(
-        (p: fs.PathOrFileDescriptor) => {
-          if (p === USER_SETTINGS_PATH)
-            return JSON.stringify(userSettingsContent);
-          return '{}';
-        },
-      );
-
-      const settings = loadSettings(MOCK_WORKSPACE_DIR);
-      expect(settings.merged.model?.chatCompression).toEqual({
-        contextPercentageThreshold: 1.5,
-      });
-      warnSpy.mockRestore();
-    });
-
-    it('should deep merge chatCompression settings', () => {
+    it('should use user compressionThreshold if workspace does not define it', () => {
      (mockFsExistsSync as Mock).mockReturnValue(true);
      const userSettingsContent = {
        general: {},
-        model: { chatCompression: { contextPercentageThreshold: 0.5 } },
+        model: { compressionThreshold: 0.5 },
      };
      const workspaceSettingsContent = {
        general: {},
-        model: { chatCompression: {} },
+        model: {},
      };

      (fs.readFileSync as Mock).mockImplementation(
@@ -1185,9 +1155,7 @@ describe('Settings Loading and Merging', () => {

      const settings = loadSettings(MOCK_WORKSPACE_DIR);

-      expect(settings.merged.model?.chatCompression).toEqual({
-        contextPercentageThreshold: 0.5,
-      });
+      expect(settings.merged.model?.compressionThreshold).toEqual(0.5);
    });

    it('should merge includeDirectories from all scopes', () => {
@@ -1972,9 +1940,6 @@ describe('Settings Loading and Merging', () => {
        },
        model: {
          name: 'gemini-pro',
-          chatCompression: {
-            contextPercentageThreshold: 0.5,
-          },
        },
        mcpServers: {
          'server-1': {
@@ -1993,9 +1958,6 @@ describe('Settings Loading and Merging', () => {
          myTheme: {},
        },
        model: 'gemini-pro',
-        chatCompression: {
-          contextPercentageThreshold: 0.5,
-        },
        mcpServers: {
          'server-1': {
            command: 'node server.js',
@@ -2035,9 +1997,6 @@ describe('Settings Loading and Merging', () => {
        },
        model: {
          name: 'gemini-pro',
-          chatCompression: {
-            contextPercentageThreshold: 0.8,
-          },
        },
        context: {
          fileName: 'CONTEXT.md',
@@ -2077,9 +2036,6 @@ describe('Settings Loading and Merging', () => {
        theme: 'dark',
        usageStatisticsEnabled: false,
        model: 'gemini-pro',
-        chatCompression: {
-          contextPercentageThreshold: 0.8,
-        },
        contextFileName: 'CONTEXT.md',
        includeDirectories: ['/src'],
        sandbox: true,
@@ -64,7 +64,7 @@ const MIGRATION_MAP: Record<string, string> = {
  autoAccept: 'tools.autoAccept',
  autoConfigureMaxOldSpaceSize: 'advanced.autoConfigureMemory',
  bugCommand: 'advanced.bugCommand',
-  chatCompression: 'model.chatCompression',
+  chatCompression: 'model.compressionThreshold',
  checkpointing: 'general.checkpointing',
  coreTools: 'tools.core',
  contextFileName: 'context.fileName',
@@ -159,6 +159,10 @@ describe('SettingsSchema', () => {
      expect(
        getSettingsSchema().ui.properties.showMemoryUsage.showInDialog,
      ).toBe(true);
+      expect(
+        getSettingsSchema().ui.properties.footer.properties
+          .hideContextPercentage.showInDialog,
+      ).toBe(true);
      expect(getSettingsSchema().general.properties.vimMode.showInDialog).toBe(
        true,
      );
@@ -14,7 +14,6 @@ import type {
  BugCommandSettings,
  TelemetrySettings,
  AuthType,
-  ChatCompressionSettings,
 } from '@google/gemini-cli-core';
 import {
  DEFAULT_TRUNCATE_TOOL_OUTPUT_LINES,
@@ -389,6 +388,15 @@ const SETTINGS_SCHEMA = {
            description: 'Hide the model name and context usage in the footer.',
            showInDialog: true,
          },
+          hideContextPercentage: {
+            type: 'boolean',
+            label: 'Hide Context Window Percentage',
+            category: 'UI',
+            requiresRestart: false,
+            default: true,
+            description: 'Hides the context window remaining percentage.',
+            showInDialog: true,
+          },
        },
      },
      hideFooter: {
@@ -578,14 +586,15 @@ const SETTINGS_SCHEMA = {
        description: 'Settings for summarizing tool output.',
        showInDialog: false,
      },
-      chatCompression: {
-        type: 'object',
-        label: 'Chat Compression',
+      compressionThreshold: {
+        type: 'number',
+        label: 'Compression Threshold',
        category: 'Model',
-        requiresRestart: false,
-        default: undefined as ChatCompressionSettings | undefined,
-        description: 'Chat compression settings.',
-        showInDialog: false,
+        requiresRestart: true,
+        default: 0.2 as number,
+        description:
+          'The fraction of context usage at which to trigger context compression (e.g. 0.2, 0.3).',
+        showInDialog: true,
      },
      skipNextSpeakerCheck: {
        type: 'boolean',
@@ -4,6 +4,7 @@
 * SPDX-License-Identifier: Apache-2.0
 */

+import { describe, it, expect, vi } from 'vitest';
 import {
  renderWithProviders,
  createMockSettings,
@@ -120,6 +121,13 @@ describe('<Footer />', () => {
    const { lastFrame } = renderWithProviders(<Footer />, {
      width: 120,
      uiState: { sessionStats: mockSessionStats },
+      settings: createMockSettings({
+        ui: {
+          footer: {
+            hideContextPercentage: false,
+          },
+        },
+      }),
    });
    expect(lastFrame()).toContain(defaultProps.model);
    expect(lastFrame()).toMatch(/\(\d+% context left\)/);
@@ -129,6 +137,13 @@ describe('<Footer />', () => {
    const { lastFrame } = renderWithProviders(<Footer />, {
      width: 99,
      uiState: { sessionStats: mockSessionStats },
+      settings: createMockSettings({
+        ui: {
+          footer: {
+            hideContextPercentage: false,
+          },
+        },
+      }),
    });
    expect(lastFrame()).toContain(defaultProps.model);
    expect(lastFrame()).toMatch(/\(\d+%\)/);
@@ -192,6 +207,13 @@ describe('<Footer />', () => {
      const { lastFrame } = renderWithProviders(<Footer />, {
        width: 120,
        uiState: { sessionStats: mockSessionStats },
+        settings: createMockSettings({
+          ui: {
+            footer: {
+              hideContextPercentage: false,
+            },
+          },
+        }),
      });
      expect(lastFrame()).toMatchSnapshot('complete-footer-wide');
    });
@@ -247,10 +269,49 @@ describe('<Footer />', () => {
      expect(lastFrame()).toMatchSnapshot('footer-only-sandbox');
    });

+    it('hides the context percentage when hideContextPercentage is true', () => {
+      const { lastFrame } = renderWithProviders(<Footer />, {
+        width: 120,
+        uiState: { sessionStats: mockSessionStats },
+        settings: createMockSettings({
+          ui: {
+            footer: {
+              hideContextPercentage: true,
+            },
+          },
+        }),
+      });
+      expect(lastFrame()).toContain(defaultProps.model);
+      expect(lastFrame()).not.toMatch(/\(\d+% context left\)/);
+    });
+
+    it('shows the context percentage when hideContextPercentage is false', () => {
+      const { lastFrame } = renderWithProviders(<Footer />, {
+        width: 120,
+        uiState: { sessionStats: mockSessionStats },
+        settings: createMockSettings({
+          ui: {
+            footer: {
+              hideContextPercentage: false,
+            },
+          },
+        }),
+      });
+      expect(lastFrame()).toContain(defaultProps.model);
+      expect(lastFrame()).toMatch(/\(\d+% context left\)/);
+    });
+
    it('renders complete footer in narrow terminal (baseline narrow)', () => {
      const { lastFrame } = renderWithProviders(<Footer />, {
        width: 79,
        uiState: { sessionStats: mockSessionStats },
+        settings: createMockSettings({
+          ui: {
+            footer: {
+              hideContextPercentage: false,
+            },
+          },
+        }),
      });
      expect(lastFrame()).toMatchSnapshot('complete-footer-narrow');
    });
@@ -61,6 +61,8 @@ export const Footer: React.FC = () => {
  const hideSandboxStatus =
    settings.merged.ui?.footer?.hideSandboxStatus || false;
  const hideModelInfo = settings.merged.ui?.footer?.hideModelInfo || false;
+  const hideContextPercentage =
+    settings.merged.ui?.footer?.hideContextPercentage ?? true;

  const pathLength = Math.max(20, Math.floor(mainAreaWidth * 0.25));
  const displayPath = shortenPath(tildeifyPath(targetDir), pathLength);
@@ -146,12 +148,17 @@ export const Footer: React.FC = () => {
        <Box alignItems="center" justifyContent="flex-end">
          <Box alignItems="center">
            <Text color={theme.text.accent}>
-              {model}{' '}
-              <ContextUsageDisplay
-                promptTokenCount={promptTokenCount}
-                model={model}
-                terminalWidth={mainAreaWidth}
-              />
+              {model}
+              {!hideContextPercentage && (
+                <>
+                  {' '}
+                  <ContextUsageDisplay
+                    promptTokenCount={promptTokenCount}
+                    model={model}
+                    terminalWidth={mainAreaWidth}
+                  />
+                </>
+              )}
            </Text>
            {showMemoryUsage && <MemoryUsageDisplay />}
          </Box>
@@ -19,25 +19,15 @@ import {
  type FallbackModelHandler,
  UserTierId,
  AuthType,
-  isGenericQuotaExceededError,
-  isProQuotaExceededError,
+  TerminalQuotaError,
  makeFakeConfig,
+  type GoogleApiError,
+  RetryableQuotaError,
 } from '@google/gemini-cli-core';
 import { useQuotaAndFallback } from './useQuotaAndFallback.js';
 import type { UseHistoryManagerReturn } from './useHistoryManager.js';
 import { AuthState, MessageType } from '../types.js';

-// Mock the error checking functions from the core package to control test scenarios
-vi.mock('@google/gemini-cli-core', async (importOriginal) => {
-  const original =
-    await importOriginal<typeof import('@google/gemini-cli-core')>();
-  return {
-    ...original,
-    isGenericQuotaExceededError: vi.fn(),
-    isProQuotaExceededError: vi.fn(),
-  };
-});
-
 // Use a type alias for SpyInstance as it's not directly exported
 type SpyInstance = ReturnType<typeof vi.spyOn>;

@@ -47,12 +37,15 @@ describe('useQuotaAndFallback', () => {
  let mockSetAuthState: Mock;
  let mockSetModelSwitchedFromQuotaError: Mock;
  let setFallbackHandlerSpy: SpyInstance;
-
-  const mockedIsGenericQuotaExceededError = isGenericQuotaExceededError as Mock;
-  const mockedIsProQuotaExceededError = isProQuotaExceededError as Mock;
+  let mockGoogleApiError: GoogleApiError;

  beforeEach(() => {
    mockConfig = makeFakeConfig();
+    mockGoogleApiError = {
+      code: 429,
+      message: 'mock error',
+      details: [],
+    };

    // Spy on the method that requires the private field and mock its return.
    // This is cleaner than modifying the config class for tests.
@@ -72,9 +65,6 @@ describe('useQuotaAndFallback', () => {

    setFallbackHandlerSpy = vi.spyOn(mockConfig, 'setFallbackModelHandler');
    vi.spyOn(mockConfig, 'setQuotaErrorOccurred');
-
-    mockedIsGenericQuotaExceededError.mockReturnValue(false);
-    mockedIsProQuotaExceededError.mockReturnValue(false);
  });

  afterEach(() => {
@@ -140,51 +130,62 @@ describe('useQuotaAndFallback', () => {
    describe('Automatic Fallback Scenarios', () => {
      const testCases = [
        {
-          errorType: 'generic',
+          description: 'other error for FREE tier',
          tier: UserTierId.FREE,
+          error: new Error('some error'),
          expectedMessageSnippets: [
-            'Automatically switching from model-A to model-B',
+            'Automatically switching from model-A to model-B for faster responses',
            'upgrade to a Gemini Code Assist Standard or Enterprise plan',
          ],
        },
        {
-          errorType: 'generic',
-          tier: UserTierId.STANDARD, // Paid tier
+          description: 'other error for LEGACY tier',
+          tier: UserTierId.LEGACY, // Paid tier
+          error: new Error('some error'),
          expectedMessageSnippets: [
-            'Automatically switching from model-A to model-B',
+            'Automatically switching from model-A to model-B for faster responses',
            'switch to using a paid API key from AI Studio',
          ],
        },
        {
-          errorType: 'other',
+          description: 'retryable quota error for FREE tier',
          tier: UserTierId.FREE,
+          error: new RetryableQuotaError(
+            'retryable quota',
+            mockGoogleApiError,
+            5,
+          ),
          expectedMessageSnippets: [
-            'Automatically switching from model-A to model-B for faster responses',
-            'upgrade to a Gemini Code Assist Standard or Enterprise plan',
+            'Your requests are being throttled right now due to server being at capacity for model-A',
+            'Automatically switching from model-A to model-B',
+            'upgrading to a Gemini Code Assist Standard or Enterprise plan',
          ],
        },
        {
-          errorType: 'other',
+          description: 'retryable quota error for LEGACY tier',
          tier: UserTierId.LEGACY, // Paid tier
+          error: new RetryableQuotaError(
+            'retryable quota',
+            mockGoogleApiError,
+            5,
+          ),
          expectedMessageSnippets: [
-            'Automatically switching from model-A to model-B for faster responses',
+            'Your requests are being throttled right now due to server being at capacity for model-A',
+            'Automatically switching from model-A to model-B',
            'switch to using a paid API key from AI Studio',
          ],
        },
      ];

-      for (const { errorType, tier, expectedMessageSnippets } of testCases) {
-        it(`should handle ${errorType} error for ${tier} tier correctly`, async () => {
-          mockedIsGenericQuotaExceededError.mockReturnValue(
-            errorType === 'generic',
-          );
-
+      for (const {
+        description,
+        tier,
+        error,
+        expectedMessageSnippets,
+      } of testCases) {
+        it(`should handle ${description} correctly`, async () => {
          const handler = getRegisteredHandler(tier);
-          const result = await handler(
-            'model-A',
-            'model-B',
-            new Error('quota exceeded'),
-          );
+          const result = await handler('model-A', 'model-B', error);

          // Automatic fallbacks should return 'stop'
          expect(result).toBe('stop');
@@ -207,10 +208,6 @@ describe('useQuotaAndFallback', () => {
    });

    describe('Interactive Fallback (Pro Quota Error)', () => {
-      beforeEach(() => {
-        mockedIsProQuotaExceededError.mockReturnValue(true);
-      });
-
      it('should set an interactive request and wait for user choice', async () => {
        const { result } = renderHook(() =>
          useQuotaAndFallback({
@@ -229,7 +226,7 @@ describe('useQuotaAndFallback', () => {
        const promise = handler(
          'gemini-pro',
          'gemini-flash',
-          new Error('pro quota'),
+          new TerminalQuotaError('pro quota', mockGoogleApiError),
        );

        await act(async () => {});
@@ -268,7 +265,7 @@ describe('useQuotaAndFallback', () => {
        const promise1 = handler(
          'gemini-pro',
          'gemini-flash',
-          new Error('pro quota 1'),
+          new TerminalQuotaError('pro quota 1', mockGoogleApiError),
        );
        await act(async () => {});

@@ -278,7 +275,7 @@ describe('useQuotaAndFallback', () => {
        const result2 = await handler(
          'gemini-pro',
          'gemini-flash',
-          new Error('pro quota 2'),
+          new TerminalQuotaError('pro quota 2', mockGoogleApiError),
        );

        // The lock should have stopped the second request
@@ -297,10 +294,6 @@ describe('useQuotaAndFallback', () => {
  });

  describe('handleProQuotaChoice', () => {
-    beforeEach(() => {
-      mockedIsProQuotaExceededError.mockReturnValue(true);
-    });
-
    it('should do nothing if there is no pending pro quota request', () => {
      const { result } = renderHook(() =>
        useQuotaAndFallback({
@@ -336,7 +329,7 @@ describe('useQuotaAndFallback', () => {
      const promise = handler(
        'gemini-pro',
        'gemini-flash',
-        new Error('pro quota'),
+        new TerminalQuotaError('pro quota', mockGoogleApiError),
      );
      await act(async () => {}); // Allow state to update

@@ -367,7 +360,7 @@ describe('useQuotaAndFallback', () => {
      const promise = handler(
        'gemini-pro',
        'gemini-flash',
-        new Error('pro quota'),
+        new TerminalQuotaError('pro quota', mockGoogleApiError),
      );
      await act(async () => {}); // Allow state to update

@@ -9,9 +9,9 @@ import {
  type Config,
  type FallbackModelHandler,
  type FallbackIntent,
-  isGenericQuotaExceededError,
-  isProQuotaExceededError,
+  TerminalQuotaError,
  UserTierId,
+  RetryableQuotaError,
 } from '@google/gemini-cli-core';
 import { useCallback, useEffect, useRef, useState } from 'react';
 import { type UseHistoryManagerReturn } from './useHistoryManager.js';
@@ -63,7 +63,7 @@ export function useQuotaAndFallback({

      let message: string;

-      if (error && isProQuotaExceededError(error)) {
+      if (error instanceof TerminalQuotaError) {
        // Pro Quota specific messages (Interactive)
        if (isPaidTier) {
          message = `⚡ You have reached your daily ${failedModel} quota limit.
@@ -76,31 +76,30 @@ export function useQuotaAndFallback({
 ⚡ Or you can utilize a Gemini API Key. See: https://goo.gle/gemini-cli-docs-auth#gemini-api-key
 ⚡ You can switch authentication methods by typing /auth`;
        }
-      } else if (error && isGenericQuotaExceededError(error)) {
-        // Generic Quota (Automatic fallback)
-        const actionMessage = `⚡ You have reached your daily quota limit.\n⚡ Automatically switching from ${failedModel} to ${fallbackModel} for the remainder of this session.`;
+      } else if (error instanceof RetryableQuotaError) {
+        // Short term quota retries exhausted (Automatic fallback)
+        const actionMessage = `⚡ Your requests are being throttled right now due to server being at capacity for ${failedModel}.\n⚡ Automatically switching from ${failedModel} to ${fallbackModel} for the remainder of this session.`;

        if (isPaidTier) {
          message = `${actionMessage}
-⚡ To continue accessing the ${failedModel} model today, consider using /auth to switch to using a paid API key from AI Studio at https://aistudio.google.com/apikey`;
+⚡ To continue accessing the ${failedModel} model, retry your request after some time or consider using /auth to switch to using a paid API key from AI Studio at https://aistudio.google.com/apikey`;
        } else {
          message = `${actionMessage}
-⚡ To increase your limits, upgrade to a Gemini Code Assist Standard or Enterprise plan with higher limits at https://goo.gle/set-up-gemini-code-assist
-⚡ Or you can utilize a Gemini API Key. See: https://goo.gle/gemini-cli-docs-auth#gemini-api-key
+⚡ Retry your requests after some time. Otherwise consider upgrading to a Gemini Code Assist Standard or Enterprise plan with higher limits at https://goo.gle/set-up-gemini-code-assist
 ⚡ You can switch authentication methods by typing /auth`;
        }
      } else {
-        // Consecutive 429s or other errors (Automatic fallback)
+        // Other errors (Automatic fallback)
        const actionMessage = `⚡ Automatically switching from ${failedModel} to ${fallbackModel} for faster responses for the remainder of this session.`;

        if (isPaidTier) {
          message = `${actionMessage}
-⚡ Possible reasons for this are that you have received multiple consecutive capacity errors or you have reached your daily ${failedModel} quota limit
-⚡ To continue accessing the ${failedModel} model today, consider using /auth to switch to using a paid API key from AI Studio at https://aistudio.google.com/apikey`;
+⚡ Your requests are being throttled temporarily due to server being at capacity for ${failedModel} or there is a service outage.
+⚡ To continue accessing the ${failedModel} model, you can retry your request after some time or consider using /auth to switch to using a paid API key from AI Studio at https://aistudio.google.com/apikey`;
        } else {
          message = `${actionMessage}
-⚡ Possible reasons for this are that you have received multiple consecutive capacity errors or you have reached your daily ${failedModel} quota limit
-⚡ To increase your limits, upgrade to a Gemini Code Assist Standard or Enterprise plan with higher limits at https://goo.gle/set-up-gemini-code-assist
+⚡ Your requests are being throttled temporarily due to server being at capacity for ${failedModel} or there is a service outage.
+⚡ To avoid being throttled, you can retry your request after some time or upgrade to a Gemini Code Assist Standard or Enterprise plan with higher limits at https://goo.gle/set-up-gemini-code-assist
 ⚡ Or you can utilize a Gemini API Key. See: https://goo.gle/gemini-cli-docs-auth#gemini-api-key
 ⚡ You can switch authentication methods by typing /auth`;
        }
@@ -119,7 +118,7 @@ export function useQuotaAndFallback({
      config.setQuotaErrorOccurred(true);

      // Interactive Fallback for Pro quota
-      if (error && isProQuotaExceededError(error)) {
+      if (error instanceof TerminalQuotaError) {
        if (isDialogPending.current) {
          return 'stop'; // A dialog is already active, so just stop this request.
        }
@@ -44,3 +44,5 @@ export { makeFakeConfig } from './src/test-utils/config.js';
 export * from './src/utils/pathReader.js';
 export { ClearcutLogger } from './src/telemetry/clearcut-logger/clearcut-logger.js';
 export { logModelSlashCommand } from './src/telemetry/loggers.js';
+export * from './src/utils/googleQuotaErrors.js';
+export type { GoogleApiError } from './src/utils/googleErrors.js';
@@ -1,6 +1,6 @@
 {
  "name": "@google/gemini-cli-core",
-  "version": "0.11.0-nightly.20251021.e72c00cf",
+  "version": "0.11.3",
  "description": "Gemini CLI Core",
  "repository": {
    "type": "git",
@@ -574,29 +574,84 @@ describe('Server Config (config.ts)', () => {
    });
  });

-  describe('UseModelRouter Configuration', () => {
-    it('should default useModelRouter to false when not provided', () => {
-      const config = new Config(baseParams);
-      expect(config.getUseModelRouter()).toBe(false);
-    });
-
-    it('should set useModelRouter to true when provided as true', () => {
-      const paramsWithModelRouter: ConfigParameters = {
+  describe('Model Router with Auth', () => {
+    it('should disable model router by default for oauth-personal', async () => {
+      const config = new Config({
        ...baseParams,
        useModelRouter: true,
-      };
-      const config = new Config(paramsWithModelRouter);
+      });
+      await config.refreshAuth(AuthType.LOGIN_WITH_GOOGLE);
      expect(config.getUseModelRouter()).toBe(true);
    });

-    it('should set useModelRouter to false when explicitly provided as false', () => {
-      const paramsWithModelRouter: ConfigParameters = {
+    it('should enable model router by default for other auth types', async () => {
+      const config = new Config({
+        ...baseParams,
+        useModelRouter: true,
+      });
+      await config.refreshAuth(AuthType.USE_GEMINI);
+      expect(config.getUseModelRouter()).toBe(true);
+    });
+
+    it('should disable model router for specified auth type', async () => {
+      const config = new Config({
+        ...baseParams,
+        useModelRouter: true,
+        disableModelRouterForAuth: [AuthType.USE_GEMINI],
+      });
+      await config.refreshAuth(AuthType.USE_GEMINI);
+      expect(config.getUseModelRouter()).toBe(false);
+    });
+
+    it('should enable model router for other auth type', async () => {
+      const config = new Config({
+        ...baseParams,
+        useModelRouter: true,
+        disableModelRouterForAuth: [],
+      });
+      await config.refreshAuth(AuthType.LOGIN_WITH_GOOGLE);
+      expect(config.getUseModelRouter()).toBe(true);
+    });
+
+    it('should keep model router disabled when useModelRouter is false', async () => {
+      const config = new Config({
        ...baseParams,
        useModelRouter: false,
-      };
-      const config = new Config(paramsWithModelRouter);
+        disableModelRouterForAuth: [AuthType.USE_GEMINI],
+      });
+      await config.refreshAuth(AuthType.LOGIN_WITH_GOOGLE);
      expect(config.getUseModelRouter()).toBe(false);
    });
+
+    it('should keep the user-chosen model after refreshAuth, even when model router is disabled for the auth type', async () => {
+      const config = new Config({
+        ...baseParams,
+        useModelRouter: true,
+        disableModelRouterForAuth: [AuthType.USE_GEMINI],
+      });
+      const chosenModel = 'gemini-1.5-pro-latest';
+      config.setModel(chosenModel);
+
+      await config.refreshAuth(AuthType.USE_GEMINI);
+
+      expect(config.getUseModelRouter()).toBe(false);
+      expect(config.getModel()).toBe(chosenModel);
+    });
+
+    it('should keep the user-chosen model after refreshAuth, when model router is enabled for the auth type', async () => {
+      const config = new Config({
+        ...baseParams,
+        useModelRouter: true,
+        disableModelRouterForAuth: [AuthType.USE_GEMINI],
+      });
+      const chosenModel = 'gemini-1.5-pro-latest';
+      config.setModel(chosenModel);
+
+      await config.refreshAuth(AuthType.LOGIN_WITH_GOOGLE);
+
+      expect(config.getUseModelRouter()).toBe(true);
+      expect(config.getModel()).toBe(chosenModel);
+    });
  });

  describe('ContinueOnFailedApiCall Configuration', () => {
@@ -47,6 +47,7 @@ import {
  DEFAULT_GEMINI_EMBEDDING_MODEL,
  DEFAULT_GEMINI_FLASH_MODEL,
  DEFAULT_GEMINI_MODEL,
+  DEFAULT_GEMINI_MODEL_AUTO,
  DEFAULT_THINKING_MODE,
 } from './models.js';
 import { shouldAttemptBrowserLaunch } from '../utils/browser.js';
@@ -95,10 +96,6 @@ export interface BugCommandSettings {
  urlTemplate: string;
 }

-export interface ChatCompressionSettings {
-  contextPercentageThreshold?: number;
-}
-
 export interface SummarizeToolOutputSettings {
  tokenBudget?: number;
 }
@@ -260,7 +257,7 @@ export interface ConfigParameters {
  folderTrust?: boolean;
  ideMode?: boolean;
  loadMemoryFromIncludeDirectories?: boolean;
-  chatCompression?: ChatCompressionSettings;
+  compressionThreshold?: number;
  interactive?: boolean;
  trustedFolder?: boolean;
  useRipgrep?: boolean;
@@ -279,6 +276,7 @@ export interface ConfigParameters {
  output?: OutputSettings;
  useModelRouter?: boolean;
  enableMessageBusIntegration?: boolean;
+  disableModelRouterForAuth?: AuthType[];
  codebaseInvestigatorSettings?: CodebaseInvestigatorSettings;
  continueOnFailedApiCall?: boolean;
  retryFetchErrors?: boolean;
@@ -352,7 +350,7 @@ export class Config {
    | undefined;
  private readonly experimentalZedIntegration: boolean = false;
  private readonly loadMemoryFromIncludeDirectories: boolean = false;
-  private readonly chatCompression: ChatCompressionSettings | undefined;
+  private readonly compressionThreshold: number | undefined;
  private readonly interactive: boolean;
  private readonly ptyInfo: string;
  private readonly trustedFolder: boolean | undefined;
@@ -374,7 +372,9 @@ export class Config {
  private readonly messageBus: MessageBus;
  private readonly policyEngine: PolicyEngine;
  private readonly outputSettings: OutputSettings;
-  private readonly useModelRouter: boolean;
+  private useModelRouter: boolean;
+  private readonly initialUseModelRouter: boolean;
+  private readonly disableModelRouterForAuth?: AuthType[];
  private readonly enableMessageBusIntegration: boolean;
  private readonly codebaseInvestigatorSettings: CodebaseInvestigatorSettings;
  private readonly continueOnFailedApiCall: boolean;
@@ -449,7 +449,7 @@ export class Config {
    this.ideMode = params.ideMode ?? false;
    this.loadMemoryFromIncludeDirectories =
      params.loadMemoryFromIncludeDirectories ?? false;
-    this.chatCompression = params.chatCompression;
+    this.compressionThreshold = params.compressionThreshold;
    this.interactive = params.interactive ?? false;
    this.ptyInfo = params.ptyInfo ?? 'child_process';
    this.trustedFolder = params.trustedFolder;
@@ -470,7 +470,9 @@ export class Config {
    this.enableToolOutputTruncation = params.enableToolOutputTruncation ?? true;
    this.useSmartEdit = params.useSmartEdit ?? true;
    this.useWriteTodos = params.useWriteTodos ?? false;
-    this.useModelRouter = params.useModelRouter ?? false;
+    this.initialUseModelRouter = params.useModelRouter ?? false;
+    this.useModelRouter = this.initialUseModelRouter;
+    this.disableModelRouterForAuth = params.disableModelRouterForAuth ?? [];
    this.enableMessageBusIntegration =
      params.enableMessageBusIntegration ?? false;
    this.codebaseInvestigatorSettings = {
@@ -541,6 +543,16 @@ export class Config {
  }

  async refreshAuth(authMethod: AuthType) {
+    this.useModelRouter = this.initialUseModelRouter;
+    if (this.disableModelRouterForAuth?.includes(authMethod)) {
+      this.useModelRouter = false;
+      if (this.model === DEFAULT_GEMINI_MODEL_AUTO) {
+        this.model = DEFAULT_GEMINI_MODEL;
+      }
+    } else if (this.useModelRouter && this.model === DEFAULT_GEMINI_MODEL) {
+      this.model = DEFAULT_GEMINI_MODEL_AUTO;
+    }
+
    // Vertex and Genai have incompatible encryption and sending history with
    // thoughtSignature from Genai to Vertex will fail, we need to strip them
    if (
@@ -961,8 +973,8 @@ export class Config {
    this.fileSystemService = fileSystemService;
  }

-  getChatCompression(): ChatCompressionSettings | undefined {
-    return this.chatCompression;
+  getCompressionThreshold(): number | undefined {
+    return this.compressionThreshold;
  }

  isInteractiveShellEnabled(): boolean {
@@ -16,7 +16,6 @@ import {

 import type { Content, GenerateContentResponse, Part } from '@google/genai';
 import {
-  findCompressSplitPoint,
  isThinkingDefault,
  isThinkingSupported,
  GeminiClient,
@@ -40,9 +39,11 @@ import { FileDiscoveryService } from '../services/fileDiscoveryService.js';
 import { setSimulate429 } from '../utils/testUtils.js';
 import { tokenLimit } from './tokenLimits.js';
 import { ideContextStore } from '../ide/ideContext.js';
-import { ClearcutLogger } from '../telemetry/clearcut-logger/clearcut-logger.js';
 import type { ModelRouterService } from '../routing/modelRouterService.js';
 import { uiTelemetryService } from '../telemetry/uiTelemetry.js';
+import { ChatCompressionService } from '../services/chatCompressionService.js';
+
+vi.mock('../services/chatCompressionService.js');

 // Mock fs module to prevent actual file system operations during tests
 const mockFileSystem = new Map<string, string>();
@@ -132,83 +133,6 @@ async function fromAsync<T>(promise: AsyncGenerator<T>): Promise<readonly T[]> {
  return results;
 }

-describe('findCompressSplitPoint', () => {
-  it('should throw an error for non-positive numbers', () => {
-    expect(() => findCompressSplitPoint([], 0)).toThrow(
-      'Fraction must be between 0 and 1',
-    );
-  });
-
-  it('should throw an error for a fraction greater than or equal to 1', () => {
-    expect(() => findCompressSplitPoint([], 1)).toThrow(
-      'Fraction must be between 0 and 1',
-    );
-  });
-
-  it('should handle an empty history', () => {
-    expect(findCompressSplitPoint([], 0.5)).toBe(0);
-  });
-
-  it('should handle a fraction in the middle', () => {
-    const history: Content[] = [
-      { role: 'user', parts: [{ text: 'This is the first message.' }] }, // JSON length: 66 (19%)
-      { role: 'model', parts: [{ text: 'This is the second message.' }] }, // JSON length: 68 (40%)
-      { role: 'user', parts: [{ text: 'This is the third message.' }] }, // JSON length: 66 (60%)
-      { role: 'model', parts: [{ text: 'This is the fourth message.' }] }, // JSON length: 68 (80%)
-      { role: 'user', parts: [{ text: 'This is the fifth message.' }] }, // JSON length: 65 (100%)
-    ];
-    expect(findCompressSplitPoint(history, 0.5)).toBe(4);
-  });
-
-  it('should handle a fraction of last index', () => {
-    const history: Content[] = [
-      { role: 'user', parts: [{ text: 'This is the first message.' }] }, // JSON length: 66 (19%)
-      { role: 'model', parts: [{ text: 'This is the second message.' }] }, // JSON length: 68 (40%)
-      { role: 'user', parts: [{ text: 'This is the third message.' }] }, // JSON length: 66 (60%)
-      { role: 'model', parts: [{ text: 'This is the fourth message.' }] }, // JSON length: 68 (80%)
-      { role: 'user', parts: [{ text: 'This is the fifth message.' }] }, // JSON length: 65 (100%)
-    ];
-    expect(findCompressSplitPoint(history, 0.9)).toBe(4);
-  });
-
-  it('should handle a fraction of after last index', () => {
-    const history: Content[] = [
-      { role: 'user', parts: [{ text: 'This is the first message.' }] }, // JSON length: 66 (24%%)
-      { role: 'model', parts: [{ text: 'This is the second message.' }] }, // JSON length: 68 (50%)
-      { role: 'user', parts: [{ text: 'This is the third message.' }] }, // JSON length: 66 (74%)
-      { role: 'model', parts: [{ text: 'This is the fourth message.' }] }, // JSON length: 68 (100%)
-    ];
-    expect(findCompressSplitPoint(history, 0.8)).toBe(4);
-  });
-
-  it('should return earlier splitpoint if no valid ones are after threshhold', () => {
-    const history: Content[] = [
-      { role: 'user', parts: [{ text: 'This is the first message.' }] },
-      { role: 'model', parts: [{ text: 'This is the second message.' }] },
-      { role: 'user', parts: [{ text: 'This is the third message.' }] },
-      { role: 'model', parts: [{ functionCall: {} }] },
-    ];
-    // Can't return 4 because the previous item has a function call.
-    expect(findCompressSplitPoint(history, 0.99)).toBe(2);
-  });
-
-  it('should handle a history with only one item', () => {
-    const historyWithEmptyParts: Content[] = [
-      { role: 'user', parts: [{ text: 'Message 1' }] },
-    ];
-    expect(findCompressSplitPoint(historyWithEmptyParts, 0.5)).toBe(0);
-  });
-
-  it('should handle history with weird parts', () => {
-    const historyWithEmptyParts: Content[] = [
-      { role: 'user', parts: [{ text: 'Message 1' }] },
-      { role: 'model', parts: [{ fileData: { fileUri: 'derp' } }] },
-      { role: 'user', parts: [{ text: 'Message 2' }] },
-    ];
-    expect(findCompressSplitPoint(historyWithEmptyParts, 0.5)).toBe(2);
-  });
-});
-
 describe('isThinkingSupported', () => {
  it('should return true for gemini-2.5', () => {
    expect(isThinkingSupported('gemini-2.5')).toBe(true);
@@ -252,6 +176,15 @@ describe('Gemini Client (client.ts)', () => {
    vi.resetAllMocks();
    vi.mocked(uiTelemetryService.setLastPromptTokenCount).mockClear();

+    vi.mocked(ChatCompressionService.prototype.compress).mockResolvedValue({
+      newHistory: null,
+      info: {
+        originalTokenCount: 0,
+        newTokenCount: 0,
+        compressionStatus: CompressionStatus.NOOP,
+      },
+    });
+
    mockGenerateContentFn = vi.fn().mockResolvedValue({
      candidates: [{ content: { parts: [{ text: '{"key": "value"}' }] } }],
    });
@@ -404,7 +337,8 @@ describe('Gemini Client (client.ts)', () => {
        { role: 'model', parts: [{ text: 'Long response' }] },
      ] as Content[],
      originalTokenCount = 1000,
-      summaryText = 'This is a summary.',
+      newTokenCount = 500,
+      compressionStatus = CompressionStatus.COMPRESSED,
    } = {}) {
      const mockOriginalChat: Partial<GeminiChat> = {
        getHistory: vi.fn((_curated?: boolean) => chatHistory),
@@ -416,47 +350,25 @@ describe('Gemini Client (client.ts)', () => {
        originalTokenCount,
      );

-      mockGenerateContentFn.mockResolvedValue({
-        candidates: [
-          {
-            content: {
-              role: 'model',
-              parts: [{ text: summaryText }],
-            },
-          },
-        ],
-      } as unknown as GenerateContentResponse);
-
-      // Calculate what the new history will be
-      const splitPoint = findCompressSplitPoint(chatHistory, 0.7); // 1 - 0.3
-      const historyToKeep = chatHistory.slice(splitPoint);
-
-      // This is the history that the new chat will have.
-      // It includes the default startChat history + the extra history from tryCompressChat
-      const newCompressedHistory: Content[] = [
-        // Mocked envParts + canned response from startChat
-        {
-          role: 'user',
-          parts: [{ text: 'Mocked env context' }],
-        },
-        {
-          role: 'model',
-          parts: [{ text: 'Got it. Thanks for the context!' }],
-        },
-        // extraHistory from tryCompressChat
-        {
-          role: 'user',
-          parts: [{ text: summaryText }],
-        },
-        {
-          role: 'model',
-          parts: [{ text: 'Got it. Thanks for the additional context!' }],
-        },
-        ...historyToKeep,
+      const newHistory: Content[] = [
+        { role: 'user', parts: [{ text: 'Summary' }] },
+        { role: 'model', parts: [{ text: 'Got it' }] },
      ];

+      vi.mocked(ChatCompressionService.prototype.compress).mockResolvedValue({
+        newHistory:
+          compressionStatus === CompressionStatus.COMPRESSED
+            ? newHistory
+            : null,
+        info: {
+          originalTokenCount,
+          newTokenCount,
+          compressionStatus,
+        },
+      });
+
      const mockNewChat: Partial<GeminiChat> = {
-        getHistory: vi.fn().mockReturnValue(newCompressedHistory),
+        getHistory: vi.fn().mockReturnValue(newHistory),
        setHistory: vi.fn(),
      };

@@ -464,39 +376,32 @@ describe('Gemini Client (client.ts)', () => {
        .fn()
        .mockResolvedValue(mockNewChat as GeminiChat);

-      const totalChars = newCompressedHistory.reduce(
-        (total, content) => total + JSON.stringify(content).length,
-        0,
-      );
-      const estimatedNewTokenCount = Math.floor(totalChars / 4);
-
      return {
        client,
        mockOriginalChat,
        mockNewChat,
-        estimatedNewTokenCount,
+        estimatedNewTokenCount: newTokenCount,
      };
    }

    describe('when compression inflates the token count', () => {
      it('allows compression to be forced/manual after a failure', async () => {
-        // Call 1 (Fails): Setup with a long summary to inflate tokens
-        const longSummary = 'long summary '.repeat(100);
-        const { client, estimatedNewTokenCount: inflatedTokenCount } = setup({
+        // Call 1 (Fails): Setup with inflated tokens
+        setup({
          originalTokenCount: 100,
-          summaryText: longSummary,
+          newTokenCount: 200,
+          compressionStatus:
+            CompressionStatus.COMPRESSION_FAILED_INFLATED_TOKEN_COUNT,
        });
-        expect(inflatedTokenCount).toBeGreaterThan(100); // Ensure setup is correct

        await client.tryCompressChat('prompt-id-4', false); // Fails

-        // Call 2 (Forced): Re-setup with a short summary
-        const shortSummary = 'short';
+        // Call 2 (Forced): Re-setup with compressed tokens
        const { estimatedNewTokenCount: compressedTokenCount } = setup({
          originalTokenCount: 100,
-          summaryText: shortSummary,
+          newTokenCount: 50,
+          compressionStatus: CompressionStatus.COMPRESSED,
        });
-        expect(compressedTokenCount).toBeLessThanOrEqual(100); // Ensure setup is correct

        const result = await client.tryCompressChat('prompt-id-4', true); // Forced

@@ -508,12 +413,12 @@ describe('Gemini Client (client.ts)', () => {
      });

      it('yields the result even if the compression inflated the tokens', async () => {
-        const longSummary = 'long summary '.repeat(100);
        const { client, estimatedNewTokenCount } = setup({
          originalTokenCount: 100,
-          summaryText: longSummary,
+          newTokenCount: 200,
+          compressionStatus:
+            CompressionStatus.COMPRESSION_FAILED_INFLATED_TOKEN_COUNT,
        });
-        expect(estimatedNewTokenCount).toBeGreaterThan(100); // Ensure setup is correct

        const result = await client.tryCompressChat('prompt-id-4', false);

@@ -530,12 +435,12 @@ describe('Gemini Client (client.ts)', () => {
      });

      it('does not manipulate the source chat', async () => {
-        const longSummary = 'long summary '.repeat(100);
-        const { client, mockOriginalChat, estimatedNewTokenCount } = setup({
+        const { client, mockOriginalChat } = setup({
          originalTokenCount: 100,
-          summaryText: longSummary,
+          newTokenCount: 200,
+          compressionStatus:
+            CompressionStatus.COMPRESSION_FAILED_INFLATED_TOKEN_COUNT,
        });
-        expect(estimatedNewTokenCount).toBeGreaterThan(100); // Ensure setup is correct

        await client.tryCompressChat('prompt-id-4', false);

@@ -543,45 +448,65 @@ describe('Gemini Client (client.ts)', () => {
        expect(client['chat']).toBe(mockOriginalChat);
      });

-      it('will not attempt to compress context after a failure', async () => {
-        const longSummary = 'long summary '.repeat(100);
-        const { client, estimatedNewTokenCount } = setup({
+      it.skip('will not attempt to compress context after a failure', async () => {
+        const { client } = setup({
          originalTokenCount: 100,
-          summaryText: longSummary,
+          newTokenCount: 200,
+          compressionStatus:
+            CompressionStatus.COMPRESSION_FAILED_INFLATED_TOKEN_COUNT,
        });
-        expect(estimatedNewTokenCount).toBeGreaterThan(100); // Ensure setup is correct

        await client.tryCompressChat('prompt-id-4', false); // This fails and sets hasFailedCompressionAttempt = true

+        // Mock the next call to return NOOP
+        vi.mocked(
+          ChatCompressionService.prototype.compress,
+        ).mockResolvedValueOnce({
+          newHistory: null,
+          info: {
+            originalTokenCount: 0,
+            newTokenCount: 0,
+            compressionStatus: CompressionStatus.NOOP,
+          },
+        });
+
        // This call should now be a NOOP
        const result = await client.tryCompressChat('prompt-id-5', false);

-        // generateContent (for summary) should only have been called once
-        expect(mockGenerateContentFn).toHaveBeenCalledTimes(1);
-        expect(result).toEqual({
-          compressionStatus: CompressionStatus.NOOP,
-          newTokenCount: 0,
-          originalTokenCount: 0,
-        });
+        expect(result.compressionStatus).toBe(CompressionStatus.NOOP);
+        expect(ChatCompressionService.prototype.compress).toHaveBeenCalledTimes(
+          2,
+        );
+        expect(
+          ChatCompressionService.prototype.compress,
+        ).toHaveBeenLastCalledWith(
+          expect.anything(),
+          'prompt-id-5',
+          false,
+          expect.anything(),
+          expect.anything(),
+          true, // hasFailedCompressionAttempt
+        );
      });
    });

    it('should not trigger summarization if token count is below threshold', async () => {
      const MOCKED_TOKEN_LIMIT = 1000;
-      vi.mocked(tokenLimit).mockReturnValue(MOCKED_TOKEN_LIMIT);
-      mockGetHistory.mockReturnValue([
-        { role: 'user', parts: [{ text: '...history...' }] },
-      ]);
      const originalTokenCount = MOCKED_TOKEN_LIMIT * 0.699;
-      vi.mocked(uiTelemetryService.getLastPromptTokenCount).mockReturnValue(
-        originalTokenCount,
-      );
+
+      vi.mocked(ChatCompressionService.prototype.compress).mockResolvedValue({
+        newHistory: null,
+        info: {
+          originalTokenCount,
+          newTokenCount: originalTokenCount,
+          compressionStatus: CompressionStatus.NOOP,
+        },
+      });

      const initialChat = client.getChat();
      const result = await client.tryCompressChat('prompt-id-2', false);
      const newChat = client.getChat();

-      expect(tokenLimit).toHaveBeenCalled();
      expect(result).toEqual({
        compressionStatus: CompressionStatus.NOOP,
        newTokenCount: originalTokenCount,
@@ -594,6 +519,8 @@ describe('Gemini Client (client.ts)', () => {
      const { client } = setup({
        chatHistory: [{ role: 'user', parts: [{ text: 'hi' }] }],
        originalTokenCount: 50,
+        newTokenCount: 50,
+        compressionStatus: CompressionStatus.NOOP,
      });

      const result = await client.tryCompressChat('prompt-id-noop', false);
@@ -603,337 +530,6 @@ describe('Gemini Client (client.ts)', () => {
        originalTokenCount: 50,
        newTokenCount: 50,
      });
-      expect(mockGenerateContentFn).not.toHaveBeenCalled();
-    });
-
-    it('logs a telemetry event when compressing', async () => {
-      vi.spyOn(ClearcutLogger.prototype, 'logChatCompressionEvent');
-      const MOCKED_TOKEN_LIMIT = 1000;
-      const MOCKED_CONTEXT_PERCENTAGE_THRESHOLD = 0.5;
-      vi.spyOn(client['config'], 'getChatCompression').mockReturnValue({
-        contextPercentageThreshold: MOCKED_CONTEXT_PERCENTAGE_THRESHOLD,
-      });
-      const history = [
-        { role: 'user', parts: [{ text: '...history...' }] },
-        { role: 'model', parts: [{ text: '...history...' }] },
-        { role: 'user', parts: [{ text: '...history...' }] },
-        { role: 'model', parts: [{ text: '...history...' }] },
-        { role: 'user', parts: [{ text: '...history...' }] },
-        { role: 'model', parts: [{ text: '...history...' }] },
-      ];
-      mockGetHistory.mockReturnValue(history);
-
-      const originalTokenCount =
-        MOCKED_TOKEN_LIMIT * MOCKED_CONTEXT_PERCENTAGE_THRESHOLD;
-
-      vi.mocked(uiTelemetryService.getLastPromptTokenCount).mockReturnValue(
-        originalTokenCount,
-      );
-
-      // We need to control the estimated new token count.
-      // We mock startChat to return a chat with a known history.
-      const summaryText = 'This is a summary.';
-      const splitPoint = findCompressSplitPoint(history, 0.7);
-      const historyToKeep = history.slice(splitPoint);
-      const newCompressedHistory: Content[] = [
-        { role: 'user', parts: [{ text: 'Mocked env context' }] },
-        { role: 'model', parts: [{ text: 'Got it. Thanks for the context!' }] },
-        { role: 'user', parts: [{ text: summaryText }] },
-        {
-          role: 'model',
-          parts: [{ text: 'Got it. Thanks for the additional context!' }],
-        },
-        ...historyToKeep,
-      ];
-      const mockNewChat: Partial<GeminiChat> = {
-        getHistory: vi.fn().mockReturnValue(newCompressedHistory),
-      };
-      client['startChat'] = vi
-        .fn()
-        .mockResolvedValue(mockNewChat as GeminiChat);
-
-      const totalChars = newCompressedHistory.reduce(
-        (total, content) => total + JSON.stringify(content).length,
-        0,
-      );
-      const newTokenCount = Math.floor(totalChars / 4);
-
-      // Mock the summary response from the chat
-      mockGenerateContentFn.mockResolvedValue({
-        candidates: [
-          {
-            content: {
-              role: 'model',
-              parts: [{ text: summaryText }],
-            },
-          },
-        ],
-      } as unknown as GenerateContentResponse);
-
-      await client.tryCompressChat('prompt-id-3', false);
-
-      expect(
-        ClearcutLogger.prototype.logChatCompressionEvent,
-      ).toHaveBeenCalledWith(
-        expect.objectContaining({
-          tokens_before: originalTokenCount,
-          tokens_after: newTokenCount,
-        }),
-      );
-      expect(uiTelemetryService.setLastPromptTokenCount).toHaveBeenCalledWith(
-        newTokenCount,
-      );
-      expect(uiTelemetryService.setLastPromptTokenCount).toHaveBeenCalledTimes(
-        1,
-      );
-    });
-
-    it('should trigger summarization if token count is at threshold with contextPercentageThreshold setting', async () => {
-      const MOCKED_TOKEN_LIMIT = 1000;
-      const MOCKED_CONTEXT_PERCENTAGE_THRESHOLD = 0.5;
-      vi.mocked(tokenLimit).mockReturnValue(MOCKED_TOKEN_LIMIT);
-      vi.spyOn(client['config'], 'getChatCompression').mockReturnValue({
-        contextPercentageThreshold: MOCKED_CONTEXT_PERCENTAGE_THRESHOLD,
-      });
-      const history = [
-        { role: 'user', parts: [{ text: '...history...' }] },
-        { role: 'model', parts: [{ text: '...history...' }] },
-        { role: 'user', parts: [{ text: '...history...' }] },
-        { role: 'model', parts: [{ text: '...history...' }] },
-        { role: 'user', parts: [{ text: '...history...' }] },
-        { role: 'model', parts: [{ text: '...history...' }] },
-      ];
-      mockGetHistory.mockReturnValue(history);
-
-      const originalTokenCount =
-        MOCKED_TOKEN_LIMIT * MOCKED_CONTEXT_PERCENTAGE_THRESHOLD;
-
-      vi.mocked(uiTelemetryService.getLastPromptTokenCount).mockReturnValue(
-        originalTokenCount,
-      );
-
-      // Mock summary and new chat
-      const summaryText = 'This is a summary.';
-      const splitPoint = findCompressSplitPoint(history, 0.7);
-      const historyToKeep = history.slice(splitPoint);
-      const newCompressedHistory: Content[] = [
-        { role: 'user', parts: [{ text: 'Mocked env context' }] },
-        { role: 'model', parts: [{ text: 'Got it. Thanks for the context!' }] },
-        { role: 'user', parts: [{ text: summaryText }] },
-        {
-          role: 'model',
-          parts: [{ text: 'Got it. Thanks for the additional context!' }],
-        },
-        ...historyToKeep,
-      ];
-      const mockNewChat: Partial<GeminiChat> = {
-        getHistory: vi.fn().mockReturnValue(newCompressedHistory),
-      };
-      client['startChat'] = vi
-        .fn()
-        .mockResolvedValue(mockNewChat as GeminiChat);
-
-      const totalChars = newCompressedHistory.reduce(
-        (total, content) => total + JSON.stringify(content).length,
-        0,
-      );
-      const newTokenCount = Math.floor(totalChars / 4);
-
-      // Mock the summary response from the chat
-      mockGenerateContentFn.mockResolvedValue({
-        candidates: [
-          {
-            content: {
-              role: 'model',
-              parts: [{ text: summaryText }],
-            },
-          },
-        ],
-      } as unknown as GenerateContentResponse);
-
-      const initialChat = client.getChat();
-      const result = await client.tryCompressChat('prompt-id-3', false);
-      const newChat = client.getChat();
-
-      expect(tokenLimit).toHaveBeenCalled();
-      expect(mockGenerateContentFn).toHaveBeenCalled();
-
-      // Assert that summarization happened and returned the correct stats
-      expect(result).toEqual({
-        compressionStatus: CompressionStatus.COMPRESSED,
-        originalTokenCount,
-        newTokenCount,
-      });
-
-      // Assert that the chat was reset
-      expect(newChat).not.toBe(initialChat);
-    });
-
-    it('should not compress across a function call response', async () => {
-      const MOCKED_TOKEN_LIMIT = 1000;
-      vi.mocked(tokenLimit).mockReturnValue(MOCKED_TOKEN_LIMIT);
-      const history: Content[] = [
-        { role: 'user', parts: [{ text: '...history 1...' }] },
-        { role: 'model', parts: [{ text: '...history 2...' }] },
-        { role: 'user', parts: [{ text: '...history 3...' }] },
-        { role: 'model', parts: [{ text: '...history 4...' }] },
-        { role: 'user', parts: [{ text: '...history 5...' }] },
-        { role: 'model', parts: [{ text: '...history 6...' }] },
-        { role: 'user', parts: [{ text: '...history 7...' }] },
-        { role: 'model', parts: [{ text: '...history 8...' }] },
-        // Normally we would break here, but we have a function response.
-        {
-          role: 'user',
-          parts: [{ functionResponse: { name: '...history 8...' } }],
-        },
-        { role: 'model', parts: [{ text: '...history 10...' }] },
-        // Instead we will break here.
-        { role: 'user', parts: [{ text: '...history 10...' }] },
-      ];
-      mockGetHistory.mockReturnValue(history);
-
-      const originalTokenCount = 1000 * 0.7;
-      vi.mocked(uiTelemetryService.getLastPromptTokenCount).mockReturnValue(
-        originalTokenCount,
-      );
-
-      // Mock summary and new chat
-      const summaryText = 'This is a summary.';
-      const splitPoint = findCompressSplitPoint(history, 0.7); // This should be 10
-      expect(splitPoint).toBe(10); // Verify split point logic
-      const historyToKeep = history.slice(splitPoint); // Should keep last user message
-      expect(historyToKeep).toEqual([
-        { role: 'user', parts: [{ text: '...history 10...' }] },
-      ]);
-
-      const newCompressedHistory: Content[] = [
-        { role: 'user', parts: [{ text: 'Mocked env context' }] },
-        { role: 'model', parts: [{ text: 'Got it. Thanks for the context!' }] },
-        { role: 'user', parts: [{ text: summaryText }] },
-        {
-          role: 'model',
-          parts: [{ text: 'Got it. Thanks for the additional context!' }],
-        },
-        ...historyToKeep,
-      ];
-      const mockNewChat: Partial<GeminiChat> = {
-        getHistory: vi.fn().mockReturnValue(newCompressedHistory),
-      };
-      client['startChat'] = vi
-        .fn()
-        .mockResolvedValue(mockNewChat as GeminiChat);
-
-      const totalChars = newCompressedHistory.reduce(
-        (total, content) => total + JSON.stringify(content).length,
-        0,
-      );
-      const newTokenCount = Math.floor(totalChars / 4);
-
-      // Mock the summary response from the chat
-      mockGenerateContentFn.mockResolvedValue({
-        candidates: [
-          {
-            content: {
-              role: 'model',
-              parts: [{ text: summaryText }],
-            },
-          },
-        ],
-      } as unknown as GenerateContentResponse);
-
-      const initialChat = client.getChat();
-      const result = await client.tryCompressChat('prompt-id-3', false);
-      const newChat = client.getChat();
-
-      expect(tokenLimit).toHaveBeenCalled();
-      expect(mockGenerateContentFn).toHaveBeenCalled();
-
-      // Assert that summarization happened and returned the correct stats
-      expect(result).toEqual({
-        compressionStatus: CompressionStatus.COMPRESSED,
-        originalTokenCount,
-        newTokenCount,
-      });
-      // Assert that the chat was reset
-      expect(newChat).not.toBe(initialChat);
-
-      // 1. standard start context message (env)
-      // 2. standard canned model response
-      // 3. compressed summary message (user)
-      // 4. standard canned model response
-      // 5. The last user message (historyToKeep)
-      expect(newChat.getHistory().length).toEqual(5);
-    });
-
-    it('should always trigger summarization when force is true, regardless of token count', async () => {
-      const history = [
-        { role: 'user', parts: [{ text: '...history...' }] },
-        { role: 'model', parts: [{ text: '...history...' }] },
-        { role: 'user', parts: [{ text: '...history...' }] },
-        { role: 'model', parts: [{ text: '...history...' }] },
-        { role: 'user', parts: [{ text: '...history...' }] },
-        { role: 'model', parts: [{ text: '...history...' }] },
-      ];
-      mockGetHistory.mockReturnValue(history);
-
-      const originalTokenCount = 100; // Well below threshold, but > estimated new count
-      vi.mocked(uiTelemetryService.getLastPromptTokenCount).mockReturnValue(
-        originalTokenCount,
-      );
-
-      // Mock summary and new chat
-      const summaryText = 'This is a summary.';
-      const splitPoint = findCompressSplitPoint(history, 0.7);
-      const historyToKeep = history.slice(splitPoint);
-      const newCompressedHistory: Content[] = [
-        { role: 'user', parts: [{ text: 'Mocked env context' }] },
-        { role: 'model', parts: [{ text: 'Got it. Thanks for the context!' }] },
-        { role: 'user', parts: [{ text: summaryText }] },
-        {
-          role: 'model',
-          parts: [{ text: 'Got it. Thanks for the additional context!' }],
-        },
-        ...historyToKeep,
-      ];
-      const mockNewChat: Partial<GeminiChat> = {
-        getHistory: vi.fn().mockReturnValue(newCompressedHistory),
-      };
-      client['startChat'] = vi
-        .fn()
-        .mockResolvedValue(mockNewChat as GeminiChat);
-
-      const totalChars = newCompressedHistory.reduce(
-        (total, content) => total + JSON.stringify(content).length,
-        0,
-      );
-      const newTokenCount = Math.floor(totalChars / 4);
-
-      // Mock the summary response from the chat
-      mockGenerateContentFn.mockResolvedValue({
-        candidates: [
-          {
-            content: {
-              role: 'model',
-              parts: [{ text: summaryText }],
-            },
-          },
-        ],
-      } as unknown as GenerateContentResponse);
-
-      const initialChat = client.getChat();
-      const result = await client.tryCompressChat('prompt-id-1', true); // force = true
-      const newChat = client.getChat();
-
-      expect(mockGenerateContentFn).toHaveBeenCalled();
-
-      expect(result).toEqual({
-        compressionStatus: CompressionStatus.COMPRESSED,
-        originalTokenCount,
-        newTokenCount,
-      });
-
-      // Assert that the chat was reset
-      expect(newChat).not.toBe(initialChat);
    });
  });

@@ -2073,7 +1669,11 @@ ${JSON.stringify(
          vi.mocked(ideContextStore.get).mockReturnValue({
            workspaceState: {
              openFiles: [
-                { ...currentActiveFile, isActive: true, timestamp: Date.now() },
+                {
+                  ...currentActiveFile,
+                  isActive: true,
+                  timestamp: Date.now(),
+                },
              ],
            },
          });
@@ -13,14 +13,13 @@ import type {
 } from '@google/genai';
 import {
  getDirectoryContextString,
-  getEnvironmentContext,
+  getInitialChatHistory,
 } from '../utils/environmentContext.js';
 import type { ServerGeminiStreamEvent, ChatCompressionInfo } from './turn.js';
 import { CompressionStatus } from './turn.js';
 import { Turn, GeminiEventType } from './turn.js';
 import type { Config } from '../config/config.js';
-import { getCoreSystemPrompt, getCompressionPrompt } from './prompts.js';
-import { getResponseText } from '../utils/partUtils.js';
+import { getCoreSystemPrompt } from './prompts.js';
 import { checkNextSpeaker } from '../utils/nextSpeakerChecker.js';
 import { reportError } from '../utils/errorReporting.js';
 import { GeminiChat } from './geminiChat.js';
@@ -37,15 +36,14 @@ import {
  getEffectiveModel,
 } from '../config/models.js';
 import { LoopDetectionService } from '../services/loopDetectionService.js';
+import { ChatCompressionService } from '../services/chatCompressionService.js';
 import { ideContextStore } from '../ide/ideContext.js';
 import {
-  logChatCompression,
  logContentRetryFailure,
  logNextSpeakerCheck,
 } from '../telemetry/loggers.js';
 import {
  ContentRetryFailureEvent,
-  makeChatCompressionEvent,
  NextSpeakerCheckEvent,
 } from '../telemetry/types.js';
 import type { IdeContext, File } from '../ide/types.js';
@@ -65,68 +63,8 @@ export function isThinkingDefault(model: string) {
  return model.startsWith('gemini-2.5') || model === DEFAULT_GEMINI_MODEL_AUTO;
 }

-/**
- * Returns the index of the oldest item to keep when compressing. May return
- * contents.length which indicates that everything should be compressed.
- *
- * Exported for testing purposes.
- */
-export function findCompressSplitPoint(
-  contents: Content[],
-  fraction: number,
-): number {
-  if (fraction <= 0 || fraction >= 1) {
-    throw new Error('Fraction must be between 0 and 1');
-  }
-
-  const charCounts = contents.map((content) => JSON.stringify(content).length);
-  const totalCharCount = charCounts.reduce((a, b) => a + b, 0);
-  const targetCharCount = totalCharCount * fraction;
-
-  let lastSplitPoint = 0; // 0 is always valid (compress nothing)
-  let cumulativeCharCount = 0;
-  for (let i = 0; i < contents.length; i++) {
-    const content = contents[i];
-    if (
-      content.role === 'user' &&
-      !content.parts?.some((part) => !!part.functionResponse)
-    ) {
-      if (cumulativeCharCount >= targetCharCount) {
-        return i;
-      }
-      lastSplitPoint = i;
-    }
-    cumulativeCharCount += charCounts[i];
-  }
-
-  // We found no split points after targetCharCount.
-  // Check if it's safe to compress everything.
-  const lastContent = contents[contents.length - 1];
-  if (
-    lastContent?.role === 'model' &&
-    !lastContent?.parts?.some((part) => part.functionCall)
-  ) {
-    return contents.length;
-  }
-
-  // Can't compress everything so just compress at last splitpoint.
-  return lastSplitPoint;
-}
-
 const MAX_TURNS = 100;

-/**
- * Threshold for compression token count as a fraction of the model's token limit.
- * If the chat history exceeds this threshold, it will be compressed.
- */
-const COMPRESSION_TOKEN_THRESHOLD = 0.7;
-
-/**
- * The fraction of the latest chat history to keep. A value of 0.3
- * means that only the last 30% of the chat history will be kept after compression.
- */
-const COMPRESSION_PRESERVE_THRESHOLD = 0.3;
-
 export class GeminiClient {
  private chat?: GeminiChat;
  private readonly generateContentConfig: GenerateContentConfig = {
@@ -136,6 +74,7 @@ export class GeminiClient {
  private sessionTurnCount = 0;

  private readonly loopDetector: LoopDetectionService;
+  private readonly compressionService: ChatCompressionService;
  private lastPromptId: string;
  private currentSequenceModel: string | null = null;
  private lastSentIdeContext: IdeContext | undefined;
@@ -149,6 +88,7 @@ export class GeminiClient {

  constructor(private readonly config: Config) {
    this.loopDetector = new LoopDetectionService(config);
+    this.compressionService = new ChatCompressionService();
    this.lastPromptId = this.config.getSessionId();
  }

@@ -233,31 +173,7 @@ export class GeminiClient {
    const toolDeclarations = toolRegistry.getFunctionDeclarations();
    const tools: Tool[] = [{ functionDeclarations: toolDeclarations }];

-    // 1. Get the environment context parts as an array
-    const envParts = await getEnvironmentContext(this.config);
-
-    // 2. Convert the array of parts into a single string
-    const envContextString = envParts
-      .map((part) => part.text || '')
-      .join('\n\n');
-
-    // 3. Combine the dynamic context with the static handshake instruction
-    const allSetupText = `
-${envContextString}
-
-Reminder: Do not return an empty response when a tool call is required.
-
-My setup is complete. I will provide my first command in the next turn.
-    `.trim();
-
-    // 4. Create the history with a single, comprehensive user turn
-    const history: Content[] = [
-      {
-        role: 'user',
-        parts: [{ text: allSetupText }],
-      },
-      ...(extraHistory ?? []),
-    ];
+    const history = await getInitialChatHistory(this.config, extraHistory);

    try {
      const userMemory = this.config.getUserMemory();
@@ -738,129 +654,27 @@ My setup is complete. I will provide my first command in the next turn.
    // before the model is chosen would result in an error.
    const model = this._getEffectiveModelForCurrentTurn();

-    const curatedHistory = this.getChat().getHistory(true);
+    const { newHistory, info } = await this.compressionService.compress(
+      this.getChat(),
+      prompt_id,
+      force,
+      model,
+      this.config,
+      this.hasFailedCompressionAttempt,
+    );

-    // Regardless of `force`, don't do anything if the history is empty.
    if (
-      curatedHistory.length === 0 ||
-      (this.hasFailedCompressionAttempt && !force)
+      info.compressionStatus ===
+      CompressionStatus.COMPRESSION_FAILED_INFLATED_TOKEN_COUNT
    ) {
-      return {
-        originalTokenCount: 0,
-        newTokenCount: 0,
-        compressionStatus: CompressionStatus.NOOP,
-      };
-    }
-
-    const originalTokenCount = uiTelemetryService.getLastPromptTokenCount();
-
-    const contextPercentageThreshold =
-      this.config.getChatCompression()?.contextPercentageThreshold;
-
-    // Don't compress if not forced and we are under the limit.
-    if (!force) {
-      const threshold =
-        contextPercentageThreshold ?? COMPRESSION_TOKEN_THRESHOLD;
-      if (originalTokenCount < threshold * tokenLimit(model)) {
-        return {
-          originalTokenCount,
-          newTokenCount: originalTokenCount,
-          compressionStatus: CompressionStatus.NOOP,
-        };
+      this.hasFailedCompressionAttempt = !force && true;
+    } else if (info.compressionStatus === CompressionStatus.COMPRESSED) {
+      if (newHistory) {
+        this.chat = await this.startChat(newHistory);
+        this.forceFullIdeContext = true;
      }
    }

-    const splitPoint = findCompressSplitPoint(
-      curatedHistory,
-      1 - COMPRESSION_PRESERVE_THRESHOLD,
-    );
-
-    const historyToCompress = curatedHistory.slice(0, splitPoint);
-    const historyToKeep = curatedHistory.slice(splitPoint);
-
-    if (historyToCompress.length === 0) {
-      return {
-        originalTokenCount,
-        newTokenCount: originalTokenCount,
-        compressionStatus: CompressionStatus.NOOP,
-      };
-    }
-
-    const summaryResponse = await this.config
-      .getContentGenerator()
-      .generateContent(
-        {
-          model,
-          contents: [
-            ...historyToCompress,
-            {
-              role: 'user',
-              parts: [
-                {
-                  text: 'First, reason in your scratchpad. Then, generate the <state_snapshot>.',
-                },
-              ],
-            },
-          ],
-          config: {
-            systemInstruction: { text: getCompressionPrompt() },
-          },
-        },
-        prompt_id,
-      );
-    const summary = getResponseText(summaryResponse) ?? '';
-
-    const chat = await this.startChat([
-      {
-        role: 'user',
-        parts: [{ text: summary }],
-      },
-      {
-        role: 'model',
-        parts: [{ text: 'Got it. Thanks for the additional context!' }],
-      },
-      ...historyToKeep,
-    ]);
-    this.forceFullIdeContext = true;
-
-    // Estimate token count 1 token ≈ 4 characters
-    const newTokenCount = Math.floor(
-      chat
-        .getHistory()
-        .reduce((total, content) => total + JSON.stringify(content).length, 0) /
-        4,
-    );
-
-    logChatCompression(
-      this.config,
-      makeChatCompressionEvent({
-        tokens_before: originalTokenCount,
-        tokens_after: newTokenCount,
-      }),
-    );
-
-    if (newTokenCount > originalTokenCount) {
-      this.hasFailedCompressionAttempt = !force && true;
-      return {
-        originalTokenCount,
-        newTokenCount,
-        compressionStatus:
-          CompressionStatus.COMPRESSION_FAILED_INFLATED_TOKEN_COUNT,
-      };
-    } else {
-      this.chat = chat; // Chat compression successful, set new state.
-      uiTelemetryService.setLastPromptTokenCount(newTokenCount);
-    }
-
-    return {
-      originalTokenCount,
-      newTokenCount,
-      compressionStatus: CompressionStatus.COMPRESSED,
-    };
+    return info;
  }
 }
-
-export const TEST_ONLY = {
-  COMPRESSION_PRESERVE_THRESHOLD,
-  COMPRESSION_TOKEN_THRESHOLD,
-};
@@ -45,6 +45,7 @@ export * from './utils/gitIgnoreParser.js';
 export * from './utils/gitUtils.js';
 export * from './utils/editor.js';
 export * from './utils/quotaErrorDetection.js';
+export * from './utils/googleQuotaErrors.js';
 export * from './utils/fileUtils.js';
 export * from './utils/retry.js';
 export * from './utils/shell-utils.js';
@@ -0,0 +1,296 @@
+/**
+ * @license
+ * Copyright 2025 Google LLC
+ * SPDX-License-Identifier: Apache-2.0
+ */
+
+import { describe, it, expect, vi, beforeEach, afterEach } from 'vitest';
+import {
+  ChatCompressionService,
+  findCompressSplitPoint,
+} from './chatCompressionService.js';
+import type { Content, GenerateContentResponse } from '@google/genai';
+import { CompressionStatus } from '../core/turn.js';
+import { uiTelemetryService } from '../telemetry/uiTelemetry.js';
+import { tokenLimit } from '../core/tokenLimits.js';
+import type { GeminiChat } from '../core/geminiChat.js';
+import type { Config } from '../config/config.js';
+import { getInitialChatHistory } from '../utils/environmentContext.js';
+import type { ContentGenerator } from '../core/contentGenerator.js';
+
+vi.mock('../telemetry/uiTelemetry.js');
+vi.mock('../core/tokenLimits.js');
+vi.mock('../telemetry/loggers.js');
+vi.mock('../utils/environmentContext.js');
+
+describe('findCompressSplitPoint', () => {
+  it('should throw an error for non-positive numbers', () => {
+    expect(() => findCompressSplitPoint([], 0)).toThrow(
+      'Fraction must be between 0 and 1',
+    );
+  });
+
+  it('should throw an error for a fraction greater than or equal to 1', () => {
+    expect(() => findCompressSplitPoint([], 1)).toThrow(
+      'Fraction must be between 0 and 1',
+    );
+  });
+
+  it('should handle an empty history', () => {
+    expect(findCompressSplitPoint([], 0.5)).toBe(0);
+  });
+
+  it('should handle a fraction in the middle', () => {
+    const history: Content[] = [
+      { role: 'user', parts: [{ text: 'This is the first message.' }] }, // JSON length: 66 (19%)
+      { role: 'model', parts: [{ text: 'This is the second message.' }] }, // JSON length: 68 (40%)
+      { role: 'user', parts: [{ text: 'This is the third message.' }] }, // JSON length: 66 (60%)
+      { role: 'model', parts: [{ text: 'This is the fourth message.' }] }, // JSON length: 68 (80%)
+      { role: 'user', parts: [{ text: 'This is the fifth message.' }] }, // JSON length: 65 (100%)
+    ];
+    expect(findCompressSplitPoint(history, 0.5)).toBe(4);
+  });
+
+  it('should handle a fraction of last index', () => {
+    const history: Content[] = [
+      { role: 'user', parts: [{ text: 'This is the first message.' }] }, // JSON length: 66 (19%)
+      { role: 'model', parts: [{ text: 'This is the second message.' }] }, // JSON length: 68 (40%)
+      { role: 'user', parts: [{ text: 'This is the third message.' }] }, // JSON length: 66 (60%)
+      { role: 'model', parts: [{ text: 'This is the fourth message.' }] }, // JSON length: 68 (80%)
+      { role: 'user', parts: [{ text: 'This is the fifth message.' }] }, // JSON length: 65 (100%)
+    ];
+    expect(findCompressSplitPoint(history, 0.9)).toBe(4);
+  });
+
+  it('should handle a fraction of after last index', () => {
+    const history: Content[] = [
+      { role: 'user', parts: [{ text: 'This is the first message.' }] }, // JSON length: 66 (24%)
+      { role: 'model', parts: [{ text: 'This is the second message.' }] }, // JSON length: 68 (50%)
+      { role: 'user', parts: [{ text: 'This is the third message.' }] }, // JSON length: 66 (74%)
+      { role: 'model', parts: [{ text: 'This is the fourth message.' }] }, // JSON length: 68 (100%)
+    ];
+    expect(findCompressSplitPoint(history, 0.8)).toBe(4);
+  });
+
+  it('should return earlier splitpoint if no valid ones are after threshold', () => {
+    const history: Content[] = [
+      { role: 'user', parts: [{ text: 'This is the first message.' }] },
+      { role: 'model', parts: [{ text: 'This is the second message.' }] },
+      { role: 'user', parts: [{ text: 'This is the third message.' }] },
+      { role: 'model', parts: [{ functionCall: { name: 'foo', args: {} } }] },
+    ];
+    // Can't return 4 because the previous item has a function call.
+    expect(findCompressSplitPoint(history, 0.99)).toBe(2);
+  });
+
+  it('should handle a history with only one item', () => {
+    const historyWithEmptyParts: Content[] = [
+      { role: 'user', parts: [{ text: 'Message 1' }] },
+    ];
+    expect(findCompressSplitPoint(historyWithEmptyParts, 0.5)).toBe(0);
+  });
+
+  it('should handle history with weird parts', () => {
+    const historyWithEmptyParts: Content[] = [
+      { role: 'user', parts: [{ text: 'Message 1' }] },
+      {
+        role: 'model',
+        parts: [{ fileData: { fileUri: 'derp', mimeType: 'text/plain' } }],
+      },
+      { role: 'user', parts: [{ text: 'Message 2' }] },
+    ];
+    expect(findCompressSplitPoint(historyWithEmptyParts, 0.5)).toBe(2);
+  });
+});
+
+describe('ChatCompressionService', () => {
+  let service: ChatCompressionService;
+  let mockChat: GeminiChat;
+  let mockConfig: Config;
+  const mockModel = 'gemini-pro';
+  const mockPromptId = 'test-prompt-id';
+
+  beforeEach(() => {
+    service = new ChatCompressionService();
+    mockChat = {
+      getHistory: vi.fn(),
+    } as unknown as GeminiChat;
+    mockConfig = {
+      getCompressionThreshold: vi.fn(),
+      getContentGenerator: vi.fn(),
+    } as unknown as Config;
+
+    vi.mocked(tokenLimit).mockReturnValue(1000);
+    vi.mocked(uiTelemetryService.getLastPromptTokenCount).mockReturnValue(500);
+    vi.mocked(getInitialChatHistory).mockImplementation(
+      async (_config, extraHistory) => extraHistory || [],
+    );
+  });
+
+  afterEach(() => {
+    vi.restoreAllMocks();
+  });
+
+  it('should return NOOP if history is empty', async () => {
+    vi.mocked(mockChat.getHistory).mockReturnValue([]);
+    const result = await service.compress(
+      mockChat,
+      mockPromptId,
+      false,
+      mockModel,
+      mockConfig,
+      false,
+    );
+    expect(result.info.compressionStatus).toBe(CompressionStatus.NOOP);
+    expect(result.newHistory).toBeNull();
+  });
+
+  it('should return NOOP if previously failed and not forced', async () => {
+    vi.mocked(mockChat.getHistory).mockReturnValue([
+      { role: 'user', parts: [{ text: 'hi' }] },
+    ]);
+    const result = await service.compress(
+      mockChat,
+      mockPromptId,
+      false,
+      mockModel,
+      mockConfig,
+      true,
+    );
+    expect(result.info.compressionStatus).toBe(CompressionStatus.NOOP);
+    expect(result.newHistory).toBeNull();
+  });
+
+  it('should return NOOP if under token threshold and not forced', async () => {
+    vi.mocked(mockChat.getHistory).mockReturnValue([
+      { role: 'user', parts: [{ text: 'hi' }] },
+    ]);
+    vi.mocked(uiTelemetryService.getLastPromptTokenCount).mockReturnValue(600);
+    vi.mocked(tokenLimit).mockReturnValue(1000);
+    // Threshold is 0.7 * 1000 = 700. 600 < 700, so NOOP.
+
+    const result = await service.compress(
+      mockChat,
+      mockPromptId,
+      false,
+      mockModel,
+      mockConfig,
+      false,
+    );
+    expect(result.info.compressionStatus).toBe(CompressionStatus.NOOP);
+    expect(result.newHistory).toBeNull();
+  });
+
+  it('should compress if over token threshold', async () => {
+    const history: Content[] = [
+      { role: 'user', parts: [{ text: 'msg1' }] },
+      { role: 'model', parts: [{ text: 'msg2' }] },
+      { role: 'user', parts: [{ text: 'msg3' }] },
+      { role: 'model', parts: [{ text: 'msg4' }] },
+    ];
+    vi.mocked(mockChat.getHistory).mockReturnValue(history);
+    vi.mocked(uiTelemetryService.getLastPromptTokenCount).mockReturnValue(800);
+    vi.mocked(tokenLimit).mockReturnValue(1000);
+    const mockGenerateContent = vi.fn().mockResolvedValue({
+      candidates: [
+        {
+          content: {
+            parts: [{ text: 'Summary' }],
+          },
+        },
+      ],
+    } as unknown as GenerateContentResponse);
+    vi.mocked(mockConfig.getContentGenerator).mockReturnValue({
+      generateContent: mockGenerateContent,
+    } as unknown as ContentGenerator);
+
+    const result = await service.compress(
+      mockChat,
+      mockPromptId,
+      false,
+      mockModel,
+      mockConfig,
+      false,
+    );
+
+    expect(result.info.compressionStatus).toBe(CompressionStatus.COMPRESSED);
+    expect(result.newHistory).not.toBeNull();
+    expect(result.newHistory![0].parts![0].text).toBe('Summary');
+    expect(mockGenerateContent).toHaveBeenCalled();
+  });
+
+  it('should force compress even if under threshold', async () => {
+    const history: Content[] = [
+      { role: 'user', parts: [{ text: 'msg1' }] },
+      { role: 'model', parts: [{ text: 'msg2' }] },
+      { role: 'user', parts: [{ text: 'msg3' }] },
+      { role: 'model', parts: [{ text: 'msg4' }] },
+    ];
+    vi.mocked(mockChat.getHistory).mockReturnValue(history);
+    vi.mocked(uiTelemetryService.getLastPromptTokenCount).mockReturnValue(100);
+    vi.mocked(tokenLimit).mockReturnValue(1000);
+
+    const mockGenerateContent = vi.fn().mockResolvedValue({
+      candidates: [
+        {
+          content: {
+            parts: [{ text: 'Summary' }],
+          },
+        },
+      ],
+    } as unknown as GenerateContentResponse);
+    vi.mocked(mockConfig.getContentGenerator).mockReturnValue({
+      generateContent: mockGenerateContent,
+    } as unknown as ContentGenerator);
+
+    const result = await service.compress(
+      mockChat,
+      mockPromptId,
+      true, // forced
+      mockModel,
+      mockConfig,
+      false,
+    );
+
+    expect(result.info.compressionStatus).toBe(CompressionStatus.COMPRESSED);
+    expect(result.newHistory).not.toBeNull();
+  });
+
+  it('should return FAILED if new token count is inflated', async () => {
+    const history: Content[] = [
+      { role: 'user', parts: [{ text: 'msg1' }] },
+      { role: 'model', parts: [{ text: 'msg2' }] },
+    ];
+    vi.mocked(mockChat.getHistory).mockReturnValue(history);
+    vi.mocked(uiTelemetryService.getLastPromptTokenCount).mockReturnValue(10);
+    vi.mocked(tokenLimit).mockReturnValue(1000);
+
+    const longSummary = 'a'.repeat(1000); // Long summary to inflate token count
+    const mockGenerateContent = vi.fn().mockResolvedValue({
+      candidates: [
+        {
+          content: {
+            parts: [{ text: longSummary }],
+          },
+        },
+      ],
+    } as unknown as GenerateContentResponse);
+    vi.mocked(mockConfig.getContentGenerator).mockReturnValue({
+      generateContent: mockGenerateContent,
+    } as unknown as ContentGenerator);
+
+    const result = await service.compress(
+      mockChat,
+      mockPromptId,
+      true,
+      mockModel,
+      mockConfig,
+      false,
+    );
+
+    expect(result.info.compressionStatus).toBe(
+      CompressionStatus.COMPRESSION_FAILED_INFLATED_TOKEN_COUNT,
+    );
+    expect(result.newHistory).toBeNull();
+  });
+});
@@ -0,0 +1,217 @@
+/**
+ * @license
+ * Copyright 2025 Google LLC
+ * SPDX-License-Identifier: Apache-2.0
+ */
+
+import type { Content } from '@google/genai';
+import type { Config } from '../config/config.js';
+import type { GeminiChat } from '../core/geminiChat.js';
+import { type ChatCompressionInfo, CompressionStatus } from '../core/turn.js';
+import { uiTelemetryService } from '../telemetry/uiTelemetry.js';
+import { tokenLimit } from '../core/tokenLimits.js';
+import { getCompressionPrompt } from '../core/prompts.js';
+import { getResponseText } from '../utils/partUtils.js';
+import { logChatCompression } from '../telemetry/loggers.js';
+import { makeChatCompressionEvent } from '../telemetry/types.js';
+import { getInitialChatHistory } from '../utils/environmentContext.js';
+
+/**
+ * Default threshold for compression token count as a fraction of the model's
+ * token limit. If the chat history exceeds this threshold, it will be compressed.
+ */
+export const DEFAULT_COMPRESSION_TOKEN_THRESHOLD = 0.2;
+
+/**
+ * The fraction of the latest chat history to keep. A value of 0.3
+ * means that only the last 30% of the chat history will be kept after compression.
+ */
+export const COMPRESSION_PRESERVE_THRESHOLD = 0.3;
+
+/**
+ * Returns the index of the oldest item to keep when compressing. May return
+ * contents.length which indicates that everything should be compressed.
+ *
+ * Exported for testing purposes.
+ */
+export function findCompressSplitPoint(
+  contents: Content[],
+  fraction: number,
+): number {
+  if (fraction <= 0 || fraction >= 1) {
+    throw new Error('Fraction must be between 0 and 1');
+  }
+
+  const charCounts = contents.map((content) => JSON.stringify(content).length);
+  const totalCharCount = charCounts.reduce((a, b) => a + b, 0);
+  const targetCharCount = totalCharCount * fraction;
+
+  let lastSplitPoint = 0; // 0 is always valid (compress nothing)
+  let cumulativeCharCount = 0;
+  for (let i = 0; i < contents.length; i++) {
+    const content = contents[i];
+    if (
+      content.role === 'user' &&
+      !content.parts?.some((part) => !!part.functionResponse)
+    ) {
+      if (cumulativeCharCount >= targetCharCount) {
+        return i;
+      }
+      lastSplitPoint = i;
+    }
+    cumulativeCharCount += charCounts[i];
+  }
+
+  // We found no split points after targetCharCount.
+  // Check if it's safe to compress everything.
+  const lastContent = contents[contents.length - 1];
+  if (
+    lastContent?.role === 'model' &&
+    !lastContent?.parts?.some((part) => part.functionCall)
+  ) {
+    return contents.length;
+  }
+
+  // Can't compress everything so just compress at last splitpoint.
+  return lastSplitPoint;
+}
+
+export class ChatCompressionService {
+  async compress(
+    chat: GeminiChat,
+    promptId: string,
+    force: boolean,
+    model: string,
+    config: Config,
+    hasFailedCompressionAttempt: boolean,
+  ): Promise<{ newHistory: Content[] | null; info: ChatCompressionInfo }> {
+    const curatedHistory = chat.getHistory(true);
+
+    // Regardless of `force`, don't do anything if the history is empty.
+    if (
+      curatedHistory.length === 0 ||
+      (hasFailedCompressionAttempt && !force)
+    ) {
+      return {
+        newHistory: null,
+        info: {
+          originalTokenCount: 0,
+          newTokenCount: 0,
+          compressionStatus: CompressionStatus.NOOP,
+        },
+      };
+    }
+
+    const originalTokenCount = uiTelemetryService.getLastPromptTokenCount();
+
+    // Don't compress if not forced and we are under the limit.
+    if (!force) {
+      const threshold =
+        config.getCompressionThreshold() ?? DEFAULT_COMPRESSION_TOKEN_THRESHOLD;
+      if (originalTokenCount < threshold * tokenLimit(model)) {
+        return {
+          newHistory: null,
+          info: {
+            originalTokenCount,
+            newTokenCount: originalTokenCount,
+            compressionStatus: CompressionStatus.NOOP,
+          },
+        };
+      }
+    }
+
+    const splitPoint = findCompressSplitPoint(
+      curatedHistory,
+      1 - COMPRESSION_PRESERVE_THRESHOLD,
+    );
+
+    const historyToCompress = curatedHistory.slice(0, splitPoint);
+    const historyToKeep = curatedHistory.slice(splitPoint);
+
+    if (historyToCompress.length === 0) {
+      return {
+        newHistory: null,
+        info: {
+          originalTokenCount,
+          newTokenCount: originalTokenCount,
+          compressionStatus: CompressionStatus.NOOP,
+        },
+      };
+    }
+
+    const summaryResponse = await config.getContentGenerator().generateContent(
+      {
+        model,
+        contents: [
+          ...historyToCompress,
+          {
+            role: 'user',
+            parts: [
+              {
+                text: 'First, reason in your scratchpad. Then, generate the <state_snapshot>.',
+              },
+            ],
+          },
+        ],
+        config: {
+          systemInstruction: { text: getCompressionPrompt() },
+        },
+      },
+      promptId,
+    );
+    const summary = getResponseText(summaryResponse) ?? '';
+
+    const extraHistory: Content[] = [
+      {
+        role: 'user',
+        parts: [{ text: summary }],
+      },
+      {
+        role: 'model',
+        parts: [{ text: 'Got it. Thanks for the additional context!' }],
+      },
+      ...historyToKeep,
+    ];
+
+    // Use a shared utility to construct the initial history for an accurate token count.
+    const fullNewHistory = await getInitialChatHistory(config, extraHistory);
+
+    // Estimate token count 1 token ≈ 4 characters
+    const newTokenCount = Math.floor(
+      fullNewHistory.reduce(
+        (total, content) => total + JSON.stringify(content).length,
+        0,
+      ) / 4,
+    );
+
+    logChatCompression(
+      config,
+      makeChatCompressionEvent({
+        tokens_before: originalTokenCount,
+        tokens_after: newTokenCount,
+      }),
+    );
+
+    if (newTokenCount > originalTokenCount) {
+      return {
+        newHistory: null,
+        info: {
+          originalTokenCount,
+          newTokenCount,
+          compressionStatus:
+            CompressionStatus.COMPRESSION_FAILED_INFLATED_TOKEN_COUNT,
+        },
+      };
+    } else {
+      uiTelemetryService.setLastPromptTokenCount(newTokenCount);
+      return {
+        newHistory: extraHistory,
+        info: {
+          originalTokenCount,
+          newTokenCount,
+          compressionStatus: CompressionStatus.COMPRESSED,
+        },
+      };
+    }
+  }
+}
@@ -142,6 +142,19 @@ describe('LoopDetectionService', () => {
      }
      expect(loggers.logLoopDetected).not.toHaveBeenCalled();
    });
+
+    it('should stop reporting a loop if disabled after detection', () => {
+      const event = createToolCallRequestEvent('testTool', { param: 'value' });
+      for (let i = 0; i < TOOL_CALL_LOOP_THRESHOLD; i++) {
+        service.addAndCheck(event);
+      }
+      expect(service.addAndCheck(event)).toBe(true);
+
+      service.disableForSession();
+
+      // Should now return false even though a loop was previously detected
+      expect(service.addAndCheck(event)).toBe(false);
+    });
  });

  describe('Content Loop Detection', () => {
@@ -123,7 +123,11 @@ export class LoopDetectionService {
   * @returns true if a loop is detected, false otherwise
   */
  addAndCheck(event: ServerGeminiStreamEvent): boolean {
-    if (this.loopDetected || this.disabledForSession) {
+    if (this.disabledForSession) {
+      return false;
+    }
+
+    if (this.loopDetected) {
      return this.loopDetected;
    }

@@ -4,7 +4,7 @@
 * SPDX-License-Identifier: Apache-2.0
 */

-import type { Part } from '@google/genai';
+import type { Part, Content } from '@google/genai';
 import type { Config } from '../config/config.js';
 import { getFolderStructure } from './getFolderStructure.js';

@@ -71,3 +71,27 @@ ${directoryContext}

  return initialParts;
 }
+
+export async function getInitialChatHistory(
+  config: Config,
+  extraHistory?: Content[],
+): Promise<Content[]> {
+  const envParts = await getEnvironmentContext(config);
+  const envContextString = envParts.map((part) => part.text || '').join('\n\n');
+
+  const allSetupText = `
+${envContextString}
+
+Reminder: Do not return an empty response when a tool call is required.
+
+My setup is complete. I will provide my first command in the next turn.
+    `.trim();
+
+  return [
+    {
+      role: 'user',
+      parts: [{ text: allSetupText }],
+    },
+    ...(extraHistory ?? []),
+  ];
+}
@@ -6,9 +6,7 @@

 import { describe, it, expect } from 'vitest';
 import { parseAndFormatApiError } from './errorParsing.js';
-import { isProQuotaExceededError } from './quotaErrorDetection.js';
 import { DEFAULT_GEMINI_FLASH_MODEL } from '../config/models.js';
-import { UserTierId } from '../code_assist/types.js';
 import { AuthType } from '../core/contentGenerator.js';
 import type { StructuredError } from '../core/turn.js';

@@ -40,22 +38,6 @@ describe('parseAndFormatApiError', () => {
    );
  });

-  it('should format a 429 API error with the personal message', () => {
-    const errorMessage =
-      'got status: 429 Too Many Requests. {"error":{"code":429,"message":"Rate limit exceeded","status":"RESOURCE_EXHAUSTED"}}';
-    const result = parseAndFormatApiError(
-      errorMessage,
-      AuthType.LOGIN_WITH_GOOGLE,
-      undefined,
-      'gemini-2.5-pro',
-      DEFAULT_GEMINI_FLASH_MODEL,
-    );
-    expect(result).toContain('[API Error: Rate limit exceeded');
-    expect(result).toContain(
-      'Possible quota limitations in place or slow response times detected. Switching to the gemini-2.5-flash model',
-    );
-  });
-
  it('should format a 429 API error with the vertex message', () => {
    const errorMessage =
      'got status: 429 Too Many Requests. {"error":{"code":429,"message":"Rate limit exceeded","status":"RESOURCE_EXHAUSTED"}}';
@@ -132,230 +114,4 @@ describe('parseAndFormatApiError', () => {
    const expected = '[API Error: An unknown error occurred.]';
    expect(parseAndFormatApiError(error)).toBe(expected);
  });
-
-  it('should format a 429 API error with Pro quota exceeded message for Google auth (Free tier)', () => {
-    const errorMessage =
-      'got status: 429 Too Many Requests. {"error":{"code":429,"message":"Quota exceeded for quota metric \'Gemini 2.5 Pro Requests\' and limit \'RequestsPerDay\' of service \'generativelanguage.googleapis.com\' for consumer \'project_number:123456789\'.","status":"RESOURCE_EXHAUSTED"}}';
-    const result = parseAndFormatApiError(
-      errorMessage,
-      AuthType.LOGIN_WITH_GOOGLE,
-      undefined,
-      'gemini-2.5-pro',
-      DEFAULT_GEMINI_FLASH_MODEL,
-    );
-    expect(result).toContain(
-      "[API Error: Quota exceeded for quota metric 'Gemini 2.5 Pro Requests'",
-    );
-    expect(result).toContain(
-      'You have reached your daily gemini-2.5-pro quota limit',
-    );
-    expect(result).toContain('upgrade to get higher limits');
-  });
-
-  it('should format a regular 429 API error with standard message for Google auth', () => {
-    const errorMessage =
-      'got status: 429 Too Many Requests. {"error":{"code":429,"message":"Rate limit exceeded","status":"RESOURCE_EXHAUSTED"}}';
-    const result = parseAndFormatApiError(
-      errorMessage,
-      AuthType.LOGIN_WITH_GOOGLE,
-      undefined,
-      'gemini-2.5-pro',
-      DEFAULT_GEMINI_FLASH_MODEL,
-    );
-    expect(result).toContain('[API Error: Rate limit exceeded');
-    expect(result).toContain(
-      'Possible quota limitations in place or slow response times detected. Switching to the gemini-2.5-flash model',
-    );
-    expect(result).not.toContain(
-      'You have reached your daily gemini-2.5-pro quota limit',
-    );
-  });
-
-  it('should format a 429 API error with generic quota exceeded message for Google auth', () => {
-    const errorMessage =
-      'got status: 429 Too Many Requests. {"error":{"code":429,"message":"Quota exceeded for quota metric \'GenerationRequests\' and limit \'RequestsPerDay\' of service \'generativelanguage.googleapis.com\' for consumer \'project_number:123456789\'.","status":"RESOURCE_EXHAUSTED"}}';
-    const result = parseAndFormatApiError(
-      errorMessage,
-      AuthType.LOGIN_WITH_GOOGLE,
-      undefined,
-      'gemini-2.5-pro',
-      DEFAULT_GEMINI_FLASH_MODEL,
-    );
-    expect(result).toContain(
-      "[API Error: Quota exceeded for quota metric 'GenerationRequests'",
-    );
-    expect(result).toContain('You have reached your daily quota limit');
-    expect(result).not.toContain(
-      'You have reached your daily Gemini 2.5 Pro quota limit',
-    );
-  });
-
-  it('should prioritize Pro quota message over generic quota message for Google auth', () => {
-    const errorMessage =
-      'got status: 429 Too Many Requests. {"error":{"code":429,"message":"Quota exceeded for quota metric \'Gemini 2.5 Pro Requests\' and limit \'RequestsPerDay\' of service \'generativelanguage.googleapis.com\' for consumer \'project_number:123456789\'.","status":"RESOURCE_EXHAUSTED"}}';
-    const result = parseAndFormatApiError(
-      errorMessage,
-      AuthType.LOGIN_WITH_GOOGLE,
-      undefined,
-      'gemini-2.5-pro',
-      DEFAULT_GEMINI_FLASH_MODEL,
-    );
-    expect(result).toContain(
-      "[API Error: Quota exceeded for quota metric 'Gemini 2.5 Pro Requests'",
-    );
-    expect(result).toContain(
-      'You have reached your daily gemini-2.5-pro quota limit',
-    );
-    expect(result).not.toContain('You have reached your daily quota limit');
-  });
-
-  it('should format a 429 API error with Pro quota exceeded message for Google auth (Standard tier)', () => {
-    const errorMessage =
-      'got status: 429 Too Many Requests. {"error":{"code":429,"message":"Quota exceeded for quota metric \'Gemini 2.5 Pro Requests\' and limit \'RequestsPerDay\' of service \'generativelanguage.googleapis.com\' for consumer \'project_number:123456789\'.","status":"RESOURCE_EXHAUSTED"}}';
-    const result = parseAndFormatApiError(
-      errorMessage,
-      AuthType.LOGIN_WITH_GOOGLE,
-      UserTierId.STANDARD,
-      'gemini-2.5-pro',
-      DEFAULT_GEMINI_FLASH_MODEL,
-    );
-    expect(result).toContain(
-      "[API Error: Quota exceeded for quota metric 'Gemini 2.5 Pro Requests'",
-    );
-    expect(result).toContain(
-      'You have reached your daily gemini-2.5-pro quota limit',
-    );
-    expect(result).toContain(
-      'We appreciate you for choosing Gemini Code Assist and the Gemini CLI',
-    );
-    expect(result).not.toContain('upgrade to get higher limits');
-  });
-
-  it('should format a 429 API error with Pro quota exceeded message for Google auth (Legacy tier)', () => {
-    const errorMessage =
-      'got status: 429 Too Many Requests. {"error":{"code":429,"message":"Quota exceeded for quota metric \'Gemini 2.5 Pro Requests\' and limit \'RequestsPerDay\' of service \'generativelanguage.googleapis.com\' for consumer \'project_number:123456789\'.","status":"RESOURCE_EXHAUSTED"}}';
-    const result = parseAndFormatApiError(
-      errorMessage,
-      AuthType.LOGIN_WITH_GOOGLE,
-      UserTierId.LEGACY,
-      'gemini-2.5-pro',
-      DEFAULT_GEMINI_FLASH_MODEL,
-    );
-    expect(result).toContain(
-      "[API Error: Quota exceeded for quota metric 'Gemini 2.5 Pro Requests'",
-    );
-    expect(result).toContain(
-      'You have reached your daily gemini-2.5-pro quota limit',
-    );
-    expect(result).toContain(
-      'We appreciate you for choosing Gemini Code Assist and the Gemini CLI',
-    );
-    expect(result).not.toContain('upgrade to get higher limits');
-  });
-
-  it('should handle different Gemini 2.5 version strings in Pro quota exceeded errors', () => {
-    const errorMessage25 =
-      'got status: 429 Too Many Requests. {"error":{"code":429,"message":"Quota exceeded for quota metric \'Gemini 2.5 Pro Requests\' and limit \'RequestsPerDay\' of service \'generativelanguage.googleapis.com\' for consumer \'project_number:123456789\'.","status":"RESOURCE_EXHAUSTED"}}';
-    const errorMessagePreview =
-      'got status: 429 Too Many Requests. {"error":{"code":429,"message":"Quota exceeded for quota metric \'Gemini 2.5-preview Pro Requests\' and limit \'RequestsPerDay\' of service \'generativelanguage.googleapis.com\' for consumer \'project_number:123456789\'.","status":"RESOURCE_EXHAUSTED"}}';
-
-    const result25 = parseAndFormatApiError(
-      errorMessage25,
-      AuthType.LOGIN_WITH_GOOGLE,
-      undefined,
-      'gemini-2.5-pro',
-      DEFAULT_GEMINI_FLASH_MODEL,
-    );
-    const resultPreview = parseAndFormatApiError(
-      errorMessagePreview,
-      AuthType.LOGIN_WITH_GOOGLE,
-      undefined,
-      'gemini-2.5-preview-pro',
-      DEFAULT_GEMINI_FLASH_MODEL,
-    );
-
-    expect(result25).toContain(
-      'You have reached your daily gemini-2.5-pro quota limit',
-    );
-    expect(resultPreview).toContain(
-      'You have reached your daily gemini-2.5-preview-pro quota limit',
-    );
-    expect(result25).toContain('upgrade to get higher limits');
-    expect(resultPreview).toContain('upgrade to get higher limits');
-  });
-
-  it('should not match non-Pro models with similar version strings', () => {
-    // Test that Flash models with similar version strings don't match
-    expect(
-      isProQuotaExceededError(
-        "Quota exceeded for quota metric 'Gemini 2.5 Flash Requests' and limit",
-      ),
-    ).toBe(false);
-    expect(
-      isProQuotaExceededError(
-        "Quota exceeded for quota metric 'Gemini 2.5-preview Flash Requests' and limit",
-      ),
-    ).toBe(false);
-
-    // Test other model types
-    expect(
-      isProQuotaExceededError(
-        "Quota exceeded for quota metric 'Gemini 2.5 Ultra Requests' and limit",
-      ),
-    ).toBe(false);
-    expect(
-      isProQuotaExceededError(
-        "Quota exceeded for quota metric 'Gemini 2.5 Standard Requests' and limit",
-      ),
-    ).toBe(false);
-
-    // Test generic quota messages
-    expect(
-      isProQuotaExceededError(
-        "Quota exceeded for quota metric 'GenerationRequests' and limit",
-      ),
-    ).toBe(false);
-    expect(
-      isProQuotaExceededError(
-        "Quota exceeded for quota metric 'EmbeddingRequests' and limit",
-      ),
-    ).toBe(false);
-  });
-
-  it('should format a generic quota exceeded message for Google auth (Standard tier)', () => {
-    const errorMessage =
-      'got status: 429 Too Many Requests. {"error":{"code":429,"message":"Quota exceeded for quota metric \'GenerationRequests\' and limit \'RequestsPerDay\' of service \'generativelanguage.googleapis.com\' for consumer \'project_number:123456789\'.","status":"RESOURCE_EXHAUSTED"}}';
-    const result = parseAndFormatApiError(
-      errorMessage,
-      AuthType.LOGIN_WITH_GOOGLE,
-      UserTierId.STANDARD,
-      'gemini-2.5-pro',
-      DEFAULT_GEMINI_FLASH_MODEL,
-    );
-    expect(result).toContain(
-      "[API Error: Quota exceeded for quota metric 'GenerationRequests'",
-    );
-    expect(result).toContain('You have reached your daily quota limit');
-    expect(result).toContain(
-      'We appreciate you for choosing Gemini Code Assist and the Gemini CLI',
-    );
-    expect(result).not.toContain('upgrade to get higher limits');
-  });
-
-  it('should format a regular 429 API error with standard message for Google auth (Standard tier)', () => {
-    const errorMessage =
-      'got status: 429 Too Many Requests. {"error":{"code":429,"message":"Rate limit exceeded","status":"RESOURCE_EXHAUSTED"}}';
-    const result = parseAndFormatApiError(
-      errorMessage,
-      AuthType.LOGIN_WITH_GOOGLE,
-      UserTierId.STANDARD,
-      'gemini-2.5-pro',
-      DEFAULT_GEMINI_FLASH_MODEL,
-    );
-    expect(result).toContain('[API Error: Rate limit exceeded');
-    expect(result).toContain(
-      'We appreciate you for choosing Gemini Code Assist and the Gemini CLI',
-    );
-    expect(result).not.toContain('upgrade to get higher limits');
-  });
 });
@@ -4,50 +4,11 @@
 * SPDX-License-Identifier: Apache-2.0
 */

-import {
-  isProQuotaExceededError,
-  isGenericQuotaExceededError,
-  isApiError,
-  isStructuredError,
-} from './quotaErrorDetection.js';
-import {
-  DEFAULT_GEMINI_MODEL,
-  DEFAULT_GEMINI_FLASH_MODEL,
-} from '../config/models.js';
-import { UserTierId } from '../code_assist/types.js';
+import { isApiError, isStructuredError } from './quotaErrorDetection.js';
+import { DEFAULT_GEMINI_FLASH_MODEL } from '../config/models.js';
+import type { UserTierId } from '../code_assist/types.js';
 import { AuthType } from '../core/contentGenerator.js';

-// Free Tier message functions
-const getRateLimitErrorMessageGoogleFree = (
-  fallbackModel: string = DEFAULT_GEMINI_FLASH_MODEL,
-) =>
-  `\nPossible quota limitations in place or slow response times detected. Switching to the ${fallbackModel} model for the rest of this session.`;
-
-const getRateLimitErrorMessageGoogleProQuotaFree = (
-  currentModel: string = DEFAULT_GEMINI_MODEL,
-  fallbackModel: string = DEFAULT_GEMINI_FLASH_MODEL,
-) =>
-  `\nYou have reached your daily ${currentModel} quota limit. You will be switched to the ${fallbackModel} model for the rest of this session. To increase your limits, upgrade to get higher limits at https://goo.gle/set-up-gemini-code-assist, or use /auth to switch to using a paid API key from AI Studio at https://aistudio.google.com/apikey`;
-
-const getRateLimitErrorMessageGoogleGenericQuotaFree = () =>
-  `\nYou have reached your daily quota limit. To increase your limits, upgrade to get higher limits at https://goo.gle/set-up-gemini-code-assist, or use /auth to switch to using a paid API key from AI Studio at https://aistudio.google.com/apikey`;
-
-// Legacy/Standard Tier message functions
-const getRateLimitErrorMessageGooglePaid = (
-  fallbackModel: string = DEFAULT_GEMINI_FLASH_MODEL,
-) =>
-  `\nPossible quota limitations in place or slow response times detected. Switching to the ${fallbackModel} model for the rest of this session. We appreciate you for choosing Gemini Code Assist and the Gemini CLI.`;
-
-const getRateLimitErrorMessageGoogleProQuotaPaid = (
-  currentModel: string = DEFAULT_GEMINI_MODEL,
-  fallbackModel: string = DEFAULT_GEMINI_FLASH_MODEL,
-) =>
-  `\nYou have reached your daily ${currentModel} quota limit. You will be switched to the ${fallbackModel} model for the rest of this session. We appreciate you for choosing Gemini Code Assist and the Gemini CLI. To continue accessing the ${currentModel} model today, consider using /auth to switch to using a paid API key from AI Studio at https://aistudio.google.com/apikey`;
-
-const getRateLimitErrorMessageGoogleGenericQuotaPaid = (
-  currentModel: string = DEFAULT_GEMINI_MODEL,
-) =>
-  `\nYou have reached your daily quota limit. We appreciate you for choosing Gemini Code Assist and the Gemini CLI. To continue accessing the ${currentModel} model today, consider using /auth to switch to using a paid API key from AI Studio at https://aistudio.google.com/apikey`;
 const RATE_LIMIT_ERROR_MESSAGE_USE_GEMINI =
  '\nPlease wait and try again later. To increase your limits, request a quota increase through AI Studio, or switch to another /auth method';
 const RATE_LIMIT_ERROR_MESSAGE_VERTEX =
@@ -59,39 +20,9 @@ const getRateLimitErrorMessageDefault = (

 function getRateLimitMessage(
  authType?: AuthType,
-  error?: unknown,
-  userTier?: UserTierId,
-  currentModel?: string,
  fallbackModel?: string,
 ): string {
  switch (authType) {
-    case AuthType.LOGIN_WITH_GOOGLE: {
-      // Determine if user is on a paid tier (Legacy or Standard) - default to FREE if not specified
-      const isPaidTier =
-        userTier === UserTierId.LEGACY || userTier === UserTierId.STANDARD;
-
-      if (isProQuotaExceededError(error)) {
-        return isPaidTier
-          ? getRateLimitErrorMessageGoogleProQuotaPaid(
-              currentModel || DEFAULT_GEMINI_MODEL,
-              fallbackModel,
-            )
-          : getRateLimitErrorMessageGoogleProQuotaFree(
-              currentModel || DEFAULT_GEMINI_MODEL,
-              fallbackModel,
-            );
-      } else if (isGenericQuotaExceededError(error)) {
-        return isPaidTier
-          ? getRateLimitErrorMessageGoogleGenericQuotaPaid(
-              currentModel || DEFAULT_GEMINI_MODEL,
-            )
-          : getRateLimitErrorMessageGoogleGenericQuotaFree();
-      } else {
-        return isPaidTier
-          ? getRateLimitErrorMessageGooglePaid(fallbackModel)
-          : getRateLimitErrorMessageGoogleFree(fallbackModel);
-      }
-    }
    case AuthType.USE_GEMINI:
      return RATE_LIMIT_ERROR_MESSAGE_USE_GEMINI;
    case AuthType.USE_VERTEX_AI:
@@ -111,13 +42,7 @@ export function parseAndFormatApiError(
  if (isStructuredError(error)) {
    let text = `[API Error: ${error.message}]`;
    if (error.status === 429) {
-      text += getRateLimitMessage(
-        authType,
-        error,
-        userTier,
-        currentModel,
-        fallbackModel,
-      );
+      text += getRateLimitMessage(authType, fallbackModel);
    }
    return text;
  }
@@ -146,13 +71,7 @@ export function parseAndFormatApiError(
        }
        let text = `[API Error: ${finalMessage} (Status: ${parsedError.error.status})]`;
        if (parsedError.error.code === 429) {
-          text += getRateLimitMessage(
-            authType,
-            parsedError,
-            userTier,
-            currentModel,
-            fallbackModel,
-          );
+          text += getRateLimitMessage(authType, fallbackModel);
        }
        return text;
      }
@@ -11,7 +11,6 @@ import {
  setSimulate429,
  disableSimulationAfterFallback,
  shouldSimulate429,
-  createSimulated429Error,
  resetRequestCounter,
 } from './testUtils.js';
 import { DEFAULT_GEMINI_FLASH_MODEL } from '../config/models.js';
@@ -19,12 +18,15 @@ import { retryWithBackoff } from './retry.js';
 import { AuthType } from '../core/contentGenerator.js';
 // Import the new types (Assuming this test file is in packages/core/src/utils/)
 import type { FallbackModelHandler } from '../fallback/types.js';
+import type { GoogleApiError } from './googleErrors.js';
+import { TerminalQuotaError } from './googleQuotaErrors.js';

 vi.mock('node:fs');

 // Update the description to reflect that this tests the retry utility's integration
 describe('Retry Utility Fallback Integration', () => {
  let config: Config;
+  let mockGoogleApiError: GoogleApiError;

  beforeEach(() => {
    vi.mocked(fs.existsSync).mockReturnValue(true);
@@ -38,6 +40,11 @@ describe('Retry Utility Fallback Integration', () => {
      cwd: '/test',
      model: 'gemini-2.5-pro',
    });
+    mockGoogleApiError = {
+      code: 429,
+      message: 'mock error',
+      details: [],
+    };

    // Reset simulation state for each test
    setSimulate429(false);
@@ -56,6 +63,7 @@ describe('Retry Utility Fallback Integration', () => {
    const result = await config.fallbackModelHandler!(
      'gemini-2.5-pro',
      DEFAULT_GEMINI_FLASH_MODEL,
+      new Error('test'),
    );

    // Verify it returns the correct intent
@@ -63,81 +71,61 @@ describe('Retry Utility Fallback Integration', () => {
  });

  // This test validates the retry utility's logic for triggering the callback.
-  it('should trigger onPersistent429 after 2 consecutive 429 errors for OAuth users', async () => {
+  it('should trigger onPersistent429 on TerminalQuotaError for OAuth users', async () => {
    let fallbackCalled = false;
-    // Removed fallbackModel variable as it's no longer relevant here.

-    // Mock function that simulates exactly 2 429 errors, then succeeds after fallback
    const mockApiCall = vi
      .fn()
-      .mockRejectedValueOnce(createSimulated429Error())
-      .mockRejectedValueOnce(createSimulated429Error())
+      .mockRejectedValueOnce(
+        new TerminalQuotaError('Daily limit', mockGoogleApiError),
+      )
+      .mockRejectedValueOnce(
+        new TerminalQuotaError('Daily limit', mockGoogleApiError),
+      )
      .mockResolvedValueOnce('success after fallback');

-    // Mock the onPersistent429 callback (this is what client.ts/geminiChat.ts provides)
    const mockPersistent429Callback = vi.fn(async (_authType?: string) => {
      fallbackCalled = true;
-      // Return true to signal retryWithBackoff to reset attempts and continue.
      return true;
    });

-    // Test with OAuth personal auth type, with maxAttempts = 2 to ensure fallback triggers
    const result = await retryWithBackoff(mockApiCall, {
      maxAttempts: 2,
      initialDelayMs: 1,
      maxDelayMs: 10,
-      shouldRetryOnError: (error: Error) => {
-        const status = (error as Error & { status?: number }).status;
-        return status === 429;
-      },
      onPersistent429: mockPersistent429Callback,
      authType: AuthType.LOGIN_WITH_GOOGLE,
    });

-    // Verify fallback mechanism was triggered
    expect(fallbackCalled).toBe(true);
    expect(mockPersistent429Callback).toHaveBeenCalledWith(
      AuthType.LOGIN_WITH_GOOGLE,
-      expect.any(Error),
+      expect.any(TerminalQuotaError),
    );
    expect(result).toBe('success after fallback');
-    // Should have: 2 failures, then fallback triggered, then 1 success after retry reset
    expect(mockApiCall).toHaveBeenCalledTimes(3);
  });

  it('should not trigger onPersistent429 for API key users', async () => {
-    let fallbackCalled = false;
+    const fallbackCallback = vi.fn();

-    // Mock function that simulates 429 errors
-    const mockApiCall = vi.fn().mockRejectedValue(createSimulated429Error());
+    const mockApiCall = vi
+      .fn()
+      .mockRejectedValueOnce(
+        new TerminalQuotaError('Daily limit', mockGoogleApiError),
+      );

-    // Mock the callback
-    const mockPersistent429Callback = vi.fn(async () => {
-      fallbackCalled = true;
-      return true;
+    const promise = retryWithBackoff(mockApiCall, {
+      maxAttempts: 2,
+      initialDelayMs: 1,
+      maxDelayMs: 10,
+      onPersistent429: fallbackCallback,
+      authType: AuthType.USE_GEMINI, // API key auth type
    });

-    // Test with API key auth type - should not trigger fallback
-    try {
-      await retryWithBackoff(mockApiCall, {
-        maxAttempts: 5,
-        initialDelayMs: 10,
-        maxDelayMs: 100,
-        shouldRetryOnError: (error: Error) => {
-          const status = (error as Error & { status?: number }).status;
-          return status === 429;
-        },
-        onPersistent429: mockPersistent429Callback,
-        authType: AuthType.USE_GEMINI, // API key auth type
-      });
-    } catch (error) {
-      // Expected to throw after max attempts
-      expect((error as Error).message).toContain('Rate limit exceeded');
-    }
-
-    // Verify fallback was NOT triggered for API key users
-    expect(fallbackCalled).toBe(false);
-    expect(mockPersistent429Callback).not.toHaveBeenCalled();
+    await expect(promise).rejects.toThrow('Daily limit');
+    expect(fallbackCallback).not.toHaveBeenCalled();
+    expect(mockApiCall).toHaveBeenCalledTimes(1);
  });

  // This test validates the test utilities themselves.
@@ -0,0 +1,356 @@
+/**
+ * @license
+ * Copyright 2025 Google LLC
+ * SPDX-License-Identifier: Apache-2.0
+ */
+
+import { describe, it, expect } from 'vitest';
+import { parseGoogleApiError } from './googleErrors.js';
+import type { QuotaFailure } from './googleErrors.js';
+
+describe('parseGoogleApiError', () => {
+  it('should return null for non-gaxios errors', () => {
+    expect(parseGoogleApiError(new Error('vanilla error'))).toBeNull();
+    expect(parseGoogleApiError(null)).toBeNull();
+    expect(parseGoogleApiError({})).toBeNull();
+  });
+
+  it('should parse a standard gaxios error', () => {
+    const mockError = {
+      response: {
+        status: 429,
+        data: {
+          error: {
+            code: 429,
+            message: 'Quota exceeded',
+            details: [
+              {
+                '@type': 'type.googleapis.com/google.rpc.QuotaFailure',
+                violations: [{ subject: 'user', description: 'daily limit' }],
+              },
+            ],
+          },
+        },
+      },
+    };
+
+    const parsed = parseGoogleApiError(mockError);
+    expect(parsed).not.toBeNull();
+    expect(parsed?.code).toBe(429);
+    expect(parsed?.message).toBe('Quota exceeded');
+    expect(parsed?.details).toHaveLength(1);
+    const detail = parsed?.details[0] as QuotaFailure;
+    expect(detail['@type']).toBe('type.googleapis.com/google.rpc.QuotaFailure');
+    expect(detail.violations[0].description).toBe('daily limit');
+  });
+
+  it('should parse an error with details stringified in the message', () => {
+    const innerError = {
+      error: {
+        code: 429,
+        message: 'Inner quota message',
+        details: [
+          {
+            '@type': 'type.googleapis.com/google.rpc.RetryInfo',
+            retryDelay: '10s',
+          },
+        ],
+      },
+    };
+
+    const mockError = {
+      response: {
+        status: 429,
+        data: {
+          error: {
+            code: 429,
+            message: JSON.stringify(innerError),
+            details: [], // Top-level details are empty
+          },
+        },
+      },
+    };
+
+    const parsed = parseGoogleApiError(mockError);
+    expect(parsed).not.toBeNull();
+    expect(parsed?.code).toBe(429);
+    expect(parsed?.message).toBe('Inner quota message');
+    expect(parsed?.details).toHaveLength(1);
+    expect(parsed?.details[0]['@type']).toBe(
+      'type.googleapis.com/google.rpc.RetryInfo',
+    );
+  });
+
+  it('should return null if details are not in the expected format', () => {
+    const mockError = {
+      response: {
+        status: 400,
+        data: {
+          error: {
+            code: 400,
+            message: 'Bad Request',
+            details: 'just a string', // Invalid details format
+          },
+        },
+      },
+    };
+    expect(parseGoogleApiError(mockError)).toBeNull();
+  });
+
+  it('should return null if there are no valid details', () => {
+    const mockError = {
+      response: {
+        status: 400,
+        data: {
+          error: {
+            code: 400,
+            message: 'Bad Request',
+            details: [
+              {
+                // missing '@type'
+                reason: 'some reason',
+              },
+            ],
+          },
+        },
+      },
+    };
+    expect(parseGoogleApiError(mockError)).toBeNull();
+  });
+
+  it('should parse a doubly nested error in the message', () => {
+    const innerError = {
+      error: {
+        code: 429,
+        message: 'Innermost quota message',
+        details: [
+          {
+            '@type': 'type.googleapis.com/google.rpc.RetryInfo',
+            retryDelay: '20s',
+          },
+        ],
+      },
+    };
+
+    const middleError = {
+      error: {
+        code: 429,
+        message: JSON.stringify(innerError),
+        details: [],
+      },
+    };
+
+    const mockError = {
+      response: {
+        status: 429,
+        data: {
+          error: {
+            code: 429,
+            message: JSON.stringify(middleError),
+            details: [],
+          },
+        },
+      },
+    };
+
+    const parsed = parseGoogleApiError(mockError);
+    expect(parsed).not.toBeNull();
+    expect(parsed?.code).toBe(429);
+    expect(parsed?.message).toBe('Innermost quota message');
+    expect(parsed?.details).toHaveLength(1);
+    expect(parsed?.details[0]['@type']).toBe(
+      'type.googleapis.com/google.rpc.RetryInfo',
+    );
+  });
+
+  it('should parse an error that is not in a response object', () => {
+    const innerError = {
+      error: {
+        code: 429,
+        message: 'Innermost quota message',
+        details: [
+          {
+            '@type': 'type.googleapis.com/google.rpc.RetryInfo',
+            retryDelay: '20s',
+          },
+        ],
+      },
+    };
+
+    const mockError = {
+      error: {
+        code: 429,
+        message: JSON.stringify(innerError),
+        details: [],
+      },
+    };
+
+    const parsed = parseGoogleApiError(mockError);
+    expect(parsed).not.toBeNull();
+    expect(parsed?.code).toBe(429);
+    expect(parsed?.message).toBe('Innermost quota message');
+    expect(parsed?.details).toHaveLength(1);
+    expect(parsed?.details[0]['@type']).toBe(
+      'type.googleapis.com/google.rpc.RetryInfo',
+    );
+  });
+
+  it('should parse an error that is a JSON string', () => {
+    const innerError = {
+      error: {
+        code: 429,
+        message: 'Innermost quota message',
+        details: [
+          {
+            '@type': 'type.googleapis.com/google.rpc.RetryInfo',
+            retryDelay: '20s',
+          },
+        ],
+      },
+    };
+
+    const mockError = {
+      error: {
+        code: 429,
+        message: JSON.stringify(innerError),
+        details: [],
+      },
+    };
+
+    const parsed = parseGoogleApiError(JSON.stringify(mockError));
+    expect(parsed).not.toBeNull();
+    expect(parsed?.code).toBe(429);
+    expect(parsed?.message).toBe('Innermost quota message');
+    expect(parsed?.details).toHaveLength(1);
+    expect(parsed?.details[0]['@type']).toBe(
+      'type.googleapis.com/google.rpc.RetryInfo',
+    );
+  });
+
+  it('should parse the user-provided nested error string', () => {
+    const userErrorString =
+      '{"error":{"message":"{\\n  \\"error\\": {\\n    \\"code\\": 429,\\n    \\"message\\": \\"You exceeded your current quota, please check your plan and billing details. For more information on this error, head to: https://ai.google.dev/gemini-api/docs/rate-limits.\\\\n* Quota exceeded for metric: generativelanguage.googleapis.com/generate_content_paid_tier_input_token_count, limit: 10000\\\\nPlease retry in 40.025771073s.\\",\\n    \\"status\\": \\"RESOURCE_EXHAUSTED\\",\\n    \\"details\\": [\\n      {\\n        \\"@type\\": \\"type.googleapis.com/google.rpc.DebugInfo\\",\\n        \\"detail\\": \\"[ORIGINAL ERROR] generic::resource_exhausted: You exceeded your current quota, please check your plan and billing details. For more information on this error, head to: https://ai.google.dev/gemini-api/docs/rate-limits.\\\\n* Quota exceeded for metric: generativelanguage.googleapis.com/generate_content_paid_tier_input_token_count, limit: 10000\\\\nPlease retry in 40.025771073s. [google.rpc.error_details_ext] { message: \\\\\\"You exceeded your current quota, please check your plan and billing details. For more information on this error, head to: https://ai.google.dev/gemini-api/docs/rate-limits.\\\\\\\\n* Quota exceeded for metric: generativelanguage.googleapis.com/generate_content_paid_tier_input_token_count, limit: 10000\\\\\\\\nPlease retry in 40.025771073s.\\\\\\" }\\"\\n      },\\n      {\\n        \\"@type\\": \\"type.googleapis.com/google.rpc.QuotaFailure\\",\\n        \\"violations\\": [\\n          {\\n            \\"quotaMetric\\": \\"generativelanguage.googleapis.com/generate_content_paid_tier_input_token_count\\",\\n            \\"quotaId\\": \\"GenerateContentPaidTierInputTokensPerModelPerMinute\\",\\n            \\"quotaDimensions\\": {\\n              \\"location\\": \\"global\\",\\n              \\"model\\": \\"gemini-2.5-pro\\"\\n            },\\n            \\"quotaValue\\": \\"10000\\"\\n          }\\n        ]\\n      },\\n      {\\n        \\"@type\\": \\"type.googleapis.com/google.rpc.Help\\",\\n        \\"links\\": [\\n          {\\n            \\"description\\": \\"Learn more about Gemini API quotas\\",\\n            \\"url\\": \\"https://ai.google.dev/gemini-api/docs/rate-limits\\"\\n          }\\n        ]\\n      },\\n      {\\n        \\"@type\\": \\"type.googleapis.com/google.rpc.RetryInfo\\",\\n        \\"retryDelay\\": \\"40s\\"\\n      }\\n    ]\\n  }\\n}\\n","code":429,"status":"Too Many Requests"}}';
+
+    const parsed = parseGoogleApiError(userErrorString);
+    expect(parsed).not.toBeNull();
+    expect(parsed?.code).toBe(429);
+    expect(parsed?.message).toContain('You exceeded your current quota');
+    expect(parsed?.details).toHaveLength(4);
+    expect(
+      parsed?.details.some(
+        (d) => d['@type'] === 'type.googleapis.com/google.rpc.QuotaFailure',
+      ),
+    ).toBe(true);
+    expect(
+      parsed?.details.some(
+        (d) => d['@type'] === 'type.googleapis.com/google.rpc.RetryInfo',
+      ),
+    ).toBe(true);
+  });
+
+  it('should parse an error that is an array', () => {
+    const mockError = [
+      {
+        error: {
+          code: 429,
+          message: 'Quota exceeded',
+          details: [
+            {
+              '@type': 'type.googleapis.com/google.rpc.QuotaFailure',
+              violations: [{ subject: 'user', description: 'daily limit' }],
+            },
+          ],
+        },
+      },
+    ];
+
+    const parsed = parseGoogleApiError(mockError);
+    expect(parsed).not.toBeNull();
+    expect(parsed?.code).toBe(429);
+    expect(parsed?.message).toBe('Quota exceeded');
+  });
+
+  it('should parse a gaxios error where data is an array', () => {
+    const mockError = {
+      response: {
+        status: 429,
+        data: [
+          {
+            error: {
+              code: 429,
+              message: 'Quota exceeded',
+              details: [
+                {
+                  '@type': 'type.googleapis.com/google.rpc.QuotaFailure',
+                  violations: [{ subject: 'user', description: 'daily limit' }],
+                },
+              ],
+            },
+          },
+        ],
+      },
+    };
+
+    const parsed = parseGoogleApiError(mockError);
+    expect(parsed).not.toBeNull();
+    expect(parsed?.code).toBe(429);
+    expect(parsed?.message).toBe('Quota exceeded');
+  });
+
+  it('should parse a gaxios error where data is a stringified array', () => {
+    const mockError = {
+      response: {
+        status: 429,
+        data: JSON.stringify([
+          {
+            error: {
+              code: 429,
+              message: 'Quota exceeded',
+              details: [
+                {
+                  '@type': 'type.googleapis.com/google.rpc.QuotaFailure',
+                  violations: [{ subject: 'user', description: 'daily limit' }],
+                },
+              ],
+            },
+          },
+        ]),
+      },
+    };
+
+    const parsed = parseGoogleApiError(mockError);
+    expect(parsed).not.toBeNull();
+    expect(parsed?.code).toBe(429);
+    expect(parsed?.message).toBe('Quota exceeded');
+  });
+
+  it('should parse an error with a malformed @type key (returned by Gemini API)', () => {
+    const malformedError = {
+      name: 'API Error',
+      message: {
+        error: {
+          message:
+            '{\n  "error": {\n    "code": 429,\n    "message": "You exceeded your current quota, please check your plan and billing details. For more information on this error, head to: https://ai.google.dev/gemini-api/docs/rate-limits.\\n* Quota exceeded for metric: generativelanguage.googleapis.com/generate_content_free_tier_requests, limit: 2\nPlease retry in 54.887755558s.",\n    "status": "RESOURCE_EXHAUSTED",\n    "details": [\n      {\n        " @type": "type.googleapis.com/google.rpc.DebugInfo",\n        "detail": "[ORIGINAL ERROR] generic::resource_exhausted: You exceeded your current quota, please check your plan and billing details. For more information on this error, head to: https://ai.google.dev/gemini-api/docs/rate-limits.\\n* Quota exceeded for metric: generativelanguage.googleapis.com/generate_content_free_tier_requests, limit: 2\\nPlease retry in 54.887755558s. [google.rpc.error_details_ext] { message: \\"You exceeded your current quota, please check your plan and billing details. For more information on this error, head to: https://ai.google.dev/gemini-api/docs/rate-limits.\\\\n* Quota exceeded for metric: generativelanguage.googleapis.com/generate_content_free_tier_requests, limit: 2\\\\nPlease retry in 54.887755558s.\\" }"\n      },\n      {\n" @type": "type.googleapis.com/google.rpc.QuotaFailure",\n        "violations": [\n          {\n            "quotaMetric": "generativelanguage.googleapis.com/generate_content_free_tier_requests",\n            "quotaId": "GenerateRequestsPerMinutePerProjectPerModel-FreeTier",\n            "quotaDimensions": {\n              "location": "global",\n"model": "gemini-2.5-pro"\n            },\n            "quotaValue": "2"\n          }\n        ]\n      },\n      {\n" @type": "type.googleapis.com/google.rpc.Help",\n        "links": [\n          {\n            "description": "Learn more about Gemini API quotas",\n            "url": "https://ai.google.dev/gemini-api/docs/rate-limits"\n          }\n        ]\n      },\n      {\n" @type": "type.googleapis.com/google.rpc.RetryInfo",\n        "retryDelay": "54s"\n      }\n    ]\n  }\n}\n',
+          code: 429,
+          status: 'Too Many Requests',
+        },
+      },
+    };
+
+    const parsed = parseGoogleApiError(malformedError);
+    expect(parsed).not.toBeNull();
+    expect(parsed?.code).toBe(429);
+    expect(parsed?.message).toContain('You exceeded your current quota');
+    expect(parsed?.details).toHaveLength(4);
+    expect(
+      parsed?.details.some(
+        (d) => d['@type'] === 'type.googleapis.com/google.rpc.QuotaFailure',
+      ),
+    ).toBe(true);
+    expect(
+      parsed?.details.some(
+        (d) => d['@type'] === 'type.googleapis.com/google.rpc.RetryInfo',
+      ),
+    ).toBe(true);
+  });
+});
@@ -0,0 +1,305 @@
+/**
+ * @license
+ * Copyright 2025 Google LLC
+ * SPDX-License-Identifier: Apache-2.0
+ */
+
+/**
+ * @fileoverview
+ * This file contains types and functions for parsing structured Google API errors.
+ */
+
+/**
+ * Based on google/rpc/error_details.proto
+ */
+
+export interface ErrorInfo {
+  '@type': 'type.googleapis.com/google.rpc.ErrorInfo';
+  reason: string;
+  domain: string;
+  metadata: { [key: string]: string };
+}
+
+export interface RetryInfo {
+  '@type': 'type.googleapis.com/google.rpc.RetryInfo';
+  retryDelay: string; // e.g. "51820.638305887s"
+}
+
+export interface DebugInfo {
+  '@type': 'type.googleapis.com/google.rpc.DebugInfo';
+  stackEntries: string[];
+  detail: string;
+}
+
+export interface QuotaFailure {
+  '@type': 'type.googleapis.com/google.rpc.QuotaFailure';
+  violations: Array<{
+    subject?: string;
+    description?: string;
+    apiService?: string;
+    quotaMetric?: string;
+    quotaId?: string;
+    quotaDimensions?: { [key: string]: string };
+    quotaValue?: string | number;
+    futureQuotaValue?: number;
+  }>;
+}
+
+export interface PreconditionFailure {
+  '@type': 'type.googleapis.com/google.rpc.PreconditionFailure';
+  violations: Array<{
+    type: string;
+    subject: string;
+    description: string;
+  }>;
+}
+
+export interface LocalizedMessage {
+  '@type': 'type.googleapis.com/google.rpc.LocalizedMessage';
+  locale: string;
+  message: string;
+}
+
+export interface BadRequest {
+  '@type': 'type.googleapis.com/google.rpc.BadRequest';
+  fieldViolations: Array<{
+    field: string;
+    description: string;
+    reason?: string;
+    localizedMessage?: LocalizedMessage;
+  }>;
+}
+
+export interface RequestInfo {
+  '@type': 'type.googleapis.com/google.rpc.RequestInfo';
+  requestId: string;
+  servingData: string;
+}
+
+export interface ResourceInfo {
+  '@type': 'type.googleapis.com/google.rpc.ResourceInfo';
+  resourceType: string;
+  resourceName: string;
+  owner: string;
+  description: string;
+}
+
+export interface Help {
+  '@type': 'type.googleapis.com/google.rpc.Help';
+  links: Array<{
+    description: string;
+    url: string;
+  }>;
+}
+
+export type GoogleApiErrorDetail =
+  | ErrorInfo
+  | RetryInfo
+  | DebugInfo
+  | QuotaFailure
+  | PreconditionFailure
+  | BadRequest
+  | RequestInfo
+  | ResourceInfo
+  | Help
+  | LocalizedMessage;
+
+export interface GoogleApiError {
+  code: number;
+  message: string;
+  details: GoogleApiErrorDetail[];
+}
+
+type ErrorShape = {
+  message?: string;
+  details?: unknown[];
+  code?: number;
+};
+
+/**
+ * Parses an error object to check if it's a structured Google API error
+ * and extracts all details.
+ *
+ * This function can handle two formats:
+ * 1. Standard Google API errors where `details` is a top-level field.
+ * 2. Errors where the entire structured error object is stringified inside
+ *    the `message` field of a wrapper error.
+ *
+ * @param error The error object to inspect.
+ * @returns A GoogleApiError object if the error matches, otherwise null.
+ */
+export function parseGoogleApiError(error: unknown): GoogleApiError | null {
+  if (!error) {
+    return null;
+  }
+
+  let errorObj: unknown = error;
+
+  // If error is a string, try to parse it.
+  if (typeof errorObj === 'string') {
+    try {
+      errorObj = JSON.parse(errorObj);
+    } catch (_) {
+      // Not a JSON string, can't parse.
+      return null;
+    }
+  }
+
+  if (Array.isArray(errorObj) && errorObj.length > 0) {
+    errorObj = errorObj[0];
+  }
+
+  if (typeof errorObj !== 'object' || errorObj === null) {
+    return null;
+  }
+
+  let currentError: ErrorShape | undefined =
+    fromGaxiosError(errorObj) ?? fromApiError(errorObj);
+
+  let depth = 0;
+  const maxDepth = 10;
+  // Handle cases where the actual error object is stringified inside the message
+  // by drilling down until we find an error that doesn't have a stringified message.
+  while (
+    currentError &&
+    typeof currentError.message === 'string' &&
+    depth < maxDepth
+  ) {
+    try {
+      const parsedMessage = JSON.parse(
+        currentError.message.replace(/\u00A0/g, '').replace(/\n/g, ' '),
+      );
+      if (parsedMessage.error) {
+        currentError = parsedMessage.error;
+        depth++;
+      } else {
+        // The message is a JSON string, but not a nested error object.
+        break;
+      }
+    } catch (_error) {
+      // It wasn't a JSON string, so we've drilled down as far as we can.
+      break;
+    }
+  }
+
+  if (!currentError) {
+    return null;
+  }
+
+  const code = currentError.code;
+  const message = currentError.message;
+  const errorDetails = currentError.details;
+
+  if (Array.isArray(errorDetails) && code && message) {
+    const details: GoogleApiErrorDetail[] = [];
+    for (const detail of errorDetails) {
+      if (detail && typeof detail === 'object') {
+        const detailObj = detail as Record<string, unknown>;
+        const typeKey = Object.keys(detailObj).find(
+          (key) => key.trim() === '@type',
+        );
+        if (typeKey) {
+          if (typeKey !== '@type') {
+            detailObj['@type'] = detailObj[typeKey];
+            delete detailObj[typeKey];
+          }
+          // We can just cast it; the consumer will have to switch on @type
+          details.push(detailObj as unknown as GoogleApiErrorDetail);
+        }
+      }
+    }
+
+    if (details.length > 0) {
+      return {
+        code,
+        message,
+        details,
+      };
+    }
+  }
+
+  return null;
+}
+
+function fromGaxiosError(errorObj: object): ErrorShape | undefined {
+  const gaxiosError = errorObj as {
+    response?: {
+      status?: number;
+      data?:
+        | {
+            error?: ErrorShape;
+          }
+        | string;
+    };
+    error?: ErrorShape;
+    code?: number;
+  };
+
+  let outerError: ErrorShape | undefined;
+  if (gaxiosError.response?.data) {
+    let data = gaxiosError.response.data;
+
+    if (typeof data === 'string') {
+      try {
+        data = JSON.parse(data);
+      } catch (_) {
+        // Not a JSON string, can't parse.
+      }
+    }
+
+    if (Array.isArray(data) && data.length > 0) {
+      data = data[0];
+    }
+
+    if (typeof data === 'object' && data !== null) {
+      if ('error' in data) {
+        outerError = (data as { error: ErrorShape }).error;
+      }
+    }
+  }
+
+  if (!outerError) {
+    // If the gaxios structure isn't there, check for a top-level `error` property.
+    if (gaxiosError.error) {
+      outerError = gaxiosError.error;
+    } else {
+      return undefined;
+    }
+  }
+  return outerError;
+}
+
+function fromApiError(errorObj: object): ErrorShape | undefined {
+  const apiError = errorObj as {
+    message?:
+      | {
+          error?: ErrorShape;
+        }
+      | string;
+    code?: number;
+  };
+
+  let outerError: ErrorShape | undefined;
+  if (apiError.message) {
+    let data = apiError.message;
+
+    if (typeof data === 'string') {
+      try {
+        data = JSON.parse(data);
+      } catch (_) {
+        // Not a JSON string, can't parse.
+      }
+    }
+
+    if (Array.isArray(data) && data.length > 0) {
+      data = data[0];
+    }
+
+    if (typeof data === 'object' && data !== null) {
+      if ('error' in data) {
+        outerError = (data as { error: ErrorShape }).error;
+      }
+    }
+  }
+  return outerError;
+}
@@ -0,0 +1,306 @@
+/**
+ * @license
+ * Copyright 2025 Google LLC
+ * SPDX-License-Identifier: Apache-2.0
+ */
+
+import { describe, it, expect, vi, afterEach } from 'vitest';
+import {
+  classifyGoogleError,
+  RetryableQuotaError,
+  TerminalQuotaError,
+} from './googleQuotaErrors.js';
+import * as errorParser from './googleErrors.js';
+import type { GoogleApiError } from './googleErrors.js';
+
+describe('classifyGoogleError', () => {
+  afterEach(() => {
+    vi.restoreAllMocks();
+  });
+
+  it('should return original error if not a Google API error', () => {
+    const regularError = new Error('Something went wrong');
+    vi.spyOn(errorParser, 'parseGoogleApiError').mockReturnValue(null);
+    const result = classifyGoogleError(regularError);
+    expect(result).toBe(regularError);
+  });
+
+  it('should return original error if code is not 429', () => {
+    const apiError: GoogleApiError = {
+      code: 500,
+      message: 'Server error',
+      details: [],
+    };
+    vi.spyOn(errorParser, 'parseGoogleApiError').mockReturnValue(apiError);
+    const originalError = new Error();
+    const result = classifyGoogleError(originalError);
+    expect(result).toBe(originalError);
+    expect(result).not.toBeInstanceOf(TerminalQuotaError);
+    expect(result).not.toBeInstanceOf(RetryableQuotaError);
+  });
+
+  it('should return TerminalQuotaError for daily quota violations in QuotaFailure', () => {
+    const apiError: GoogleApiError = {
+      code: 429,
+      message: 'Quota exceeded',
+      details: [
+        {
+          '@type': 'type.googleapis.com/google.rpc.QuotaFailure',
+          violations: [
+            {
+              subject: 'user',
+              description: 'daily limit',
+              quotaId: 'RequestsPerDay-limit',
+            },
+          ],
+        },
+      ],
+    };
+    vi.spyOn(errorParser, 'parseGoogleApiError').mockReturnValue(apiError);
+    const result = classifyGoogleError(new Error());
+    expect(result).toBeInstanceOf(TerminalQuotaError);
+    expect((result as TerminalQuotaError).cause).toBe(apiError);
+  });
+
+  it('should return TerminalQuotaError for daily quota violations in ErrorInfo', () => {
+    const apiError: GoogleApiError = {
+      code: 429,
+      message: 'Quota exceeded',
+      details: [
+        {
+          '@type': 'type.googleapis.com/google.rpc.ErrorInfo',
+          reason: 'QUOTA_EXCEEDED',
+          domain: 'googleapis.com',
+          metadata: {
+            quota_limit: 'RequestsPerDay_PerProject_PerUser',
+          },
+        },
+      ],
+    };
+    vi.spyOn(errorParser, 'parseGoogleApiError').mockReturnValue(apiError);
+    const result = classifyGoogleError(new Error());
+    expect(result).toBeInstanceOf(TerminalQuotaError);
+  });
+
+  it('should return TerminalQuotaError for long retry delays', () => {
+    const apiError: GoogleApiError = {
+      code: 429,
+      message: 'Too many requests',
+      details: [
+        {
+          '@type': 'type.googleapis.com/google.rpc.RetryInfo',
+          retryDelay: '301s', // > 5 minutes
+        },
+      ],
+    };
+    vi.spyOn(errorParser, 'parseGoogleApiError').mockReturnValue(apiError);
+    const result = classifyGoogleError(new Error());
+    expect(result).toBeInstanceOf(TerminalQuotaError);
+  });
+
+  it('should return RetryableQuotaError for short retry delays', () => {
+    const apiError: GoogleApiError = {
+      code: 429,
+      message: 'Too many requests',
+      details: [
+        {
+          '@type': 'type.googleapis.com/google.rpc.RetryInfo',
+          retryDelay: '45.123s',
+        },
+      ],
+    };
+    vi.spyOn(errorParser, 'parseGoogleApiError').mockReturnValue(apiError);
+    const result = classifyGoogleError(new Error());
+    expect(result).toBeInstanceOf(RetryableQuotaError);
+    expect((result as RetryableQuotaError).retryDelayMs).toBe(45123);
+  });
+
+  it('should return RetryableQuotaError for per-minute quota violations in QuotaFailure', () => {
+    const apiError: GoogleApiError = {
+      code: 429,
+      message: 'Quota exceeded',
+      details: [
+        {
+          '@type': 'type.googleapis.com/google.rpc.QuotaFailure',
+          violations: [
+            {
+              subject: 'user',
+              description: 'per minute limit',
+              quotaId: 'RequestsPerMinute-limit',
+            },
+          ],
+        },
+      ],
+    };
+    vi.spyOn(errorParser, 'parseGoogleApiError').mockReturnValue(apiError);
+    const result = classifyGoogleError(new Error());
+    expect(result).toBeInstanceOf(RetryableQuotaError);
+    expect((result as RetryableQuotaError).retryDelayMs).toBe(60000);
+  });
+
+  it('should return RetryableQuotaError for per-minute quota violations in ErrorInfo', () => {
+    const apiError: GoogleApiError = {
+      code: 429,
+      message: 'Quota exceeded',
+      details: [
+        {
+          '@type': 'type.googleapis.com/google.rpc.ErrorInfo',
+          reason: 'QUOTA_EXCEEDED',
+          domain: 'googleapis.com',
+          metadata: {
+            quota_limit: 'RequestsPerMinute_PerProject_PerUser',
+          },
+        },
+      ],
+    };
+    vi.spyOn(errorParser, 'parseGoogleApiError').mockReturnValue(apiError);
+    const result = classifyGoogleError(new Error());
+    expect(result).toBeInstanceOf(RetryableQuotaError);
+    expect((result as RetryableQuotaError).retryDelayMs).toBe(60000);
+  });
+
+  it('should return RetryableQuotaError for another short retry delay', () => {
+    const apiError: GoogleApiError = {
+      code: 429,
+      message:
+        'You exceeded your current quota, please check your plan and billing details. For more information on this error, head to: https://ai.google.dev/gemini-api/docs/rate-limits.\n* Quota exceeded for metric: generativelanguage.googleapis.com/generate_content_free_tier_requests, limit: 2\nPlease retry in 56.185908122s.',
+      details: [
+        {
+          '@type': 'type.googleapis.com/google.rpc.QuotaFailure',
+          violations: [
+            {
+              quotaMetric:
+                'generativelanguage.googleapis.com/generate_content_free_tier_requests',
+              quotaId: 'GenerateRequestsPerMinutePerProjectPerModel-FreeTier',
+              quotaDimensions: {
+                location: 'global',
+                model: 'gemini-2.5-pro',
+              },
+              quotaValue: '2',
+            },
+          ],
+        },
+        {
+          '@type': 'type.googleapis.com/google.rpc.Help',
+          links: [
+            {
+              description: 'Learn more about Gemini API quotas',
+              url: 'https://ai.google.dev/gemini-api/docs/rate-limits',
+            },
+          ],
+        },
+        {
+          '@type': 'type.googleapis.com/google.rpc.RetryInfo',
+          retryDelay: '56s',
+        },
+      ],
+    };
+    vi.spyOn(errorParser, 'parseGoogleApiError').mockReturnValue(apiError);
+    const result = classifyGoogleError(new Error());
+    expect(result).toBeInstanceOf(RetryableQuotaError);
+    expect((result as RetryableQuotaError).retryDelayMs).toBe(56000);
+  });
+
+  it('should return RetryableQuotaError for Cloud Code RATE_LIMIT_EXCEEDED with retry delay', () => {
+    const apiError: GoogleApiError = {
+      code: 429,
+      message:
+        'You have exhausted your capacity on this model. Your quota will reset after 0s.',
+      details: [
+        {
+          '@type': 'type.googleapis.com/google.rpc.ErrorInfo',
+          reason: 'RATE_LIMIT_EXCEEDED',
+          domain: 'cloudcode-pa.googleapis.com',
+          metadata: {
+            uiMessage: 'true',
+            model: 'gemini-2.5-pro',
+            quotaResetDelay: '539.477544ms',
+            quotaResetTimeStamp: '2025-10-20T19:14:08Z',
+          },
+        },
+        {
+          '@type': 'type.googleapis.com/google.rpc.RetryInfo',
+          retryDelay: '0.539477544s',
+        },
+      ],
+    };
+    vi.spyOn(errorParser, 'parseGoogleApiError').mockReturnValue(apiError);
+    const result = classifyGoogleError(new Error());
+    expect(result).toBeInstanceOf(RetryableQuotaError);
+    expect((result as RetryableQuotaError).retryDelayMs).toBeCloseTo(
+      539.477544,
+    );
+  });
+
+  it('should return TerminalQuotaError for Cloud Code QUOTA_EXHAUSTED', () => {
+    const apiError: GoogleApiError = {
+      code: 429,
+      message:
+        'You have exhausted your capacity on this model. Your quota will reset after 0s.',
+      details: [
+        {
+          '@type': 'type.googleapis.com/google.rpc.ErrorInfo',
+          reason: 'QUOTA_EXHAUSTED',
+          domain: 'cloudcode-pa.googleapis.com',
+          metadata: {
+            uiMessage: 'true',
+            model: 'gemini-2.5-pro',
+            quotaResetDelay: '539.477544ms',
+            quotaResetTimeStamp: '2025-10-20T19:14:08Z',
+          },
+        },
+        {
+          '@type': 'type.googleapis.com/google.rpc.RetryInfo',
+          retryDelay: '0.539477544s',
+        },
+      ],
+    };
+    vi.spyOn(errorParser, 'parseGoogleApiError').mockReturnValue(apiError);
+    const result = classifyGoogleError(new Error());
+    expect(result).toBeInstanceOf(TerminalQuotaError);
+  });
+
+  it('should prioritize daily limit over retry info', () => {
+    const apiError: GoogleApiError = {
+      code: 429,
+      message: 'Quota exceeded',
+      details: [
+        {
+          '@type': 'type.googleapis.com/google.rpc.QuotaFailure',
+          violations: [
+            {
+              subject: 'user',
+              description: 'daily limit',
+              quotaId: 'RequestsPerDay-limit',
+            },
+          ],
+        },
+        {
+          '@type': 'type.googleapis.com/google.rpc.RetryInfo',
+          retryDelay: '10s',
+        },
+      ],
+    };
+    vi.spyOn(errorParser, 'parseGoogleApiError').mockReturnValue(apiError);
+    const result = classifyGoogleError(new Error());
+    expect(result).toBeInstanceOf(TerminalQuotaError);
+  });
+
+  it('should return original error for 429 without specific details', () => {
+    const apiError: GoogleApiError = {
+      code: 429,
+      message: 'Too many requests',
+      details: [
+        {
+          '@type': 'type.googleapis.com/google.rpc.DebugInfo',
+          detail: 'some debug info',
+          stackEntries: [],
+        },
+      ],
+    };
+    vi.spyOn(errorParser, 'parseGoogleApiError').mockReturnValue(apiError);
+    const originalError = new Error();
+    const result = classifyGoogleError(originalError);
+    expect(result).toBe(originalError);
+  });
+});
@@ -0,0 +1,192 @@
+/**
+ * @license
+ * Copyright 2025 Google LLC
+ * SPDX-License-Identifier: Apache-2.0
+ */
+
+import type {
+  ErrorInfo,
+  GoogleApiError,
+  QuotaFailure,
+  RetryInfo,
+} from './googleErrors.js';
+import { parseGoogleApiError } from './googleErrors.js';
+
+/**
+ * A non-retryable error indicating a hard quota limit has been reached (e.g., daily limit).
+ */
+export class TerminalQuotaError extends Error {
+  constructor(
+    message: string,
+    override readonly cause: GoogleApiError,
+  ) {
+    super(message);
+    this.name = 'TerminalQuotaError';
+  }
+}
+
+/**
+ * A retryable error indicating a temporary quota issue (e.g., per-minute limit).
+ */
+export class RetryableQuotaError extends Error {
+  retryDelayMs: number;
+
+  constructor(
+    message: string,
+    override readonly cause: GoogleApiError,
+    retryDelaySeconds: number,
+  ) {
+    super(message);
+    this.name = 'RetryableQuotaError';
+    this.retryDelayMs = retryDelaySeconds * 1000;
+  }
+}
+
+/**
+ * Parses a duration string (e.g., "34.074824224s", "60s") and returns the time in seconds.
+ * @param duration The duration string to parse.
+ * @returns The duration in seconds, or null if parsing fails.
+ */
+function parseDurationInSeconds(duration: string): number | null {
+  if (!duration.endsWith('s')) {
+    return null;
+  }
+  const seconds = parseFloat(duration.slice(0, -1));
+  return isNaN(seconds) ? null : seconds;
+}
+
+/**
+ * Analyzes a caught error and classifies it as a specific quota-related error if applicable.
+ *
+ * It decides whether an error is a `TerminalQuotaError` or a `RetryableQuotaError` based on
+ * the following logic:
+ * - If the error indicates a daily limit, it's a `TerminalQuotaError`.
+ * - If the error suggests a retry delay of more than 2 minutes, it's a `TerminalQuotaError`.
+ * - If the error suggests a retry delay of 2 minutes or less, it's a `RetryableQuotaError`.
+ * - If the error indicates a per-minute limit, it's a `RetryableQuotaError`.
+ *
+ * @param error The error to classify.
+ * @returns A `TerminalQuotaError`, `RetryableQuotaError`, or the original `unknown` error.
+ */
+export function classifyGoogleError(error: unknown): unknown {
+  const googleApiError = parseGoogleApiError(error);
+
+  if (!googleApiError || googleApiError.code !== 429) {
+    return error; // Not a 429 error we can handle.
+  }
+
+  const quotaFailure = googleApiError.details.find(
+    (d): d is QuotaFailure =>
+      d['@type'] === 'type.googleapis.com/google.rpc.QuotaFailure',
+  );
+
+  const errorInfo = googleApiError.details.find(
+    (d): d is ErrorInfo =>
+      d['@type'] === 'type.googleapis.com/google.rpc.ErrorInfo',
+  );
+
+  const retryInfo = googleApiError.details.find(
+    (d): d is RetryInfo =>
+      d['@type'] === 'type.googleapis.com/google.rpc.RetryInfo',
+  );
+
+  // 1. Check for long-term limits in QuotaFailure or ErrorInfo
+  if (quotaFailure) {
+    for (const violation of quotaFailure.violations) {
+      const quotaId = violation.quotaId ?? '';
+      if (quotaId.includes('PerDay') || quotaId.includes('Daily')) {
+        return new TerminalQuotaError(
+          `You have exhausted your daily quota on this model.`,
+          googleApiError,
+        );
+      }
+    }
+  }
+
+  if (errorInfo) {
+    // New Cloud Code API quota handling
+    if (errorInfo.domain) {
+      const validDomains = [
+        'cloudcode-pa.googleapis.com',
+        'staging-cloudcode-pa.googleapis.com',
+        'autopush-cloudcode-pa.googleapis.com',
+      ];
+      if (validDomains.includes(errorInfo.domain)) {
+        if (errorInfo.reason === 'RATE_LIMIT_EXCEEDED') {
+          let delaySeconds = 10; // Default retry of 10s
+          if (retryInfo?.retryDelay) {
+            const parsedDelay = parseDurationInSeconds(retryInfo.retryDelay);
+            if (parsedDelay) {
+              delaySeconds = parsedDelay;
+            }
+          }
+          return new RetryableQuotaError(
+            `${googleApiError.message}`,
+            googleApiError,
+            delaySeconds,
+          );
+        }
+        if (errorInfo.reason === 'QUOTA_EXHAUSTED') {
+          return new TerminalQuotaError(
+            `${googleApiError.message}`,
+            googleApiError,
+          );
+        }
+      }
+    }
+
+    // Existing Cloud Code API quota handling
+    const quotaLimit = errorInfo.metadata?.['quota_limit'] ?? '';
+    if (quotaLimit.includes('PerDay') || quotaLimit.includes('Daily')) {
+      return new TerminalQuotaError(
+        `You have exhausted your daily quota on this model.`,
+        googleApiError,
+      );
+    }
+  }
+
+  // 2. Check for long delays in RetryInfo
+  if (retryInfo?.retryDelay) {
+    const delaySeconds = parseDurationInSeconds(retryInfo.retryDelay);
+    if (delaySeconds) {
+      if (delaySeconds > 120) {
+        return new TerminalQuotaError(
+          `${googleApiError.message}\nSuggested retry after ${retryInfo.retryDelay}.`,
+          googleApiError,
+        );
+      }
+      // This is a retryable error with a specific delay.
+      return new RetryableQuotaError(
+        `${googleApiError.message}\nSuggested retry after ${retryInfo.retryDelay}.`,
+        googleApiError,
+        delaySeconds,
+      );
+    }
+  }
+
+  // 3. Check for short-term limits in QuotaFailure or ErrorInfo
+  if (quotaFailure) {
+    for (const violation of quotaFailure.violations) {
+      const quotaId = violation.quotaId ?? '';
+      if (quotaId.includes('PerMinute')) {
+        return new RetryableQuotaError(
+          `${googleApiError.message}\nSuggested retry after 60s.`,
+          googleApiError,
+          60,
+        );
+      }
+    }
+  }
+
+  if (errorInfo) {
+    const quotaLimit = errorInfo.metadata?.['quota_limit'] ?? '';
+    if (quotaLimit.includes('PerMinute')) {
+      return new RetryableQuotaError(
+        `${errorInfo.reason}\nSuggested retry after 60s.`,
+        googleApiError,
+        60,
+      );
+    }
+  }
+  return error; // Fallback to original error if no specific classification fits.
+}
@@ -33,68 +33,3 @@ export function isStructuredError(error: unknown): error is StructuredError {
    typeof (error as StructuredError).message === 'string'
  );
 }
-
-export function isProQuotaExceededError(error: unknown): boolean {
-  // Check for Pro quota exceeded errors by looking for the specific pattern
-  // This will match patterns like:
-  // - "Quota exceeded for quota metric 'Gemini 2.5 Pro Requests'"
-  // - "Quota exceeded for quota metric 'Gemini 2.5-preview Pro Requests'"
-  // We use string methods instead of regex to avoid ReDoS vulnerabilities
-
-  const checkMessage = (message: string): boolean =>
-    message.includes("Quota exceeded for quota metric 'Gemini") &&
-    message.includes("Pro Requests'");
-
-  if (typeof error === 'string') {
-    return checkMessage(error);
-  }
-
-  if (isStructuredError(error)) {
-    return checkMessage(error.message);
-  }
-
-  if (isApiError(error)) {
-    return checkMessage(error.error.message);
-  }
-
-  // Check if it's a Gaxios error with response data
-  if (error && typeof error === 'object' && 'response' in error) {
-    const gaxiosError = error as {
-      response?: {
-        data?: unknown;
-      };
-    };
-    if (gaxiosError.response && gaxiosError.response.data) {
-      if (typeof gaxiosError.response.data === 'string') {
-        return checkMessage(gaxiosError.response.data);
-      }
-      if (
-        typeof gaxiosError.response.data === 'object' &&
-        gaxiosError.response.data !== null &&
-        'error' in gaxiosError.response.data
-      ) {
-        const errorData = gaxiosError.response.data as {
-          error?: { message?: string };
-        };
-        return checkMessage(errorData.error?.message || '');
-      }
-    }
-  }
-  return false;
-}
-
-export function isGenericQuotaExceededError(error: unknown): boolean {
-  if (typeof error === 'string') {
-    return error.includes('Quota exceeded for quota metric');
-  }
-
-  if (isStructuredError(error)) {
-    return error.message.includes('Quota exceeded for quota metric');
-  }
-
-  if (isApiError(error)) {
-    return error.error.message.includes('Quota exceeded for quota metric');
-  }
-
-  return false;
-}
@@ -7,10 +7,15 @@
 /* eslint-disable @typescript-eslint/no-explicit-any */
 import { describe, it, expect, vi, beforeEach, afterEach } from 'vitest';
 import { ApiError } from '@google/genai';
+import { AuthType } from '../core/contentGenerator.js';
 import type { HttpError } from './retry.js';
 import { retryWithBackoff } from './retry.js';
 import { setSimulate429 } from './testUtils.js';
 import { debugLogger } from './debugLogger.js';
+import {
+  TerminalQuotaError,
+  RetryableQuotaError,
+} from './googleQuotaErrors.js';

 // Helper to create a mock function that fails a certain number of times
 const createFailingFunction = (
@@ -100,26 +105,26 @@ describe('retryWithBackoff', () => {

    // Expect it to fail with the error from the 5th attempt.
    await Promise.all([
-      expect(promise).rejects.toThrow('Simulated error attempt 5'),
+      expect(promise).rejects.toThrow('Simulated error attempt 3'),
      vi.runAllTimersAsync(),
    ]);

-    expect(mockFn).toHaveBeenCalledTimes(5);
+    expect(mockFn).toHaveBeenCalledTimes(3);
  });

-  it('should default to 5 maxAttempts if options.maxAttempts is undefined', async () => {
-    // This function will fail more than 5 times to ensure all retries are used.
+  it('should default to 3 maxAttempts if options.maxAttempts is undefined', async () => {
+    // This function will fail more than 3 times to ensure all retries are used.
    const mockFn = createFailingFunction(10);

    const promise = retryWithBackoff(mockFn, { maxAttempts: undefined });

    // Expect it to fail with the error from the 5th attempt.
    await Promise.all([
-      expect(promise).rejects.toThrow('Simulated error attempt 5'),
+      expect(promise).rejects.toThrow('Simulated error attempt 3'),
      vi.runAllTimersAsync(),
    ]);

-    expect(mockFn).toHaveBeenCalledTimes(5);
+    expect(mockFn).toHaveBeenCalledTimes(3);
  });

  it('should not retry if shouldRetry returns false', async () => {
@@ -336,15 +341,13 @@ describe('retryWithBackoff', () => {
  });

  describe('Flash model fallback for OAuth users', () => {
-    it('should trigger fallback for OAuth personal users after persistent 429 errors', async () => {
+    it('should trigger fallback for OAuth personal users on TerminalQuotaError', async () => {
      const fallbackCallback = vi.fn().mockResolvedValue('gemini-2.5-flash');

      let fallbackOccurred = false;
      const mockFn = vi.fn().mockImplementation(async () => {
        if (!fallbackOccurred) {
-          const error: HttpError = new Error('Rate limit exceeded');
-          error.status = 429;
-          throw error;
+          throw new TerminalQuotaError('Daily limit reached', {} as any);
        }
        return 'success';
      });
@@ -352,154 +355,62 @@ describe('retryWithBackoff', () => {
      const promise = retryWithBackoff(mockFn, {
        maxAttempts: 3,
        initialDelayMs: 100,
-        onPersistent429: async (authType?: string) => {
+        onPersistent429: async (authType?: string, error?: unknown) => {
          fallbackOccurred = true;
-          return await fallbackCallback(authType);
+          return await fallbackCallback(authType, error);
        },
        authType: 'oauth-personal',
      });

-      // Advance all timers to complete retries
-      await vi.runAllTimersAsync();
-
-      // Should succeed after fallback
-      await expect(promise).resolves.toBe('success');
-
-      // Verify callback was called with correct auth type
-      expect(fallbackCallback).toHaveBeenCalledWith('oauth-personal');
-
-      // Should retry again after fallback
-      expect(mockFn).toHaveBeenCalledTimes(3); // 2 initial attempts + 1 after fallback
-    });
-
-    it('should NOT trigger fallback for API key users', async () => {
-      const fallbackCallback = vi.fn();
-
-      const mockFn = vi.fn(async () => {
-        const error: HttpError = new Error('Rate limit exceeded');
-        error.status = 429;
-        throw error;
-      });
-
-      const promise = retryWithBackoff(mockFn, {
-        maxAttempts: 3,
-        initialDelayMs: 100,
-        onPersistent429: fallbackCallback,
-        authType: 'gemini-api-key',
-      });
-
-      // Handle the promise properly to avoid unhandled rejections
-      const resultPromise = promise.catch((error) => error);
-      await vi.runAllTimersAsync();
-      const result = await resultPromise;
-
-      // Should fail after all retries without fallback
-      expect(result).toBeInstanceOf(Error);
-      expect(result.message).toBe('Rate limit exceeded');
-
-      // Callback should not be called for API key users
-      expect(fallbackCallback).not.toHaveBeenCalled();
-    });
-
-    it('should reset attempt counter and continue after successful fallback', async () => {
-      let fallbackCalled = false;
-      const fallbackCallback = vi.fn().mockImplementation(async () => {
-        fallbackCalled = true;
-        return 'gemini-2.5-flash';
-      });
-
-      const mockFn = vi.fn().mockImplementation(async () => {
-        if (!fallbackCalled) {
-          const error: HttpError = new Error('Rate limit exceeded');
-          error.status = 429;
-          throw error;
-        }
-        return 'success';
-      });
-
-      const promise = retryWithBackoff(mockFn, {
-        maxAttempts: 3,
-        initialDelayMs: 100,
-        onPersistent429: fallbackCallback,
-        authType: 'oauth-personal',
-      });
-
      await vi.runAllTimersAsync();

      await expect(promise).resolves.toBe('success');
-      expect(fallbackCallback).toHaveBeenCalledOnce();
-    });
-
-    it('should continue with original error if fallback is rejected', async () => {
-      const fallbackCallback = vi.fn().mockResolvedValue(null); // User rejected fallback
-
-      const mockFn = vi.fn(async () => {
-        const error: HttpError = new Error('Rate limit exceeded');
-        error.status = 429;
-        throw error;
-      });
-
-      const promise = retryWithBackoff(mockFn, {
-        maxAttempts: 3,
-        initialDelayMs: 100,
-        onPersistent429: fallbackCallback,
-        authType: 'oauth-personal',
-      });
-
-      // Handle the promise properly to avoid unhandled rejections
-      const resultPromise = promise.catch((error) => error);
-      await vi.runAllTimersAsync();
-      const result = await resultPromise;
-
-      // Should fail with original error when fallback is rejected
-      expect(result).toBeInstanceOf(Error);
-      expect(result.message).toBe('Rate limit exceeded');
      expect(fallbackCallback).toHaveBeenCalledWith(
        'oauth-personal',
-        expect.any(Error),
+        expect.any(TerminalQuotaError),
      );
+      expect(mockFn).toHaveBeenCalledTimes(2);
    });

-    it('should handle mixed error types (only count consecutive 429s)', async () => {
-      const fallbackCallback = vi.fn().mockResolvedValue('gemini-2.5-flash');
-      let attempts = 0;
-      let fallbackOccurred = false;
-
+    it('should use retryDelayMs from RetryableQuotaError', async () => {
+      const setTimeoutSpy = vi.spyOn(global, 'setTimeout');
      const mockFn = vi.fn().mockImplementation(async () => {
-        attempts++;
-        if (fallbackOccurred) {
-          return 'success';
-        }
-        if (attempts === 1) {
-          // First attempt: 500 error (resets consecutive count)
-          const error: HttpError = new Error('Server error');
-          error.status = 500;
-          throw error;
-        } else {
-          // Remaining attempts: 429 errors
-          const error: HttpError = new Error('Rate limit exceeded');
-          error.status = 429;
-          throw error;
-        }
+        throw new RetryableQuotaError('Per-minute limit', {} as any, 12.345);
      });

      const promise = retryWithBackoff(mockFn, {
-        maxAttempts: 5,
+        maxAttempts: 2,
        initialDelayMs: 100,
-        onPersistent429: async (authType?: string) => {
-          fallbackOccurred = true;
-          return await fallbackCallback(authType);
-        },
-        authType: 'oauth-personal',
      });

+      // Attach the rejection expectation *before* running timers
+      // eslint-disable-next-line vitest/valid-expect
+      const assertionPromise = expect(promise).rejects.toThrow();
      await vi.runAllTimersAsync();
+      await assertionPromise;

-      await expect(promise).resolves.toBe('success');
-
-      // Should trigger fallback after 2 consecutive 429s (attempts 2-3)
-      expect(fallbackCallback).toHaveBeenCalledWith('oauth-personal');
+      expect(setTimeoutSpy).toHaveBeenCalledWith(expect.any(Function), 12345);
    });
+
+    it.each([[AuthType.USE_GEMINI], [AuthType.USE_VERTEX_AI], [undefined]])(
+      'should not trigger fallback for non-Google auth users (authType: %s) on TerminalQuotaError',
+      async (authType) => {
+        const fallbackCallback = vi.fn();
+        const mockFn = vi.fn().mockImplementation(async () => {
+          throw new TerminalQuotaError('Daily limit reached', {} as any);
+        });
+
+        const promise = retryWithBackoff(mockFn, {
+          maxAttempts: 3,
+          onPersistent429: fallbackCallback,
+          authType,
+        });
+
+        await expect(promise).rejects.toThrow('Daily limit reached');
+        expect(fallbackCallback).not.toHaveBeenCalled();
+        expect(mockFn).toHaveBeenCalledTimes(1);
+      },
+    );
  });
  it('should abort the retry loop when the signal is aborted', async () => {
    const abortController = new AbortController();
@@ -8,9 +8,10 @@ import type { GenerateContentResponse } from '@google/genai';
 import { ApiError } from '@google/genai';
 import { AuthType } from '../core/contentGenerator.js';
 import {
-  isProQuotaExceededError,
-  isGenericQuotaExceededError,
-} from './quotaErrorDetection.js';
+  classifyGoogleError,
+  RetryableQuotaError,
+  TerminalQuotaError,
+} from './googleQuotaErrors.js';
 import { delay, createAbortError } from './delay.js';
 import { debugLogger } from './debugLogger.js';

@@ -37,7 +38,7 @@ export interface RetryOptions {
 }

 const DEFAULT_RETRY_OPTIONS: RetryOptions = {
-  maxAttempts: 5,
+  maxAttempts: 3,
  initialDelayMs: 5000,
  maxDelayMs: 30000, // 30 seconds
  shouldRetryOnError: defaultShouldRetry,
@@ -118,7 +119,6 @@ export async function retryWithBackoff<T>(

  let attempt = 0;
  let currentDelay = initialDelayMs;
-  let consecutive429Count = 0;

  while (attempt < maxAttempts) {
    if (signal?.aborted) {
@@ -145,94 +145,54 @@ export async function retryWithBackoff<T>(
        throw error;
      }

-      const errorStatus = getErrorStatus(error);
+      const classifiedError = classifyGoogleError(error);

-      // Check for Pro quota exceeded error first - immediate fallback for OAuth users
-      if (
-        errorStatus === 429 &&
-        authType === AuthType.LOGIN_WITH_GOOGLE &&
-        isProQuotaExceededError(error) &&
-        onPersistent429
-      ) {
-        try {
-          const fallbackModel = await onPersistent429(authType, error);
-          if (fallbackModel !== false && fallbackModel !== null) {
-            // Reset attempt counter and try with new model
-            attempt = 0;
-            consecutive429Count = 0;
-            currentDelay = initialDelayMs;
-            // With the model updated, we continue to the next attempt
-            continue;
-          } else {
-            // Fallback handler returned null/false, meaning don't continue - stop retry process
-            throw error;
+      if (classifiedError instanceof TerminalQuotaError) {
+        if (onPersistent429 && authType === AuthType.LOGIN_WITH_GOOGLE) {
+          try {
+            const fallbackModel = await onPersistent429(
+              authType,
+              classifiedError,
+            );
+            if (fallbackModel) {
+              attempt = 0; // Reset attempts and retry with the new model.
+              currentDelay = initialDelayMs;
+              continue;
+            }
+          } catch (fallbackError) {
+            debugLogger.warn('Fallback to Flash model failed:', fallbackError);
          }
-        } catch (fallbackError) {
-          // If fallback fails, continue with original error
-          debugLogger.warn('Fallback to Flash model failed:', fallbackError);
        }
+        throw classifiedError; // Throw if no fallback or fallback failed.
      }

-      // Check for generic quota exceeded error (but not Pro, which was handled above) - immediate fallback for OAuth users
-      if (
-        errorStatus === 429 &&
-        authType === AuthType.LOGIN_WITH_GOOGLE &&
-        !isProQuotaExceededError(error) &&
-        isGenericQuotaExceededError(error) &&
-        onPersistent429
-      ) {
-        try {
-          const fallbackModel = await onPersistent429(authType, error);
-          if (fallbackModel !== false && fallbackModel !== null) {
-            // Reset attempt counter and try with new model
-            attempt = 0;
-            consecutive429Count = 0;
-            currentDelay = initialDelayMs;
-            // With the model updated, we continue to the next attempt
-            continue;
-          } else {
-            // Fallback handler returned null/false, meaning don't continue - stop retry process
-            throw error;
+      if (classifiedError instanceof RetryableQuotaError) {
+        if (attempt >= maxAttempts) {
+          if (onPersistent429 && authType === AuthType.LOGIN_WITH_GOOGLE) {
+            try {
+              const fallbackModel = await onPersistent429(
+                authType,
+                classifiedError,
+              );
+              if (fallbackModel) {
+                attempt = 0; // Reset attempts and retry with the new model.
+                currentDelay = initialDelayMs;
+                continue;
+              }
+            } catch (fallbackError) {
+              console.warn('Model fallback failed:', fallbackError);
+            }
          }
-        } catch (fallbackError) {
-          // If fallback fails, continue with original error
-          debugLogger.warn('Fallback to Flash model failed:', fallbackError);
+          throw classifiedError;
        }
+        console.warn(
+          `Attempt ${attempt} failed: ${classifiedError.message}. Retrying after ${classifiedError.retryDelayMs}ms...`,
+        );
+        await delay(classifiedError.retryDelayMs, signal);
+        continue;
      }

-      // Track consecutive 429 errors
-      if (errorStatus === 429) {
-        consecutive429Count++;
-      } else {
-        consecutive429Count = 0;
-      }
-
-      // If we have persistent 429s and a fallback callback for OAuth
-      if (
-        consecutive429Count >= 2 &&
-        onPersistent429 &&
-        authType === AuthType.LOGIN_WITH_GOOGLE
-      ) {
-        try {
-          const fallbackModel = await onPersistent429(authType, error);
-          if (fallbackModel !== false && fallbackModel !== null) {
-            // Reset attempt counter and try with new model
-            attempt = 0;
-            consecutive429Count = 0;
-            currentDelay = initialDelayMs;
-            // With the model updated, we continue to the next attempt
-            continue;
-          } else {
-            // Fallback handler returned null/false, meaning don't continue - stop retry process
-            throw error;
-          }
-        } catch (fallbackError) {
-          // If fallback fails, continue with original error
-          debugLogger.warn('Fallback to Flash model failed:', fallbackError);
-        }
-      }
-
-      // Check if we've exhausted retries or shouldn't retry
+      // Generic retry logic for other errors
      if (
        attempt >= maxAttempts ||
        !shouldRetryOnError(error as Error, retryFetchErrors)
@@ -240,31 +200,17 @@ export async function retryWithBackoff<T>(
        throw error;
      }

-      const { delayDurationMs, errorStatus: delayErrorStatus } =
-        getDelayDurationAndStatus(error);
+      const errorStatus = getErrorStatus(error);
+      logRetryAttempt(attempt, error, errorStatus);

-      if (delayDurationMs > 0) {
-        // Respect Retry-After header if present and parsed
-        debugLogger.warn(
-          `Attempt ${attempt} failed with status ${delayErrorStatus ?? 'unknown'}. Retrying after explicit delay of ${delayDurationMs}ms...`,
-          error,
-        );
-        await delay(delayDurationMs, signal);
-        // Reset currentDelay for next potential non-429 error, or if Retry-After is not present next time
-        currentDelay = initialDelayMs;
-      } else {
-        // Fall back to exponential backoff with jitter
-        logRetryAttempt(attempt, error, errorStatus);
-        // Add jitter: +/- 30% of currentDelay
-        const jitter = currentDelay * 0.3 * (Math.random() * 2 - 1);
-        const delayWithJitter = Math.max(0, currentDelay + jitter);
-        await delay(delayWithJitter, signal);
-        currentDelay = Math.min(maxDelayMs, currentDelay * 2);
-      }
+      // Exponential backoff with jitter for non-quota errors
+      const jitter = currentDelay * 0.3 * (Math.random() * 2 - 1);
+      const delayWithJitter = Math.max(0, currentDelay + jitter);
+      await delay(delayWithJitter, signal);
+      currentDelay = Math.min(maxDelayMs, currentDelay * 2);
    }
  }
-  // This line should theoretically be unreachable due to the throw in the catch block.
-  // Added for type safety and to satisfy the compiler that a promise is always returned.
+
  throw new Error('Retry attempts exhausted');
 }

@@ -295,62 +241,6 @@ export function getErrorStatus(error: unknown): number | undefined {
  return undefined;
 }

-/**
- * Extracts the Retry-After delay from an error object's headers.
- * @param error The error object.
- * @returns The delay in milliseconds, or 0 if not found or invalid.
- */
-function getRetryAfterDelayMs(error: unknown): number {
-  if (typeof error === 'object' && error !== null) {
-    // Check for error.response.headers (common in axios errors)
-    if (
-      'response' in error &&
-      typeof (error as { response?: unknown }).response === 'object' &&
-      (error as { response?: unknown }).response !== null
-    ) {
-      const response = (error as { response: { headers?: unknown } }).response;
-      if (
-        'headers' in response &&
-        typeof response.headers === 'object' &&
-        response.headers !== null
-      ) {
-        const headers = response.headers as { 'retry-after'?: unknown };
-        const retryAfterHeader = headers['retry-after'];
-        if (typeof retryAfterHeader === 'string') {
-          const retryAfterSeconds = parseInt(retryAfterHeader, 10);
-          if (!isNaN(retryAfterSeconds)) {
-            return retryAfterSeconds * 1000;
-          }
-          // It might be an HTTP date
-          const retryAfterDate = new Date(retryAfterHeader);
-          if (!isNaN(retryAfterDate.getTime())) {
-            return Math.max(0, retryAfterDate.getTime() - Date.now());
-          }
-        }
-      }
-    }
-  }
-  return 0;
-}
-
-/**
- * Determines the delay duration based on the error, prioritizing Retry-After header.
- * @param error The error object.
- * @returns An object containing the delay duration in milliseconds and the error status.
- */
-function getDelayDurationAndStatus(error: unknown): {
-  delayDurationMs: number;
-  errorStatus: number | undefined;
-} {
-  const errorStatus = getErrorStatus(error);
-  let delayDurationMs = 0;
-
-  if (errorStatus === 429) {
-    delayDurationMs = getRetryAfterDelayMs(error);
-  }
-  return { delayDurationMs, errorStatus };
-}
-
 /**
 * Logs a message for a retry attempt when using exponential backoff.
 * @param attempt The current attempt number.
@@ -1,6 +1,6 @@
 {
  "name": "@google/gemini-cli-test-utils",
-  "version": "0.11.0-nightly.20251021.e72c00cf",
+  "version": "0.11.3",
  "private": true,
  "main": "src/index.ts",
  "license": "Apache-2.0",
@@ -2,7 +2,7 @@
  "name": "gemini-cli-vscode-ide-companion",
  "displayName": "Gemini CLI Companion",
  "description": "Enable Gemini CLI with direct access to your IDE workspace.",
-  "version": "0.11.0-nightly.20251021.e72c00cf",
+  "version": "0.11.3",
  "publisher": "google",
  "icon": "assets/icon.png",
  "repository": {
Author	SHA1	Message	Date
gemini-cli-robot	e5161610ff	chore(release): v0.11.3	2025-10-31 20:24:24 +00:00
Gaurav Ghosh	44b3c97486	refactor: simplify daily quota error messages	2025-10-31 11:11:19 -07:00
Sandy Tao	b1bbef433d	fix(core): ensure loop detection respects session disable flag (#12347 )	2025-10-31 10:31:04 -07:00
Tommaso Sciortino	8a72585992	Mark `model.compressionThreshold` as requiring a restart (#12378 )	2025-10-31 10:31:03 -07:00
Abhi	73b3211ea7	Remove context percentage in footer by default (#12326 )	2025-10-31 10:31:02 -07:00
Tommaso Sciortino	ce40a6534f	Make compression threshold editable in the UI. (#12317 )	2025-10-31 10:30:34 -07:00
Tommaso Sciortino	5be5575df0	Change default compression threshold (#12306 )	2025-10-31 09:32:34 -07:00
Sandy Tao	847c6e7fd0	refactor(core): extract ChatCompressionService from GeminiClient (#12001 )	2025-10-31 09:21:04 -07:00
gemini-cli-robot	f4f3727936	chore(release): v0.11.2	2025-10-31 02:33:21 +00:00
gemini-cli-robot	f36dec6ad4	fix(patch): cherry-pick `643f2c0` to release/v0.11.1-pr-12300 to patch version v0.11.1 and create version 0.11.2 (#12335 ) Co-authored-by: Victor May <mayvic@google.com>	2025-10-30 19:20:45 -07:00
gemini-cli-robot	5213d9f3f9	chore(release): v0.11.1	2025-10-30 23:11:54 +00:00
gemini-cli-robot	d9f6cebe1f	fix(patch): cherry-pick `ee92db7` to release/v0.11.0-pr-11624 to patch version v0.11.0 and create version 0.11.1 (#12321 ) Co-authored-by: Gaurav <39389231+gsquared94@users.noreply.github.com>	2025-10-30 15:55:48 -07:00
gemini-cli-robot	92f5355d81	chore(release): v0.11.0	2025-10-29 01:17:14 +00:00
gemini-cli-robot	c9c2e79dde	chore(release): v0.11.0-preview.1	2025-10-28 22:56:55 +00:00
gemini-cli-robot	a3947a8de5	fix(patch): cherry-pick `601a639` to release/v0.11.0-preview.0-pr-11889 to patch version v0.11.0-preview.0 and create version 0.11.0-preview.1 (#12188 ) Co-authored-by: Victor May <mayvic@google.com> Co-authored-by: matt korwel <matt.korwel@gmail.com>	2025-10-28 15:44:17 -07:00
gemini-cli-robot	9cf8b40349	chore(release): v0.11.0-preview.0	2025-10-22 17:11:36 +00:00