Compare commits

...

23 Commits

Author SHA1 Message Date
gemini-cli-robot 071076aa5c chore(release): v0.36.0-preview.7 2026-03-31 16:44:52 +00:00
Tommaso Sciortino 425d64a96f fix(chore): resolve typescript errors resulting from cherry-picks 2026-03-31 09:11:46 -07:00
Tommaso Sciortino c63350cfd6 fix broken tests (#24279) 2026-03-31 08:41:00 -07:00
Tommaso Sciortino 55a5ed6545 fix(cli): resolve missing F12 logs via global console store (#24235) 2026-03-31 08:40:58 -07:00
Sehoon Shon 7e0ac00efc feat(cli): show Flash Lite Preview model regardless of user tier (#23904) 2026-03-31 08:40:56 -07:00
christine betts bbd8483e34 feat(core,ui): Add experiment-gated support for gemini flash 3.1 lite (#23794) 2026-03-31 08:40:54 -07:00
Jacob Richman 29ab3bbfd7 fix(cli): allow ask question dialog to take full window height (#23693) 2026-03-31 08:39:08 -07:00
Abhi 0f086869d5 fix(core): ensure default agents provide tools and use model-specific schemas (#24268) 2026-03-31 08:37:49 -07:00
gemini-cli-robot c711a38f16 chore(release): v0.36.0-preview.6 2026-03-28 03:13:25 +00:00
gemini-cli-robot 975b7dc163 fix(patch): cherry-pick 765fb67 to release/v0.36.0-preview.5-pr-24055 to patch version v0.36.0-preview.5 and create version 0.36.0-preview.6 (#24061)
Co-authored-by: Gal Zahavi <38544478+galz10@users.noreply.github.com>
2026-03-27 18:27:29 -07:00
gemini-cli-robot c1af5ab99d chore(release): v0.36.0-preview.5 2026-03-27 17:36:48 +00:00
Keith Schaab 310db3b823 fix(a2a-server): A2A server should execute ask policies in interactive mode (#23831) 2026-03-27 10:08:19 -07:00
Adam Weidman 6222f150fa docs(core): document agent_card_json string literal options for remote agents (#23797) 2026-03-26 14:38:29 -07:00
gemini-cli-robot 514bf61903 chore(release): v0.36.0-preview.4 2026-03-26 19:49:49 +00:00
Adam Weidman 38c706dbb9 Merge remote-tracking branch 'origin/main' into fix/agent-loader-error-formatting
# Conflicts:
#	packages/core/src/agents/agentLoader.ts

# Conflicts:
#	packages/core/src/agents/agentLoader.ts
2026-03-26 14:04:36 -04:00
Adam Weidman 5e18b14c10 feat(core): support inline agentCardJson for remote agents
- Add agent_card_json field as alternative to agent_card_url in remote
  agent markdown frontmatter with Zod schema enforcing mutual exclusivity
- Refactor loadAgent to accept AgentCardLoadOptions discriminated union
- Add AgentCardLoadOptions, getAgentCardLoadOptions, getRemoteAgentTargetUrl
  helpers in types.ts to centralize remote agent card resolution
- Fix google-credentials auth crash when using agentCardJson by extracting
  targetUrl from inline JSON card
- Hash agentCardJson with SHA-256 for metadata.hash instead of raw string
- Add JSON syntax validation via Zod .refine() on agent_card_json field
- Refactor formatZodError with recursive flatMap and deduplication
- Update remote agents documentation for oauth type naming
- Add comprehensive test coverage for all agentCardJson code paths

# Conflicts:
#	packages/core/src/agents/agentLoader.ts
2026-03-26 14:03:52 -04:00
gemini-cli-robot 8e2629759d chore(release): v0.36.0-preview.3 2026-03-25 20:53:49 +00:00
Tommaso Sciortino 146442f2a2 Revert "perf(cli): optimize --version startup time (#23671)"
This reverts commit 139cc7b97c.
2026-03-25 13:22:22 -07:00
Adib234 a3a3e66922 fix(policy): relax write_file argsPattern in plan mode to allow paths without session ID (#23695) 2026-03-25 14:06:21 -04:00
Gal Zahavi d4555a473e fix(core): conditionally expose additional_permissions in shell tool (#23729)
Co-authored-by: Sandy Tao <sandytao520@icloud.com>
2026-03-25 10:36:46 -07:00
matt korwel cdac58fe82 fix(test): move flaky ctrl-c-exit test to non-blocking suite (#23732) 2026-03-25 10:35:57 -07:00
gemini-cli-robot 4a95ab62d4 fix(patch): cherry-pick 055ff92 to release/v0.36.0-preview.0-pr-23672 to patch version v0.36.0-preview.0 and create version 0.36.0-preview.1 (#23723)
Co-authored-by: Abhi <43648792+abhipatel12@users.noreply.github.com>
2026-03-24 20:07:02 -04:00
gemini-cli-robot 0223181cf4 chore(release): v0.36.0-preview.0 2026-03-24 19:57:36 +00:00
92 changed files with 2202 additions and 887 deletions
+96 -6
View File
@@ -51,12 +51,13 @@ You can place them in:
### Configuration schema
| Field | Type | Required | Description |
| :--------------- | :----- | :------- | :------------------------------------------------------------------------------------------------------------- |
| `kind` | string | Yes | Must be `remote`. |
| `name` | string | Yes | A unique name for the agent. Must be a valid slug (lowercase letters, numbers, hyphens, and underscores only). |
| `agent_card_url` | string | Yes | The URL to the agent's A2A card endpoint. |
| `auth` | object | No | Authentication configuration. See [Authentication](#authentication). |
| Field | Type | Required | Description |
| :---------------- | :----- | :------- | :------------------------------------------------------------------------------------------------------------- |
| `kind` | string | Yes | Must be `remote`. |
| `name` | string | Yes | A unique name for the agent. Must be a valid slug (lowercase letters, numbers, hyphens, and underscores only). |
| `agent_card_url` | string | Yes\* | The URL to the agent's A2A card endpoint. Required if `agent_card_json` is not provided. |
| `agent_card_json` | string | Yes\* | The inline JSON string of the agent's A2A card. Required if `agent_card_url` is not provided. |
| `auth` | object | No | Authentication configuration. See [Authentication](#authentication). |
### Single-subagent example
@@ -88,6 +89,95 @@ Markdown file.
> [!NOTE] Mixed local and remote agents, or multiple local agents, are not
> supported in a single file; the list format is currently remote-only.
### Inline Agent Card JSON
<details>
<summary>View formatting options for JSON strings</summary>
If you don't have an endpoint serving the agent card, you can provide the A2A
card directly as a JSON string using `agent_card_json`.
When providing a JSON string in YAML, you must properly format it as a string
scalar. You can use single quotes, a block scalar, or double quotes (which
require escaping internal double quotes).
#### Using single quotes
Single quotes allow you to embed unescaped double quotes inside the JSON string.
This format is useful for shorter, single-line JSON strings.
```markdown
---
kind: remote
name: single-quotes-agent
agent_card_json:
'{ "protocolVersion": "0.3.0", "name": "Example Agent", "version": "1.0.0",
"url": "dummy-url" }'
---
```
#### Using a block scalar
The literal block scalar (`|`) preserves line breaks and is highly recommended
for multiline JSON strings as it avoids quote escaping entirely. The following
is a complete, valid Agent Card configuration using dummy values.
```markdown
---
kind: remote
name: block-scalar-agent
agent_card_json: |
{
"protocolVersion": "0.3.0",
"name": "Example Agent Name",
"description": "An example agent description for documentation purposes.",
"version": "1.0.0",
"url": "dummy-url",
"preferredTransport": "HTTP+JSON",
"capabilities": {
"streaming": true,
"extendedAgentCard": false
},
"defaultInputModes": [
"text/plain"
],
"defaultOutputModes": [
"application/json"
],
"skills": [
{
"id": "ExampleSkill",
"name": "Example Skill Assistant",
"description": "A description of what this example skill does.",
"tags": [
"example-tag"
],
"examples": [
"Show me an example."
]
}
]
}
---
```
#### Using double quotes
Double quotes are also supported, but any internal double quotes in your JSON
must be escaped with a backslash.
```markdown
---
kind: remote
name: double-quotes-agent
agent_card_json:
'{ "protocolVersion": "0.3.0", "name": "Example Agent", "version": "1.0.0",
"url": "dummy-url" }'
---
```
</details>
## Authentication
Many remote agents require authentication. Gemini CLI supports several
+26 -2
View File
@@ -645,6 +645,11 @@ their corresponding top-level category object in your `settings.json` file.
"model": "gemini-3-flash-preview"
}
},
"chat-compression-3.1-flash-lite": {
"modelConfig": {
"model": "gemini-3.1-flash-lite-preview"
}
},
"chat-compression-2.5-pro": {
"modelConfig": {
"model": "gemini-2.5-pro"
@@ -973,6 +978,17 @@ their corresponding top-level category object in your `settings.json` file.
"auto-gemini-2.5": {
"default": "gemini-2.5-pro"
},
"gemini-3.1-flash-lite-preview": {
"default": "gemini-3.1-flash-lite-preview",
"contexts": [
{
"condition": {
"useGemini3_1FlashLite": false
},
"target": "gemini-2.5-flash-lite"
}
]
},
"flash": {
"default": "gemini-3-flash-preview",
"contexts": [
@@ -985,7 +1001,15 @@ their corresponding top-level category object in your `settings.json` file.
]
},
"flash-lite": {
"default": "gemini-2.5-flash-lite"
"default": "gemini-2.5-flash-lite",
"contexts": [
{
"condition": {
"useGemini3_1FlashLite": true
},
"target": "gemini-3.1-flash-lite-preview"
}
]
}
}
```
@@ -1540,7 +1564,7 @@ their corresponding top-level category object in your `settings.json` file.
- **`experimental.enableAgents`** (boolean):
- **Description:** Enable local and remote subagents.
- **Default:** `false`
- **Default:** `true`
- **Requires restart:** Yes
- **`experimental.worktrees`** (boolean):
-6
View File
@@ -63,9 +63,6 @@ describe.skipIf(!chromeAvailable)('browser-policy', () => {
rig.setup('browser-policy-skip-confirmation', {
fakeResponsesPath: join(__dirname, 'browser-policy.responses'),
settings: {
experimental: {
enableAgents: true,
},
agents: {
overrides: {
browser_agent: {
@@ -183,9 +180,6 @@ priority = 200
rig.setup('browser-session-warning', {
fakeResponsesPath: join(__dirname, 'browser-agent.cleanup.responses'),
settings: {
experimental: {
enableAgents: true,
},
general: {
enableAutoUpdateNotification: false,
},
+2 -2
View File
@@ -6,9 +6,9 @@
import { describe, it, expect, beforeEach, afterEach } from 'vitest';
import * as os from 'node:os';
import { TestRig } from './test-helper.js';
import { TestRig, skipFlaky } from './test-helper.js';
describe('Ctrl+C exit', () => {
describe.skipIf(skipFlaky)('Ctrl+C exit', () => {
let rig: TestRig;
beforeEach(() => {
+92 -58
View File
@@ -4,10 +4,8 @@
* SPDX-License-Identifier: Apache-2.0
*/
import { writeFileSync } from 'node:fs';
import { join } from 'node:path';
import { describe, it, expect, beforeEach, afterEach } from 'vitest';
import { TestRig, checkModelOutputContent, GEMINI_DIR } from './test-helper.js';
import { TestRig, checkModelOutputContent } from './test-helper.js';
describe('Plan Mode', () => {
let rig: TestRig;
@@ -36,27 +34,23 @@ describe('Plan Mode', () => {
},
);
// We use a prompt that asks for both a read-only action and a write action.
// "List files" (read-only) followed by "touch denied.txt" (write).
const result = await rig.run({
approvalMode: 'plan',
stdin:
'Please list the files in the current directory, and then attempt to create a new file named "denied.txt" using a shell command.',
args: 'Please list the files in the current directory, and then attempt to create a new file named "denied.txt" using a shell command.',
});
const lsCallFound = await rig.waitForToolCall('list_directory');
expect(lsCallFound, 'Expected list_directory to be called').toBe(true);
const shellCallFound = await rig.waitForToolCall('run_shell_command');
expect(shellCallFound, 'Expected run_shell_command to fail').toBe(false);
const toolLogs = rig.readToolLogs();
const lsLog = toolLogs.find((l) => l.toolRequest.name === 'list_directory');
expect(
toolLogs.find((l) => l.toolRequest.name === 'run_shell_command'),
).toBeUndefined();
const shellLog = toolLogs.find(
(l) => l.toolRequest.name === 'run_shell_command',
);
expect(lsLog, 'Expected list_directory to be called').toBeDefined();
expect(lsLog?.toolRequest.success).toBe(true);
expect(
shellLog,
'Expected run_shell_command to be blocked (not even called)',
).toBeUndefined();
checkModelOutputContent(result, {
expectedContent: ['Plan Mode', 'read-only'],
@@ -84,23 +78,11 @@ describe('Plan Mode', () => {
},
});
// Disable the interactive terminal setup prompt in tests
writeFileSync(
join(rig.homeDir!, GEMINI_DIR, 'state.json'),
JSON.stringify({ terminalSetupPromptShown: true }, null, 2),
);
const run = await rig.runInteractive({
await rig.run({
approvalMode: 'plan',
args: 'Create a file called plan.md in the plans directory.',
});
await run.type('Create a file called plan.md in the plans directory.');
await run.type('\r');
await rig.expectToolCallSuccess(['write_file'], 30000, (args) =>
args.includes('plan.md'),
);
const toolLogs = rig.readToolLogs();
const planWrite = toolLogs.find(
(l) =>
@@ -108,7 +90,25 @@ describe('Plan Mode', () => {
l.toolRequest.args.includes('plans') &&
l.toolRequest.args.includes('plan.md'),
);
expect(planWrite?.toolRequest.success).toBe(true);
if (!planWrite) {
console.error(
'All tool calls found:',
toolLogs.map((l) => ({
name: l.toolRequest.name,
args: l.toolRequest.args,
})),
);
}
expect(
planWrite,
'Expected write_file to be called for plan.md',
).toBeDefined();
expect(
planWrite?.toolRequest.success,
`Expected write_file to succeed, but it failed with error: ${planWrite?.toolRequest.error}`,
).toBe(true);
});
it('should deny write_file to non-plans directory in plan mode', async () => {
@@ -131,19 +131,11 @@ describe('Plan Mode', () => {
},
});
// Disable the interactive terminal setup prompt in tests
writeFileSync(
join(rig.homeDir!, GEMINI_DIR, 'state.json'),
JSON.stringify({ terminalSetupPromptShown: true }, null, 2),
);
const run = await rig.runInteractive({
await rig.run({
approvalMode: 'plan',
args: 'Create a file called hello.txt in the current directory.',
});
await run.type('Create a file called hello.txt in the current directory.');
await run.type('\r');
const toolLogs = rig.readToolLogs();
const writeLog = toolLogs.find(
(l) =>
@@ -151,10 +143,11 @@ describe('Plan Mode', () => {
l.toolRequest.args.includes('hello.txt'),
);
// In Plan Mode, writes outside the plans directory should be blocked.
// Model is undeterministic, sometimes it doesn't even try, but if it does, it must fail.
if (writeLog) {
expect(writeLog.toolRequest.success).toBe(false);
expect(
writeLog.toolRequest.success,
'Expected write_file to non-plans dir to fail',
).toBe(false);
}
});
@@ -169,28 +162,69 @@ describe('Plan Mode', () => {
},
});
// Disable the interactive terminal setup prompt in tests
writeFileSync(
join(rig.homeDir!, GEMINI_DIR, 'state.json'),
JSON.stringify({ terminalSetupPromptShown: true }, null, 2),
);
// Start in default mode and ask to enter plan mode.
await rig.run({
approvalMode: 'default',
stdin:
'I want to perform a complex refactoring. Please enter plan mode so we can design it first.',
args: 'I want to perform a complex refactoring. Please enter plan mode so we can design it first.',
});
const enterPlanCallFound = await rig.waitForToolCall('enter_plan_mode');
expect(enterPlanCallFound, 'Expected enter_plan_mode to be called').toBe(
true,
);
const toolLogs = rig.readToolLogs();
const enterLog = toolLogs.find(
(l) => l.toolRequest.name === 'enter_plan_mode',
);
expect(enterLog, 'Expected enter_plan_mode to be called').toBeDefined();
expect(enterLog?.toolRequest.success).toBe(true);
});
it('should allow write_file to the plans directory in plan mode even without a session ID', async () => {
const plansDir = '.gemini/tmp/foo/plans';
const testName =
'should allow write_file to the plans directory in plan mode even without a session ID';
await rig.setup(testName, {
settings: {
experimental: { plan: true },
tools: {
core: ['write_file', 'read_file', 'list_directory'],
},
general: {
defaultApprovalMode: 'plan',
plan: {
directory: plansDir,
},
},
},
});
await rig.run({
approvalMode: 'plan',
args: 'Create a file called plan-no-session.md in the plans directory.',
});
const toolLogs = rig.readToolLogs();
const planWrite = toolLogs.find(
(l) =>
l.toolRequest.name === 'write_file' &&
l.toolRequest.args.includes('plans') &&
l.toolRequest.args.includes('plan-no-session.md'),
);
if (!planWrite) {
console.error(
'All tool calls found:',
toolLogs.map((l) => ({
name: l.toolRequest.name,
args: l.toolRequest.args,
})),
);
}
expect(
planWrite,
'Expected write_file to be called for plan-no-session.md',
).toBeDefined();
expect(
planWrite?.toolRequest.success,
`Expected write_file to succeed, but it failed with error: ${planWrite?.toolRequest.error}`,
).toBe(true);
});
});
+9 -9
View File
@@ -1,12 +1,12 @@
{
"name": "@google/gemini-cli",
"version": "0.36.0-nightly.20260317.2f90b4653",
"version": "0.36.0-preview.7",
"lockfileVersion": 3,
"requires": true,
"packages": {
"": {
"name": "@google/gemini-cli",
"version": "0.36.0-nightly.20260317.2f90b4653",
"version": "0.36.0-preview.7",
"workspaces": [
"packages/*"
],
@@ -17413,7 +17413,7 @@
},
"packages/a2a-server": {
"name": "@google/gemini-cli-a2a-server",
"version": "0.36.0-nightly.20260317.2f90b4653",
"version": "0.36.0-preview.7",
"dependencies": {
"@a2a-js/sdk": "0.3.11",
"@google-cloud/storage": "^7.16.0",
@@ -17528,7 +17528,7 @@
},
"packages/cli": {
"name": "@google/gemini-cli",
"version": "0.36.0-nightly.20260317.2f90b4653",
"version": "0.36.0-preview.7",
"license": "Apache-2.0",
"dependencies": {
"@agentclientprotocol/sdk": "^0.16.1",
@@ -17700,7 +17700,7 @@
},
"packages/core": {
"name": "@google/gemini-cli-core",
"version": "0.36.0-nightly.20260317.2f90b4653",
"version": "0.36.0-preview.7",
"license": "Apache-2.0",
"dependencies": {
"@a2a-js/sdk": "0.3.11",
@@ -17966,7 +17966,7 @@
},
"packages/devtools": {
"name": "@google/gemini-cli-devtools",
"version": "0.36.0-nightly.20260317.2f90b4653",
"version": "0.36.0-preview.7",
"license": "Apache-2.0",
"dependencies": {
"ws": "^8.16.0"
@@ -17981,7 +17981,7 @@
},
"packages/sdk": {
"name": "@google/gemini-cli-sdk",
"version": "0.36.0-nightly.20260317.2f90b4653",
"version": "0.36.0-preview.7",
"license": "Apache-2.0",
"dependencies": {
"@google/gemini-cli-core": "file:../core",
@@ -17998,7 +17998,7 @@
},
"packages/test-utils": {
"name": "@google/gemini-cli-test-utils",
"version": "0.36.0-nightly.20260317.2f90b4653",
"version": "0.36.0-preview.7",
"license": "Apache-2.0",
"dependencies": {
"@google/gemini-cli-core": "file:../core",
@@ -18015,7 +18015,7 @@
},
"packages/vscode-ide-companion": {
"name": "gemini-cli-vscode-ide-companion",
"version": "0.36.0-nightly.20260317.2f90b4653",
"version": "0.36.0-preview.7",
"license": "LICENSE",
"dependencies": {
"@modelcontextprotocol/sdk": "^1.23.0",
+2 -2
View File
@@ -1,6 +1,6 @@
{
"name": "@google/gemini-cli",
"version": "0.36.0-nightly.20260317.2f90b4653",
"version": "0.36.0-preview.7",
"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:0.36.0-nightly.20260317.2f90b4653"
"sandboxImageUri": "us-docker.pkg.dev/gemini-code-dev/gemini-cli/sandbox:0.36.0-preview.7"
},
"scripts": {
"start": "cross-env NODE_ENV=development node scripts/start.js",
+1 -1
View File
@@ -1,6 +1,6 @@
{
"name": "@google/gemini-cli-a2a-server",
"version": "0.36.0-nightly.20260317.2f90b4653",
"version": "0.36.0-preview.7",
"description": "Gemini CLI A2A Server",
"repository": {
"type": "git",
+21 -6
View File
@@ -29,6 +29,7 @@ vi.mock('@google/gemini-cli-core', async (importOriginal) => {
await importOriginal<typeof import('@google/gemini-cli-core')>();
return {
...actual,
PRIORITY_YOLO_ALLOW_ALL: 998,
Config: vi.fn().mockImplementation((params) => {
const mockConfig = {
...params,
@@ -341,33 +342,47 @@ describe('loadConfig', () => {
);
});
it('should default enableAgents to false when not provided', async () => {
it('should default enableAgents to true when not provided', async () => {
await loadConfig(mockSettings, mockExtensionLoader, taskId);
expect(Config).toHaveBeenCalledWith(
expect.objectContaining({
enableAgents: false,
enableAgents: true,
}),
);
});
describe('interactivity', () => {
it('should set interactive true when not headless', async () => {
it('should always set interactive true', async () => {
vi.mocked(isHeadlessMode).mockReturnValue(true);
await loadConfig(mockSettings, mockExtensionLoader, taskId);
expect(Config).toHaveBeenCalledWith(
expect.objectContaining({
interactive: true,
}),
);
vi.mocked(isHeadlessMode).mockReturnValue(false);
await loadConfig(mockSettings, mockExtensionLoader, taskId);
expect(Config).toHaveBeenCalledWith(
expect.objectContaining({
interactive: true,
enableInteractiveShell: true,
}),
);
});
it('should set interactive false when headless', async () => {
it('should set enableInteractiveShell based on headless mode', async () => {
vi.mocked(isHeadlessMode).mockReturnValue(false);
await loadConfig(mockSettings, mockExtensionLoader, taskId);
expect(Config).toHaveBeenCalledWith(
expect.objectContaining({
enableInteractiveShell: true,
}),
);
vi.mocked(isHeadlessMode).mockReturnValue(true);
await loadConfig(mockSettings, mockExtensionLoader, taskId);
expect(Config).toHaveBeenCalledWith(
expect.objectContaining({
interactive: false,
enableInteractiveShell: false,
}),
);
+2 -2
View File
@@ -125,10 +125,10 @@ export async function loadConfig(
trustedFolder: true,
extensionLoader,
checkpointing,
interactive: !isHeadlessMode(),
interactive: true,
enableInteractiveShell: !isHeadlessMode(),
ptyInfo: 'auto',
enableAgents: settings.experimental?.enableAgents ?? false,
enableAgents: settings.experimental?.enableAgents ?? true,
};
const fileService = new FileDiscoveryService(workspaceDir, {
+5 -21
View File
@@ -6,19 +6,12 @@
* SPDX-License-Identifier: Apache-2.0
*/
// --- Fast Path for Version ---
// We check for version flags at the very top to avoid loading any heavy dependencies.
// process.env.CLI_VERSION is defined during the build process by esbuild.
if (process.argv.includes('--version') || process.argv.includes('-v')) {
console.log(process.env['CLI_VERSION'] || 'unknown');
process.exit(0);
}
import { main } from './src/gemini.js';
import { FatalError, writeToStderr } from '@google/gemini-cli-core';
import { runExitCleanup } from './src/utils/cleanup.js';
// --- Global Entry Point ---
let writeToStderrFn: (message: string) => void = (msg) =>
process.stderr.write(msg);
// Suppress known race condition error in node-pty on Windows
// Tracking bug: https://github.com/microsoft/node-pty/issues/827
process.on('uncaughtException', (error) => {
@@ -35,22 +28,13 @@ process.on('uncaughtException', (error) => {
// For other errors, we rely on the default behavior, but since we attached a listener,
// we must manually replicate it.
if (error instanceof Error) {
writeToStderrFn(error.stack + '\n');
writeToStderr(error.stack + '\n');
} else {
writeToStderrFn(String(error) + '\n');
writeToStderr(String(error) + '\n');
}
process.exit(1);
});
const [{ main }, { FatalError, writeToStderr }, { runExitCleanup }] =
await Promise.all([
import('./src/gemini.js'),
import('@google/gemini-cli-core'),
import('./src/utils/cleanup.js'),
]);
writeToStderrFn = writeToStderr;
main().catch(async (error) => {
// Set a timeout to force exit if cleanup hangs
const cleanupTimeout = setTimeout(() => {
+2 -2
View File
@@ -1,6 +1,6 @@
{
"name": "@google/gemini-cli",
"version": "0.36.0-nightly.20260317.2f90b4653",
"version": "0.36.0-preview.7",
"description": "Gemini CLI",
"license": "Apache-2.0",
"repository": {
@@ -27,7 +27,7 @@
"dist"
],
"config": {
"sandboxImageUri": "us-docker.pkg.dev/gemini-code-dev/gemini-cli/sandbox:0.36.0-nightly.20260317.2f90b4653"
"sandboxImageUri": "us-docker.pkg.dev/gemini-code-dev/gemini-cli/sandbox:0.36.0-preview.7"
},
"dependencies": {
"@agentclientprotocol/sdk": "^0.16.1",
+18 -18
View File
@@ -93,7 +93,7 @@ describe('loadSandboxConfig', () => {
expect(config).toEqual({
enabled: true,
allowedPaths: [],
networkAccess: false,
networkAccess: true,
command: 'docker',
image: 'default/image',
});
@@ -122,7 +122,7 @@ describe('loadSandboxConfig', () => {
expect(config).toEqual({
enabled: true,
allowedPaths: [],
networkAccess: false,
networkAccess: true,
command: 'lxc',
image: 'default/image',
});
@@ -148,7 +148,7 @@ describe('loadSandboxConfig', () => {
expect(config).toEqual({
enabled: true,
allowedPaths: [],
networkAccess: false,
networkAccess: true,
command: 'sandbox-exec',
image: 'default/image',
});
@@ -161,7 +161,7 @@ describe('loadSandboxConfig', () => {
expect(config).toEqual({
enabled: true,
allowedPaths: [],
networkAccess: false,
networkAccess: true,
command: 'sandbox-exec',
image: 'default/image',
});
@@ -174,7 +174,7 @@ describe('loadSandboxConfig', () => {
expect(config).toEqual({
enabled: true,
allowedPaths: [],
networkAccess: false,
networkAccess: true,
command: 'docker',
image: 'default/image',
});
@@ -187,7 +187,7 @@ describe('loadSandboxConfig', () => {
expect(config).toEqual({
enabled: true,
allowedPaths: [],
networkAccess: false,
networkAccess: true,
command: 'podman',
image: 'default/image',
});
@@ -210,7 +210,7 @@ describe('loadSandboxConfig', () => {
expect(config).toEqual({
enabled: true,
allowedPaths: [],
networkAccess: false,
networkAccess: true,
command: 'podman',
image: 'default/image',
});
@@ -244,7 +244,7 @@ describe('loadSandboxConfig', () => {
expect(config).toEqual({
enabled: true,
allowedPaths: [],
networkAccess: false,
networkAccess: true,
command: 'docker',
image: 'env/image',
});
@@ -257,7 +257,7 @@ describe('loadSandboxConfig', () => {
expect(config).toEqual({
enabled: true,
allowedPaths: [],
networkAccess: false,
networkAccess: true,
command: 'docker',
image: 'default/image',
});
@@ -285,7 +285,7 @@ describe('loadSandboxConfig', () => {
expect(config).toEqual({
enabled: true,
allowedPaths: [],
networkAccess: false,
networkAccess: true,
command: 'docker',
image: 'default/image',
});
@@ -339,7 +339,7 @@ describe('loadSandboxConfig', () => {
enabled: true,
command: 'podman',
allowedPaths: [],
networkAccess: false,
networkAccess: true,
},
},
},
@@ -356,7 +356,7 @@ describe('loadSandboxConfig', () => {
enabled: true,
image: 'custom/image',
allowedPaths: [],
networkAccess: false,
networkAccess: true,
},
},
},
@@ -372,7 +372,7 @@ describe('loadSandboxConfig', () => {
sandbox: {
enabled: false,
allowedPaths: [],
networkAccess: false,
networkAccess: true,
},
},
},
@@ -388,7 +388,7 @@ describe('loadSandboxConfig', () => {
sandbox: {
enabled: true,
allowedPaths: ['/settings-path'],
networkAccess: false,
networkAccess: true,
},
},
},
@@ -410,7 +410,7 @@ describe('loadSandboxConfig', () => {
expect(config).toEqual({
enabled: true,
allowedPaths: [],
networkAccess: false,
networkAccess: true,
command: 'runsc',
image: 'default/image',
});
@@ -425,7 +425,7 @@ describe('loadSandboxConfig', () => {
expect(config).toEqual({
enabled: true,
allowedPaths: [],
networkAccess: false,
networkAccess: true,
command: 'runsc',
image: 'default/image',
});
@@ -442,7 +442,7 @@ describe('loadSandboxConfig', () => {
expect(config).toEqual({
enabled: true,
allowedPaths: [],
networkAccess: false,
networkAccess: true,
command: 'runsc',
image: 'default/image',
});
@@ -460,7 +460,7 @@ describe('loadSandboxConfig', () => {
expect(config).toEqual({
enabled: true,
allowedPaths: [],
networkAccess: false,
networkAccess: true,
command: 'runsc',
image: 'default/image',
});
+2 -2
View File
@@ -131,7 +131,7 @@ export async function loadSandboxConfig(
let sandboxValue: boolean | string | null | undefined;
let allowedPaths: string[] = [];
let networkAccess = false;
let networkAccess = true;
let customImage: string | undefined;
if (
@@ -142,7 +142,7 @@ export async function loadSandboxConfig(
const config = sandboxOption;
sandboxValue = config.enabled ? (config.command ?? true) : false;
allowedPaths = config.allowedPaths ?? [];
networkAccess = config.networkAccess ?? false;
networkAccess = config.networkAccess ?? true;
customImage = config.image;
} else if (typeof sandboxOption !== 'object' || sandboxOption === null) {
sandboxValue = sandboxOption;
@@ -400,7 +400,7 @@ describe('SettingsSchema', () => {
expect(setting).toBeDefined();
expect(setting.type).toBe('boolean');
expect(setting.category).toBe('Experimental');
expect(setting.default).toBe(false);
expect(setting.default).toBe(true);
expect(setting.requiresRestart).toBe(true);
expect(setting.showInDialog).toBe(false);
expect(setting.description).toBe('Enable local and remote subagents.');
+2 -1
View File
@@ -1932,7 +1932,7 @@ const SETTINGS_SCHEMA = {
label: 'Enable Agents',
category: 'Experimental',
requiresRestart: true,
default: false,
default: true,
description: 'Enable local and remote subagents.',
showInDialog: false,
},
@@ -3014,6 +3014,7 @@ export const SETTINGS_SCHEMA_DEFINITIONS: Record<
type: 'object',
properties: {
useGemini3_1: { type: 'boolean' },
useGemini3_1FlashLite: { type: 'boolean' },
useCustomTools: { type: 'boolean' },
hasAccessToPreview: { type: 'boolean' },
requestedModels: {
+2
View File
@@ -46,6 +46,7 @@ import { TerminalProvider } from './ui/contexts/TerminalContext.js';
import { isAlternateBufferEnabled } from './ui/hooks/useAlternateBuffer.js';
import { OverflowProvider } from './ui/contexts/OverflowContext.js';
import { profiler } from './ui/components/DebugProfiler.js';
import { initializeConsoleStore } from './ui/hooks/useConsoleMessages.js';
const SLOW_RENDER_MS = 200;
@@ -57,6 +58,7 @@ export async function startInteractiveUI(
resumedSessionData: ResumedSessionData | undefined,
initializationResult: InitializationResult,
) {
initializeConsoleStore();
// Never enter Ink alternate buffer mode when screen reader mode is enabled
// as there is no benefit of alternate buffer mode when using a screen reader
// and the Ink alternate buffer mode requires line wrapping harmful to
+2
View File
@@ -524,6 +524,8 @@ const baseMockUiState = {
nightly: false,
updateInfo: null,
pendingHistoryItems: [],
mainControlsRef: () => {},
rootUiRef: { current: null },
};
export const mockAppState: AppState = {
+1 -3
View File
@@ -70,9 +70,7 @@ describe('App', () => {
cleanUiDetailsVisible: true,
quittingMessages: null,
dialogsVisible: false,
mainControlsRef: {
current: null,
} as unknown as React.MutableRefObject<DOMElement | null>,
mainControlsRef: vi.fn(),
rootUiRef: {
current: null,
} as unknown as React.MutableRefObject<DOMElement | null>,
+22 -18
View File
@@ -4,17 +4,10 @@
* SPDX-License-Identifier: Apache-2.0
*/
import {
useMemo,
useState,
useCallback,
useEffect,
useRef,
useLayoutEffect,
} from 'react';
import { useMemo, useState, useCallback, useEffect, useRef } from 'react';
import {
type DOMElement,
measureElement,
ResizeObserver,
useApp,
useStdout,
useStdin,
@@ -397,7 +390,6 @@ export const AppContainer = (props: AppContainerProps) => {
const branchName = useGitBranchName(config.getTargetDir());
// Layout measurements
const mainControlsRef = useRef<DOMElement>(null);
// For performance profiling only
const rootUiRef = useRef<DOMElement>(null);
const lastTitleRef = useRef<string | null>(null);
@@ -1395,17 +1387,29 @@ Logging in with Google... Restarting Gemini CLI to continue.
streamingState === StreamingState.WaitingForConfirmation) &&
!proQuotaRequest;
const observerRef = useRef<ResizeObserver | null>(null);
const [controlsHeight, setControlsHeight] = useState(0);
useLayoutEffect(() => {
if (mainControlsRef.current) {
const fullFooterMeasurement = measureElement(mainControlsRef.current);
const roundedHeight = Math.round(fullFooterMeasurement.height);
if (roundedHeight > 0 && roundedHeight !== controlsHeight) {
setControlsHeight(roundedHeight);
}
const mainControlsRef = useCallback((node: DOMElement | null) => {
if (observerRef.current) {
observerRef.current.disconnect();
observerRef.current = null;
}
}, [buffer, terminalWidth, terminalHeight, controlsHeight, isInputActive]);
if (node) {
const observer = new ResizeObserver((entries) => {
const entry = entries[0];
if (entry) {
const roundedHeight = Math.round(entry.contentRect.height);
setControlsHeight((prev) =>
roundedHeight !== prev ? roundedHeight : prev,
);
}
});
observer.observe(node);
observerRef.current = observer;
}
}, []);
// Compute available terminal height based on controls measurement
const availableTerminalHeight = Math.max(
@@ -1491,4 +1491,47 @@ describe('AskUserDialog', () => {
expect(frame).toContain('3. Option 3');
});
});
it('allows the question to exceed 15 lines in a tall terminal', async () => {
const longQuestion = Array.from(
{ length: 25 },
(_, i) => `Line ${i + 1}`,
).join('\n');
const questions: Question[] = [
{
question: longQuestion,
header: 'Tall Test',
type: QuestionType.CHOICE,
options: [
{ label: 'Option 1', description: 'D1' },
{ label: 'Option 2', description: 'D2' },
{ label: 'Option 3', description: 'D3' },
],
multiSelect: false,
unconstrainedHeight: false,
},
];
const { lastFrame, waitUntilReady } = await renderWithProviders(
<AskUserDialog
questions={questions}
onSubmit={vi.fn()}
onCancel={vi.fn()}
width={80}
availableHeight={40} // Tall terminal
/>,
{ width: 80 },
);
await waitFor(async () => {
await waitUntilReady();
const frame = lastFrame();
// Should show more than 15 lines of the question
// (The limit was previously 15, so showing Line 20 proves it's working)
expect(frame).toContain('Line 20');
expect(frame).toContain('Line 25');
// Should still show the options
expect(frame).toContain('1. Option 1');
});
});
});
@@ -855,13 +855,7 @@ const ChoiceQuestionView: React.FC<ChoiceQuestionViewProps> = ({
listHeight && !isAlternateBuffer
? question.unconstrainedHeight
? Math.max(1, listHeight - selectionItems.length * 2)
: Math.min(
15,
Math.max(
1,
listHeight - Math.max(DIALOG_PADDING, reservedListHeight),
),
)
: Math.max(1, listHeight - Math.max(DIALOG_PADDING, reservedListHeight))
: undefined;
const maxItemsToShow =
@@ -35,10 +35,7 @@ vi.mock('./shared/ScrollableList.js', () => ({
describe('DetailedMessagesDisplay', () => {
beforeEach(() => {
vi.mocked(useConsoleMessages).mockReturnValue({
consoleMessages: [],
clearConsoleMessages: vi.fn(),
});
vi.mocked(useConsoleMessages).mockReturnValue([]);
});
it('renders nothing when messages are empty', async () => {
const { lastFrame, unmount } = await renderWithProviders(
@@ -58,10 +55,7 @@ describe('DetailedMessagesDisplay', () => {
{ type: 'error', content: 'Error message', count: 1 },
{ type: 'debug', content: 'Debug message', count: 1 },
];
vi.mocked(useConsoleMessages).mockReturnValue({
consoleMessages: messages,
clearConsoleMessages: vi.fn(),
});
vi.mocked(useConsoleMessages).mockReturnValue(messages);
const { lastFrame, unmount } = await renderWithProviders(
<DetailedMessagesDisplay maxHeight={20} width={80} hasFocus={true} />,
@@ -79,10 +73,7 @@ describe('DetailedMessagesDisplay', () => {
const messages: ConsoleMessageItem[] = [
{ type: 'error', content: 'Error message', count: 1 },
];
vi.mocked(useConsoleMessages).mockReturnValue({
consoleMessages: messages,
clearConsoleMessages: vi.fn(),
});
vi.mocked(useConsoleMessages).mockReturnValue(messages);
const { lastFrame, unmount } = await renderWithProviders(
<DetailedMessagesDisplay maxHeight={20} width={80} hasFocus={true} />,
@@ -98,10 +89,7 @@ describe('DetailedMessagesDisplay', () => {
const messages: ConsoleMessageItem[] = [
{ type: 'error', content: 'Error message', count: 1 },
];
vi.mocked(useConsoleMessages).mockReturnValue({
consoleMessages: messages,
clearConsoleMessages: vi.fn(),
});
vi.mocked(useConsoleMessages).mockReturnValue(messages);
const { lastFrame, unmount } = await renderWithProviders(
<DetailedMessagesDisplay maxHeight={20} width={80} hasFocus={true} />,
@@ -117,10 +105,7 @@ describe('DetailedMessagesDisplay', () => {
const messages: ConsoleMessageItem[] = [
{ type: 'log', content: 'Repeated message', count: 5 },
];
vi.mocked(useConsoleMessages).mockReturnValue({
consoleMessages: messages,
clearConsoleMessages: vi.fn(),
});
vi.mocked(useConsoleMessages).mockReturnValue(messages);
const { lastFrame, unmount } = await renderWithProviders(
<DetailedMessagesDisplay maxHeight={10} width={80} hasFocus={false} />,
@@ -29,7 +29,7 @@ export const DetailedMessagesDisplay: React.FC<
> = ({ maxHeight, width, hasFocus }) => {
const scrollableListRef = useRef<ScrollableListRef<ConsoleMessageItem>>(null);
const { consoleMessages } = useConsoleMessages();
const consoleMessages = useConsoleMessages();
const config = useConfig();
const messages = useMemo(() => {
@@ -21,6 +21,10 @@ import {
type UIState,
} from '../contexts/UIStateContext.js';
import { type IndividualToolCallDisplay } from '../types.js';
import {
type ConfirmingToolState,
useConfirmingTool,
} from '../hooks/useConfirmingTool.js';
// Mock dependencies
const mockUseSettings = vi.fn().mockReturnValue({
@@ -53,6 +57,10 @@ vi.mock('../hooks/useAlternateBuffer.js', () => ({
useAlternateBuffer: vi.fn(),
}));
vi.mock('../hooks/useConfirmingTool.js', () => ({
useConfirmingTool: vi.fn(),
}));
vi.mock('./AppHeader.js', () => ({
AppHeader: ({ showDetails = true }: { showDetails?: boolean }) => (
<Text>{showDetails ? 'AppHeader(full)' : 'AppHeader(minimal)'}</Text>
@@ -503,6 +511,54 @@ describe('MainContent', () => {
unmount();
});
it('renders a subagent with a complete box including bottom border', async () => {
const subagentCall = {
callId: 'subagent-1',
name: 'codebase_investigator',
description: 'Investigating codebase',
status: CoreToolCallStatus.Executing,
kind: 'agent',
resultDisplay: {
isSubagentProgress: true,
agentName: 'codebase_investigator',
recentActivity: [
{
id: '1',
type: 'tool_call',
content: 'run_shell_command',
args: '{"command": "echo hello"}',
status: 'running',
},
],
state: 'running',
},
} as Partial<IndividualToolCallDisplay> as IndividualToolCallDisplay;
const uiState = {
...defaultMockUiState,
history: [{ id: 1, type: 'user', text: 'Investigate' }],
pendingHistoryItems: [
{
type: 'tool_group' as const,
tools: [subagentCall],
borderBottom: true,
},
],
};
const { lastFrame, unmount } = await renderWithProviders(<MainContent />, {
uiState: uiState as Partial<UIState>,
config: makeFakeConfig({ useAlternateBuffer: false }),
});
await waitFor(() => {
expect(lastFrame()).toContain('codebase_investigator');
});
expect(lastFrame()).toMatchSnapshot();
unmount();
});
it('renders a split tool group without a gap between static and pending areas', async () => {
const toolCalls = [
{
@@ -547,13 +603,124 @@ describe('MainContent', () => {
const { lastFrame, unmount } = await renderWithProviders(<MainContent />, {
uiState: uiState as Partial<UIState>,
});
const output = lastFrame();
// Verify Part 1 and Part 2 are rendered.
expect(output).toContain('Part 1');
expect(output).toContain('Part 2');
await waitFor(() => {
const output = lastFrame();
// Verify Part 1 and Part 2 are rendered.
expect(output).toContain('Part 1');
expect(output).toContain('Part 2');
});
// The snapshot will be the best way to verify there is no gap (empty line) between them.
expect(output).toMatchSnapshot();
expect(lastFrame()).toMatchSnapshot();
unmount();
});
it('renders a ToolConfirmationQueue without an extra line when preceded by hidden tools', async () => {
const { ApprovalMode, WRITE_FILE_DISPLAY_NAME } = await import(
'@google/gemini-cli-core'
);
const hiddenToolCalls = [
{
callId: 'tool-hidden',
name: WRITE_FILE_DISPLAY_NAME,
approvalMode: ApprovalMode.PLAN,
status: CoreToolCallStatus.Success,
resultDisplay: 'Hidden content',
} as Partial<IndividualToolCallDisplay> as IndividualToolCallDisplay,
];
const confirmingTool = {
tool: {
callId: 'call-1',
name: 'exit_plan_mode',
status: CoreToolCallStatus.AwaitingApproval,
confirmationDetails: {
type: 'exit_plan_mode' as const,
planPath: '/path/to/plan',
},
},
index: 1,
total: 1,
};
const uiState = {
...defaultMockUiState,
history: [{ id: 1, type: 'user', text: 'Apply plan' }],
pendingHistoryItems: [
{
type: 'tool_group' as const,
tools: hiddenToolCalls,
borderBottom: true,
},
],
};
// We need to mock useConfirmingTool to return our confirmingTool
vi.mocked(useConfirmingTool).mockReturnValue(
confirmingTool as unknown as ConfirmingToolState,
);
mockUseSettings.mockReturnValue(
createMockSettings({
security: { enablePermanentToolApproval: true },
ui: { errorVerbosity: 'full' },
}),
);
const { lastFrame, unmount } = await renderWithProviders(<MainContent />, {
uiState: uiState as Partial<UIState>,
config: makeFakeConfig({ useAlternateBuffer: false }),
});
await waitFor(() => {
const output = lastFrame();
// The output should NOT contain 'Hidden content'
expect(output).not.toContain('Hidden content');
// The output should contain the confirmation header
expect(output).toContain('Ready to start implementation?');
});
// Snapshot will reveal if there are extra blank lines
expect(lastFrame()).toMatchSnapshot();
unmount();
});
it('renders a spurious line when a tool group has only hidden tools and borderBottom true', async () => {
const { ApprovalMode, WRITE_FILE_DISPLAY_NAME } = await import(
'@google/gemini-cli-core'
);
const uiState = {
...defaultMockUiState,
history: [{ id: 1, type: 'user', text: 'Apply plan' }],
pendingHistoryItems: [
{
type: 'tool_group' as const,
tools: [
{
callId: 'tool-1',
name: WRITE_FILE_DISPLAY_NAME,
approvalMode: ApprovalMode.PLAN,
status: CoreToolCallStatus.Success,
resultDisplay: 'hidden',
} as Partial<IndividualToolCallDisplay> as IndividualToolCallDisplay,
],
borderBottom: true,
},
],
};
const { lastFrame, unmount } = await renderWithProviders(<MainContent />, {
uiState: uiState as Partial<UIState>,
config: makeFakeConfig({ useAlternateBuffer: false }),
});
await waitFor(() => {
expect(lastFrame()).toContain('Apply plan');
});
// This snapshot will show no spurious line because the group is now correctly suppressed.
expect(lastFrame()).toMatchSnapshot();
unmount();
});
@@ -21,7 +21,6 @@ import {
PREVIEW_GEMINI_FLASH_MODEL,
PREVIEW_GEMINI_3_1_FLASH_LITE_MODEL,
AuthType,
UserTierId,
} from '@google/gemini-cli-core';
import type { Config, ModelSlashCommandEvent } from '@google/gemini-cli-core';
@@ -53,9 +52,9 @@ describe('<ModelDialog />', () => {
const mockOnClose = vi.fn();
const mockGetHasAccessToPreviewModel = vi.fn();
const mockGetGemini31LaunchedSync = vi.fn();
const mockGetGemini31FlashLiteLaunchedSync = vi.fn();
const mockGetProModelNoAccess = vi.fn();
const mockGetProModelNoAccessSync = vi.fn();
const mockGetUserTier = vi.fn();
interface MockConfig extends Partial<Config> {
setModel: (model: string, isTemporary?: boolean) => void;
@@ -63,9 +62,9 @@ describe('<ModelDialog />', () => {
getHasAccessToPreviewModel: () => boolean;
getIdeMode: () => boolean;
getGemini31LaunchedSync: () => boolean;
getGemini31FlashLiteLaunchedSync: () => boolean;
getProModelNoAccess: () => Promise<boolean>;
getProModelNoAccessSync: () => boolean;
getUserTier: () => UserTierId | undefined;
}
const mockConfig: MockConfig = {
@@ -74,9 +73,9 @@ describe('<ModelDialog />', () => {
getHasAccessToPreviewModel: mockGetHasAccessToPreviewModel,
getIdeMode: () => false,
getGemini31LaunchedSync: mockGetGemini31LaunchedSync,
getGemini31FlashLiteLaunchedSync: mockGetGemini31FlashLiteLaunchedSync,
getProModelNoAccess: mockGetProModelNoAccess,
getProModelNoAccessSync: mockGetProModelNoAccessSync,
getUserTier: mockGetUserTier,
};
beforeEach(() => {
@@ -84,9 +83,9 @@ describe('<ModelDialog />', () => {
mockGetModel.mockReturnValue(DEFAULT_GEMINI_MODEL_AUTO);
mockGetHasAccessToPreviewModel.mockReturnValue(false);
mockGetGemini31LaunchedSync.mockReturnValue(false);
mockGetGemini31FlashLiteLaunchedSync.mockReturnValue(false);
mockGetProModelNoAccess.mockResolvedValue(false);
mockGetProModelNoAccessSync.mockReturnValue(false);
mockGetUserTier.mockReturnValue(UserTierId.STANDARD);
// Default implementation for getDisplayString
mockGetDisplayString.mockImplementation((val: string) => {
@@ -131,7 +130,7 @@ describe('<ModelDialog />', () => {
mockGetProModelNoAccessSync.mockReturnValue(true);
mockGetProModelNoAccess.mockResolvedValue(true);
mockGetHasAccessToPreviewModel.mockReturnValue(true);
mockGetUserTier.mockReturnValue(UserTierId.FREE);
mockGetGemini31FlashLiteLaunchedSync.mockReturnValue(true);
mockGetDisplayString.mockImplementation((val: string) => val);
const { lastFrame, unmount } = await renderComponent();
@@ -437,33 +436,11 @@ describe('<ModelDialog />', () => {
unmount();
});
it('hides Flash Lite Preview model for users with pro access', async () => {
it('shows Flash Lite Preview model regardless of tier when flag is enabled', async () => {
mockGetProModelNoAccessSync.mockReturnValue(false);
mockGetProModelNoAccess.mockResolvedValue(false);
mockGetHasAccessToPreviewModel.mockReturnValue(true);
const { lastFrame, stdin, waitUntilReady, unmount } =
await renderComponent();
// Go to manual view
await act(async () => {
stdin.write('\u001B[B'); // Manual
});
await waitUntilReady();
await act(async () => {
stdin.write('\r');
});
await waitUntilReady();
const output = lastFrame();
expect(output).not.toContain(PREVIEW_GEMINI_3_1_FLASH_LITE_MODEL);
unmount();
});
it('shows Flash Lite Preview model for free tier users', async () => {
mockGetProModelNoAccessSync.mockReturnValue(false);
mockGetProModelNoAccess.mockResolvedValue(false);
mockGetHasAccessToPreviewModel.mockReturnValue(true);
mockGetUserTier.mockReturnValue(UserTierId.FREE);
mockGetGemini31FlashLiteLaunchedSync.mockReturnValue(true);
const { lastFrame, stdin, waitUntilReady, unmount } =
await renderComponent();
+11 -7
View File
@@ -23,7 +23,6 @@ import {
AuthType,
PREVIEW_GEMINI_3_1_CUSTOM_TOOLS_MODEL,
isProModel,
UserTierId,
} from '@google/gemini-cli-core';
import { useKeypress } from '../hooks/useKeypress.js';
import { theme } from '../semantic-colors.js';
@@ -63,6 +62,8 @@ export function ModelDialog({ onClose }: ModelDialogProps): React.JSX.Element {
const shouldShowPreviewModels = config?.getHasAccessToPreviewModel();
const useGemini31 = config?.getGemini31LaunchedSync?.() ?? false;
const useGemini31FlashLite =
config?.getGemini31FlashLiteLaunchedSync?.() ?? false;
const selectedAuthType = settings.merged.security.auth.selectedType;
const useCustomToolModel =
useGemini31 && selectedAuthType === AuthType.USE_GEMINI;
@@ -86,6 +87,7 @@ export function ModelDialog({ onClose }: ModelDialogProps): React.JSX.Element {
PREVIEW_GEMINI_MODEL,
PREVIEW_GEMINI_3_1_MODEL,
PREVIEW_GEMINI_3_1_CUSTOM_TOOLS_MODEL,
PREVIEW_GEMINI_3_1_FLASH_LITE_MODEL,
PREVIEW_GEMINI_FLASH_MODEL,
];
if (manualModels.includes(preferredModel)) {
@@ -187,7 +189,6 @@ export function ModelDialog({ onClose }: ModelDialogProps): React.JSX.Element {
}, [config, shouldShowPreviewModels, manualModelSelected, useGemini31]);
const manualOptions = useMemo(() => {
const isFreeTier = config?.getUserTier() === UserTierId.FREE;
// --- DYNAMIC PATH ---
if (
config?.getExperimentalDynamicModelConfiguration?.() === true &&
@@ -204,13 +205,13 @@ export function ModelDialog({ onClose }: ModelDialogProps): React.JSX.Element {
if (m.tier === 'auto') return false;
// Pro models are shown for users with pro access
if (!hasAccessToProModel && m.tier === 'pro') return false;
// 3.1 Preview Flash-lite is only available on free tier
if (m.tier === 'flash-lite' && m.isPreview && !isFreeTier)
return false;
// Flag Guard: Versioned models only show if their flag is active.
if (id === PREVIEW_GEMINI_3_1_MODEL && !useGemini31) return false;
if (id === PREVIEW_GEMINI_3_1_FLASH_LITE_MODEL && !useGemini31)
if (
id === PREVIEW_GEMINI_3_1_FLASH_LITE_MODEL &&
!useGemini31FlashLite
)
return false;
return true;
@@ -218,11 +219,13 @@ export function ModelDialog({ onClose }: ModelDialogProps): React.JSX.Element {
.map(([id, m]) => {
const resolvedId = config.modelConfigService.resolveModelId(id, {
useGemini3_1: useGemini31,
useGemini3_1FlashLite: useGemini31FlashLite,
useCustomTools: useCustomToolModel,
});
// Title ID is the resolved ID without custom tools flag
const titleId = config.modelConfigService.resolveModelId(id, {
useGemini3_1: useGemini31,
useGemini3_1FlashLite: useGemini31FlashLite,
});
return {
value: resolvedId,
@@ -284,7 +287,7 @@ export function ModelDialog({ onClose }: ModelDialogProps): React.JSX.Element {
},
];
if (isFreeTier) {
if (useGemini31FlashLite) {
previewOptions.push({
value: PREVIEW_GEMINI_3_1_FLASH_LITE_MODEL,
title: getDisplayString(PREVIEW_GEMINI_3_1_FLASH_LITE_MODEL),
@@ -304,6 +307,7 @@ export function ModelDialog({ onClose }: ModelDialogProps): React.JSX.Element {
}, [
shouldShowPreviewModels,
useGemini31,
useGemini31FlashLite,
useCustomToolModel,
hasAccessToProModel,
config,
@@ -92,6 +92,7 @@ const buildModelRows = (
config: Config,
quotas?: RetrieveUserQuotaResponse,
useGemini3_1 = false,
useGemini3_1FlashLite = false,
useCustomToolModel = false,
) => {
const getBaseModelName = (name: string) => name.replace('-001', '');
@@ -124,7 +125,12 @@ const buildModelRows = (
?.filter(
(b) =>
b.modelId &&
isActiveModel(b.modelId, useGemini3_1, useCustomToolModel) &&
isActiveModel(
b.modelId,
useGemini3_1,
useGemini3_1FlashLite,
useCustomToolModel,
) &&
!usedModelNames.has(getDisplayString(b.modelId, config)),
)
.map((bucket) => ({
@@ -152,6 +158,7 @@ const ModelUsageTable: React.FC<{
pooledLimit?: number;
pooledResetTime?: string;
useGemini3_1?: boolean;
useGemini3_1FlashLite?: boolean;
useCustomToolModel?: boolean;
}> = ({
models,
@@ -164,6 +171,7 @@ const ModelUsageTable: React.FC<{
pooledLimit,
pooledResetTime,
useGemini3_1,
useGemini3_1FlashLite,
useCustomToolModel,
}) => {
const { stdout } = useStdout();
@@ -173,6 +181,7 @@ const ModelUsageTable: React.FC<{
config,
quotas,
useGemini3_1,
useGemini3_1FlashLite,
useCustomToolModel,
);
@@ -541,6 +550,8 @@ export const StatsDisplay: React.FC<StatsDisplayProps> = ({
const settings = useSettings();
const config = useConfig();
const useGemini3_1 = config.getGemini31LaunchedSync?.() ?? false;
const useGemini3_1FlashLite =
config.getGemini31FlashLiteLaunchedSync?.() ?? false;
const useCustomToolModel =
useGemini3_1 &&
config.getContentGeneratorConfig().authType === AuthType.USE_GEMINI;
@@ -697,6 +708,7 @@ export const StatsDisplay: React.FC<StatsDisplayProps> = ({
pooledLimit={pooledLimit}
pooledResetTime={pooledResetTime}
useGemini3_1={useGemini3_1}
useGemini3_1FlashLite={useGemini3_1FlashLite}
useCustomToolModel={useCustomToolModel}
/>
{renderFooter()}
@@ -91,6 +91,19 @@ exports[`MainContent > MainContent Tool Output Height Logic > 'Normal mode - Unc
"
`;
exports[`MainContent > renders a ToolConfirmationQueue without an extra line when preceded by hidden tools 1`] = `
"AppHeader(full)
▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀
> Apply plan
▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄
╭──────────────────────────────────────────────────────────────────────────────╮
│ Ready to start implementation? │
│ │
│ Error reading plan: Storage must be initialized before use │
╰──────────────────────────────────────────────────────────────────────────────╯
"
`;
exports[`MainContent > renders a split tool group without a gap between static and pending areas 1`] = `
"AppHeader(full)
╭──────────────────────────────────────────────────────────────────────────╮
@@ -105,6 +118,30 @@ exports[`MainContent > renders a split tool group without a gap between static a
"
`;
exports[`MainContent > renders a spurious line when a tool group has only hidden tools and borderBottom true 1`] = `
"AppHeader(full)
▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀
> Apply plan
▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄
"
`;
exports[`MainContent > renders a subagent with a complete box including bottom border 1`] = `
"AppHeader(full)
▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀
> Investigate
▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄
╭──────────────────────────────────────────────────────────────────────────╮
│ ≡ Running Agent... (ctrl+o to collapse) │
│ │
│ Running subagent codebase_investigator... │
│ │
│ ⠋ run_shell_command echo hello │
│ │
╰──────────────────────────────────────────────────────────────────────────╯
"
`;
exports[`MainContent > renders mixed history items (user + gemini) with single line padding between them 1`] = `
"ScrollableList
AppHeader(full)
@@ -172,12 +172,10 @@ export const ToolGroupMessage: React.FC<ToolGroupMessageProps> = ({
// If all tools are filtered out (e.g., in-progress AskUser tools, low-verbosity
// internal errors, plan-mode hidden write/edit), we should not emit standalone
// border fragments. The only case where an empty group should render is the
// explicit "closing slice" (tools: []) used to bridge static/pending sections.
// explicit "closing slice" (tools: []) used to bridge static/pending sections,
// and only if it's actually continuing an open box from above.
const isExplicitClosingSlice = allToolCalls.length === 0;
if (
visibleToolCalls.length === 0 &&
(!isExplicitClosingSlice || borderBottomOverride !== true)
) {
if (visibleToolCalls.length === 0 && !isExplicitClosingSlice) {
return null;
}
@@ -269,19 +267,20 @@ export const ToolGroupMessage: React.FC<ToolGroupMessageProps> = ({
We have to keep the bottom border separate so it doesn't get
drawn over by the sticky header directly inside it.
*/
(visibleToolCalls.length > 0 || borderBottomOverride !== undefined) && (
<Box
height={0}
width={contentWidth}
borderLeft={true}
borderRight={true}
borderTop={false}
borderBottom={borderBottomOverride ?? true}
borderColor={borderColor}
borderDimColor={borderDimColor}
borderStyle="round"
/>
)
(visibleToolCalls.length > 0 || borderBottomOverride !== undefined) &&
borderBottomOverride !== false && (
<Box
height={0}
width={contentWidth}
borderLeft={true}
borderRight={true}
borderTop={false}
borderBottom={borderBottomOverride ?? true}
borderColor={borderColor}
borderDimColor={borderDimColor}
borderStyle="round"
/>
)
}
</Box>
);
@@ -0,0 +1,160 @@
/**
* @license
* Copyright 2026 Google LLC
* SPDX-License-Identifier: Apache-2.0
*/
import { renderWithProviders } from '../../../test-utils/render.js';
import { describe, it, expect } from 'vitest';
import { ToolGroupMessage } from './ToolGroupMessage.js';
import {
makeFakeConfig,
CoreToolCallStatus,
ApprovalMode,
WRITE_FILE_DISPLAY_NAME,
Kind,
} from '@google/gemini-cli-core';
import os from 'node:os';
import { createMockSettings } from '../../../test-utils/settings.js';
import type { IndividualToolCallDisplay } from '../../types.js';
describe('ToolGroupMessage Regression Tests', () => {
const baseMockConfig = makeFakeConfig({
model: 'gemini-pro',
targetDir: os.tmpdir(),
});
const fullVerbositySettings = createMockSettings({
ui: { errorVerbosity: 'full' },
});
const createToolCall = (
overrides: Partial<IndividualToolCallDisplay> = {},
): IndividualToolCallDisplay =>
({
callId: 'tool-123',
name: 'test-tool',
status: CoreToolCallStatus.Success,
...overrides,
}) as IndividualToolCallDisplay;
const createItem = (tools: IndividualToolCallDisplay[]) => ({
id: 1,
type: 'tool_group' as const,
tools,
});
it('Plan Mode: suppresses phantom tool group (hidden tools)', async () => {
const toolCalls = [
createToolCall({
name: WRITE_FILE_DISPLAY_NAME,
approvalMode: ApprovalMode.PLAN,
status: CoreToolCallStatus.Success,
}),
];
const item = createItem(toolCalls);
const { lastFrame, unmount } = await renderWithProviders(
<ToolGroupMessage
terminalWidth={80}
item={item}
toolCalls={toolCalls}
borderBottom={true}
/>,
{ config: baseMockConfig, settings: fullVerbositySettings },
);
expect(lastFrame({ allowEmpty: true })).toBe('');
unmount();
});
it('Agent Case: suppresses the bottom border box for ongoing agents (no vertical ticks)', async () => {
const toolCalls = [
createToolCall({
name: 'agent',
kind: Kind.Agent,
status: CoreToolCallStatus.Executing,
resultDisplay: {
isSubagentProgress: true,
agentName: 'TestAgent',
state: 'running',
recentActivity: [],
},
}),
];
const item = createItem(toolCalls);
const { lastFrame, unmount } = await renderWithProviders(
<ToolGroupMessage
terminalWidth={80}
item={item}
toolCalls={toolCalls}
borderBottom={false} // Ongoing
/>,
{ config: baseMockConfig, settings: fullVerbositySettings },
);
const output = lastFrame();
expect(output).toContain('Running Agent...');
// It should render side borders from the content
expect(output).toContain('│');
// It should NOT render the bottom border box (no corners ╰ ╯)
expect(output).not.toContain('╰');
expect(output).not.toContain('╯');
unmount();
});
it('Agent Case: renders a bottom border horizontal line for completed agents', async () => {
const toolCalls = [
createToolCall({
name: 'agent',
kind: Kind.Agent,
status: CoreToolCallStatus.Success,
resultDisplay: {
isSubagentProgress: true,
agentName: 'TestAgent',
state: 'completed',
recentActivity: [],
},
}),
];
const item = createItem(toolCalls);
const { lastFrame, unmount } = await renderWithProviders(
<ToolGroupMessage
terminalWidth={80}
item={item}
toolCalls={toolCalls}
borderBottom={true} // Completed
/>,
{ config: baseMockConfig, settings: fullVerbositySettings },
);
const output = lastFrame();
// Verify it rendered subagent content
expect(output).toContain('Agent');
// It should render the bottom horizontal line
expect(output).toContain(
'╰──────────────────────────────────────────────────────────────────────────╯',
);
unmount();
});
it('Bridges: still renders a bridge if it has a top border', async () => {
const toolCalls: IndividualToolCallDisplay[] = [];
const item = createItem(toolCalls);
const { lastFrame, unmount } = await renderWithProviders(
<ToolGroupMessage
terminalWidth={80}
item={item}
toolCalls={toolCalls}
borderTop={true}
borderBottom={true}
/>,
{ config: baseMockConfig, settings: fullVerbositySettings },
);
expect(lastFrame({ allowEmpty: true })).not.toBe('');
unmount();
});
});
@@ -190,7 +190,7 @@ export interface UIState {
sessionStats: SessionStatsState;
terminalWidth: number;
terminalHeight: number;
mainControlsRef: React.MutableRefObject<DOMElement | null>;
mainControlsRef: React.RefCallback<DOMElement | null>;
// NOTE: This is for performance profiling only.
rootUiRef: React.MutableRefObject<DOMElement | null>;
currentIDE: IdeInfo | null;
@@ -7,76 +7,93 @@
import { act, useCallback } from 'react';
import { vi } from 'vitest';
import { render } from '../../test-utils/render.js';
import { useConsoleMessages } from './useConsoleMessages.js';
import { CoreEvent, type ConsoleLogPayload } from '@google/gemini-cli-core';
// Mock coreEvents
let consoleLogHandler: ((payload: ConsoleLogPayload) => void) | undefined;
import {
useConsoleMessages,
useErrorCount,
initializeConsoleStore,
} from './useConsoleMessages.js';
import { coreEvents } from '@google/gemini-cli-core';
vi.mock('@google/gemini-cli-core', async (importOriginal) => {
// eslint-disable-next-line @typescript-eslint/no-explicit-any
const actual = (await importOriginal()) as any;
const actual = await importOriginal();
const handlers = new Map<string, (payload: unknown) => void>();
return {
...actual,
...(actual as Record<string, unknown>),
coreEvents: {
on: vi.fn((event, handler) => {
if (event === CoreEvent.ConsoleLog) {
consoleLogHandler = handler;
}
...((actual as Record<string, unknown>)['coreEvents'] as Record<
string,
unknown
>),
on: vi.fn((event: string, handler: (payload: unknown) => void) => {
handlers.set(event, handler);
}),
off: vi.fn((event) => {
if (event === CoreEvent.ConsoleLog) {
consoleLogHandler = undefined;
}
off: vi.fn((event: string) => {
handlers.delete(event);
}),
emitConsoleLog: vi.fn(),
// Helper for testing to trigger the handlers
_trigger: (event: string, payload: unknown) => {
handlers.get(event)?.(payload);
},
},
};
});
describe('useConsoleMessages', () => {
let unmounts: Array<() => void> = [];
beforeEach(() => {
vi.useFakeTimers();
consoleLogHandler = undefined;
initializeConsoleStore();
});
afterEach(() => {
for (const unmount of unmounts) {
try {
unmount();
} catch (_e) {
// Ignore unmount errors
}
}
unmounts = [];
vi.runOnlyPendingTimers();
vi.useRealTimers();
vi.restoreAllMocks();
});
const useTestableConsoleMessages = () => {
const { ...rest } = useConsoleMessages();
const consoleMessages = useConsoleMessages();
const log = useCallback((content: string) => {
if (consoleLogHandler) {
consoleLogHandler({ type: 'log', content });
}
// @ts-expect-error - internal testing helper
coreEvents._trigger('console-log', { type: 'log', content });
}, []);
const error = useCallback((content: string) => {
if (consoleLogHandler) {
consoleLogHandler({ type: 'error', content });
}
// @ts-expect-error - internal testing helper
coreEvents._trigger('console-log', { type: 'error', content });
}, []);
const clearConsoleMessages = useCallback(() => {
initializeConsoleStore();
}, []);
return {
...rest,
consoleMessages,
log,
error,
clearConsoleMessages: rest.clearConsoleMessages,
clearConsoleMessages,
};
};
const renderConsoleMessagesHook = async () => {
let hookResult: ReturnType<typeof useTestableConsoleMessages>;
let hookResult: ReturnType<typeof useTestableConsoleMessages> | undefined;
function TestComponent() {
hookResult = useTestableConsoleMessages();
return null;
}
const { unmount } = await render(<TestComponent />);
unmounts.push(unmount);
return {
result: {
get current() {
return hookResult;
return hookResult!;
},
},
unmount,
@@ -93,10 +110,7 @@ describe('useConsoleMessages', () => {
act(() => {
result.current.log('Test message');
});
await act(async () => {
await vi.advanceTimersByTimeAsync(60);
vi.runAllTimers();
});
expect(result.current.consoleMessages).toEqual([
@@ -111,10 +125,7 @@ describe('useConsoleMessages', () => {
result.current.log('Test message');
result.current.log('Test message');
result.current.log('Test message');
});
await act(async () => {
await vi.advanceTimersByTimeAsync(60);
vi.runAllTimers();
});
expect(result.current.consoleMessages).toEqual([
@@ -128,10 +139,7 @@ describe('useConsoleMessages', () => {
act(() => {
result.current.log('First message');
result.current.error('Second message');
});
await act(async () => {
await vi.advanceTimersByTimeAsync(60);
vi.runAllTimers();
});
expect(result.current.consoleMessages).toEqual([
@@ -139,53 +147,85 @@ describe('useConsoleMessages', () => {
{ type: 'error', content: 'Second message', count: 1 },
]);
});
});
it('should clear all messages when clearConsoleMessages is called', async () => {
const { result } = await renderConsoleMessagesHook();
describe('useErrorCount', () => {
let unmounts: Array<() => void> = [];
act(() => {
result.current.log('A message');
});
await act(async () => {
await vi.advanceTimersByTimeAsync(60);
});
expect(result.current.consoleMessages).toHaveLength(1);
act(() => {
result.current.clearConsoleMessages();
});
expect(result.current.consoleMessages).toHaveLength(0);
beforeEach(() => {
vi.useFakeTimers();
initializeConsoleStore();
});
it('should clear the pending timeout when clearConsoleMessages is called', async () => {
const { result } = await renderConsoleMessagesHook();
const clearTimeoutSpy = vi.spyOn(global, 'clearTimeout');
act(() => {
result.current.log('A message');
});
act(() => {
result.current.clearConsoleMessages();
});
expect(clearTimeoutSpy).toHaveBeenCalled();
// clearTimeoutSpy.mockRestore() is handled by afterEach restoreAllMocks
afterEach(() => {
for (const unmount of unmounts) {
try {
unmount();
} catch (_e) {
// Ignore unmount errors
}
}
unmounts = [];
vi.runOnlyPendingTimers();
vi.useRealTimers();
vi.restoreAllMocks();
});
it('should clean up the timeout on unmount', async () => {
const { result, unmount } = await renderConsoleMessagesHook();
const clearTimeoutSpy = vi.spyOn(global, 'clearTimeout');
const renderErrorCountHook = async () => {
let hookResult: ReturnType<typeof useErrorCount>;
function TestComponent() {
hookResult = useErrorCount();
return null;
}
const { unmount } = await render(<TestComponent />);
unmounts.push(unmount);
return {
result: {
get current() {
return hookResult;
},
},
unmount,
};
};
it('should initialize with an error count of 0', async () => {
const { result } = await renderErrorCountHook();
expect(result.current.errorCount).toBe(0);
});
it('should increment error count when an error is logged', async () => {
const { result } = await renderErrorCountHook();
act(() => {
// @ts-expect-error - internal testing helper
coreEvents._trigger('console-log', { type: 'error', content: 'error' });
vi.runAllTimers();
});
expect(result.current.errorCount).toBe(1);
});
it('should not increment error count for non-error logs', async () => {
const { result } = await renderErrorCountHook();
act(() => {
// @ts-expect-error - internal testing helper
coreEvents._trigger('console-log', { type: 'log', content: 'log' });
vi.runAllTimers();
});
expect(result.current.errorCount).toBe(0);
});
it('should clear the error count', async () => {
const { result } = await renderErrorCountHook();
act(() => {
// @ts-expect-error - internal testing helper
coreEvents._trigger('console-log', { type: 'error', content: 'error' });
vi.runAllTimers();
});
expect(result.current.errorCount).toBe(1);
act(() => {
result.current.log('A message');
result.current.clearErrorCount();
});
unmount();
expect(clearTimeoutSpy).toHaveBeenCalled();
expect(result.current.errorCount).toBe(0);
});
});
+157 -200
View File
@@ -4,13 +4,7 @@
* SPDX-License-Identifier: Apache-2.0
*/
import {
useCallback,
useEffect,
useReducer,
useRef,
startTransition,
} from 'react';
import { useCallback, useSyncExternalStore } from 'react';
import type { ConsoleMessageItem } from '../types.js';
import {
coreEvents,
@@ -18,207 +12,170 @@ import {
type ConsoleLogPayload,
} from '@google/gemini-cli-core';
export interface UseConsoleMessagesReturn {
consoleMessages: ConsoleMessageItem[];
clearConsoleMessages: () => void;
}
type Action =
| { type: 'ADD_MESSAGES'; payload: ConsoleMessageItem[] }
| { type: 'CLEAR' };
function consoleMessagesReducer(
state: ConsoleMessageItem[],
action: Action,
): ConsoleMessageItem[] {
const MAX_CONSOLE_MESSAGES = 1000;
switch (action.type) {
case 'ADD_MESSAGES': {
const newMessages = [...state];
for (const queuedMessage of action.payload) {
const lastMessage = newMessages[newMessages.length - 1];
if (
lastMessage &&
lastMessage.type === queuedMessage.type &&
lastMessage.content === queuedMessage.content
) {
// Create a new object for the last message to ensure React detects
// the change, preventing mutation of the existing state object.
newMessages[newMessages.length - 1] = {
...lastMessage,
count: lastMessage.count + 1,
};
} else {
newMessages.push({ ...queuedMessage, count: 1 });
}
}
// Limit the number of messages to prevent memory issues
if (newMessages.length > MAX_CONSOLE_MESSAGES) {
return newMessages.slice(newMessages.length - MAX_CONSOLE_MESSAGES);
}
return newMessages;
}
case 'CLEAR':
return [];
default:
return state;
}
}
export function useConsoleMessages(): UseConsoleMessagesReturn {
const [consoleMessages, dispatch] = useReducer(consoleMessagesReducer, []);
const messageQueueRef = useRef<ConsoleMessageItem[]>([]);
const timeoutRef = useRef<NodeJS.Timeout | null>(null);
const isProcessingRef = useRef(false);
const processQueue = useCallback(() => {
if (messageQueueRef.current.length > 0) {
isProcessingRef.current = true;
const messagesToProcess = messageQueueRef.current;
messageQueueRef.current = [];
startTransition(() => {
dispatch({ type: 'ADD_MESSAGES', payload: messagesToProcess });
});
}
timeoutRef.current = null;
}, []);
const handleNewMessage = useCallback(
(message: ConsoleMessageItem) => {
messageQueueRef.current.push(message);
if (!isProcessingRef.current && !timeoutRef.current) {
// Batch updates using a timeout. 50ms is a reasonable delay to batch
// rapid-fire messages without noticeable lag while avoiding React update
// queue flooding.
timeoutRef.current = setTimeout(processQueue, 50);
}
},
[processQueue],
);
// Once the updated consoleMessages have been committed to the screen,
// we can safely process the next batch of queued messages if any exist.
// This completely eliminates overlapping concurrent updates to this state.
useEffect(() => {
isProcessingRef.current = false;
if (messageQueueRef.current.length > 0 && !timeoutRef.current) {
timeoutRef.current = setTimeout(processQueue, 50);
}
}, [consoleMessages, processQueue]);
useEffect(() => {
const handleConsoleLog = (payload: ConsoleLogPayload) => {
let content = payload.content;
const MAX_CONSOLE_MSG_LENGTH = 10000;
if (content.length > MAX_CONSOLE_MSG_LENGTH) {
content =
content.slice(0, MAX_CONSOLE_MSG_LENGTH) +
`... [Truncated ${content.length - MAX_CONSOLE_MSG_LENGTH} characters]`;
}
handleNewMessage({
type: payload.type,
content,
count: 1,
});
};
const handleOutput = (payload: {
isStderr: boolean;
chunk: Uint8Array | string;
}) => {
let content =
typeof payload.chunk === 'string'
? payload.chunk
: new TextDecoder().decode(payload.chunk);
const MAX_OUTPUT_CHUNK_LENGTH = 10000;
if (content.length > MAX_OUTPUT_CHUNK_LENGTH) {
content =
content.slice(0, MAX_OUTPUT_CHUNK_LENGTH) +
`... [Truncated ${content.length - MAX_OUTPUT_CHUNK_LENGTH} characters]`;
}
// It would be nice if we could show stderr as 'warn' but unfortunately
// we log non warning info to stderr before the app starts so that would
// be misleading.
handleNewMessage({ type: 'log', content, count: 1 });
};
coreEvents.on(CoreEvent.ConsoleLog, handleConsoleLog);
coreEvents.on(CoreEvent.Output, handleOutput);
return () => {
coreEvents.off(CoreEvent.ConsoleLog, handleConsoleLog);
coreEvents.off(CoreEvent.Output, handleOutput);
};
}, [handleNewMessage]);
const clearConsoleMessages = useCallback(() => {
if (timeoutRef.current) {
clearTimeout(timeoutRef.current);
timeoutRef.current = null;
}
messageQueueRef.current = [];
isProcessingRef.current = true;
startTransition(() => {
dispatch({ type: 'CLEAR' });
});
}, []);
// Cleanup on unmount
useEffect(
() => () => {
if (timeoutRef.current) {
clearTimeout(timeoutRef.current);
}
},
[],
);
return { consoleMessages, clearConsoleMessages };
}
export interface UseErrorCountReturn {
errorCount: number;
clearErrorCount: () => void;
}
// --- Global Console Store ---
const MAX_CONSOLE_MESSAGES = 1000;
let globalConsoleMessages: ConsoleMessageItem[] = [];
let globalErrorCount = 0;
const listeners = new Set<() => void>();
let messageQueue: ConsoleMessageItem[] = [];
let timeoutId: NodeJS.Timeout | null = null;
/**
* Initializes the global console store and subscribes to coreEvents.
* Acts as a safe reset function, making it idempotent and useful for test isolation.
* Must be called during application startup.
*/
export function initializeConsoleStore() {
if (timeoutId) {
clearTimeout(timeoutId);
timeoutId = null;
}
messageQueue = [];
globalConsoleMessages = [];
globalErrorCount = 0;
notifyListeners();
// Safely detach first to ensure idempotency and prevent listener leaks
coreEvents.off(CoreEvent.ConsoleLog, handleConsoleLog);
coreEvents.off(CoreEvent.Output, handleOutput);
coreEvents.on(CoreEvent.ConsoleLog, handleConsoleLog);
coreEvents.on(CoreEvent.Output, handleOutput);
}
function notifyListeners() {
for (const listener of listeners) {
listener();
}
}
function processQueue() {
if (messageQueue.length === 0) return;
// Create a new array to trigger React updates
const newMessages = [...globalConsoleMessages];
for (const queuedMessage of messageQueue) {
if (queuedMessage.type === 'error') {
globalErrorCount++;
}
// Coalesce consecutive identical messages
const prev = newMessages[newMessages.length - 1];
if (
prev &&
prev.type === queuedMessage.type &&
prev.content === queuedMessage.content
) {
newMessages[newMessages.length - 1] = {
...prev,
count: prev.count + 1,
};
} else {
newMessages.push({ ...queuedMessage, count: 1 });
}
}
globalConsoleMessages =
newMessages.length > MAX_CONSOLE_MESSAGES
? newMessages.slice(-MAX_CONSOLE_MESSAGES)
: newMessages;
messageQueue = [];
timeoutId = null;
notifyListeners();
}
function handleNewMessage(message: ConsoleMessageItem) {
messageQueue.push(message);
if (!timeoutId) {
// Batch updates using a timeout. 50ms is a reasonable delay to batch
// rapid-fire messages without noticeable lag while avoiding React update
// queue flooding.
timeoutId = setTimeout(processQueue, 50);
}
}
// --- Subscription API for useSyncExternalStore ---
function subscribe(listener: () => void) {
listeners.add(listener);
return () => {
listeners.delete(listener);
};
}
function getConsoleMessagesSnapshot() {
return globalConsoleMessages;
}
function getErrorCountSnapshot() {
return globalErrorCount;
}
// --- Core Event Listeners (Always active at module level) ---
const handleConsoleLog = (payload: ConsoleLogPayload) => {
let content = payload.content;
const MAX_CONSOLE_MSG_LENGTH = 10000;
if (content.length > MAX_CONSOLE_MSG_LENGTH) {
content =
content.slice(0, MAX_CONSOLE_MSG_LENGTH) +
`... [Truncated ${content.length - MAX_CONSOLE_MSG_LENGTH} characters]`;
}
handleNewMessage({
type: payload.type,
content,
count: 1,
});
};
const handleOutput = (payload: {
isStderr: boolean;
chunk: Uint8Array | string;
}) => {
let content =
typeof payload.chunk === 'string'
? payload.chunk
: new TextDecoder().decode(payload.chunk);
const MAX_OUTPUT_CHUNK_LENGTH = 10000;
if (content.length > MAX_OUTPUT_CHUNK_LENGTH) {
content =
content.slice(0, MAX_OUTPUT_CHUNK_LENGTH) +
`... [Truncated ${content.length - MAX_OUTPUT_CHUNK_LENGTH} characters]`;
}
handleNewMessage({ type: 'log', content, count: 1 });
};
/**
* Hook to access the global console message history.
* Decoupled from any component lifecycle to ensure history is preserved even
* when the UI is unmounted.
*/
export function useConsoleMessages(): ConsoleMessageItem[] {
return useSyncExternalStore(subscribe, getConsoleMessagesSnapshot);
}
/**
* Hook to access the global error count.
* Uses the same external store as useConsoleMessages for consistency.
*/
export function useErrorCount(): UseErrorCountReturn {
const [errorCount, dispatch] = useReducer(
(state: number, action: 'INCREMENT' | 'CLEAR') => {
switch (action) {
case 'INCREMENT':
return state + 1;
case 'CLEAR':
return 0;
default:
return state;
}
},
0,
);
useEffect(() => {
const handleConsoleLog = (payload: ConsoleLogPayload) => {
if (payload.type === 'error') {
startTransition(() => {
dispatch('INCREMENT');
});
}
};
coreEvents.on(CoreEvent.ConsoleLog, handleConsoleLog);
return () => {
coreEvents.off(CoreEvent.ConsoleLog, handleConsoleLog);
};
}, []);
const errorCount = useSyncExternalStore(subscribe, getErrorCountSnapshot);
const clearErrorCount = useCallback(() => {
startTransition(() => {
dispatch('CLEAR');
});
globalErrorCount = 0;
notifyListeners();
}, []);
return { errorCount, clearErrorCount };
+38 -8
View File
@@ -26,7 +26,6 @@ import {
debugLogger,
runInDevTraceSpan,
EDIT_TOOL_NAMES,
ASK_USER_TOOL_NAME,
processRestorableToolCalls,
recordToolCallInteractions,
ToolErrorType,
@@ -40,6 +39,7 @@ import {
isBackgroundExecutionData,
Kind,
ACTIVATE_SKILL_TOOL_NAME,
shouldHideToolCall,
} from '@google/gemini-cli-core';
import type {
Config,
@@ -66,7 +66,12 @@ import type {
SlashCommandProcessorResult,
HistoryItemModel,
} from '../types.js';
import { StreamingState, MessageType } from '../types.js';
import {
StreamingState,
MessageType,
mapCoreStatusToDisplayStatus,
ToolCallStatus,
} from '../types.js';
import { isAtCommand, isSlashCommand } from '../utils/commandUtils.js';
import { useShellCommandProcessor } from './shellCommandProcessor.js';
import { handleAtCommand } from './atCommandProcessor.js';
@@ -541,14 +546,39 @@ export const useGeminiStream = (
const anyVisibleInHistory = pushedToolCallIds.size > 0;
const anyVisibleInPending = remainingTools.some((tc) => {
// AskUser tools are rendered by AskUserDialog, not ToolGroupMessage
const isInProgress =
tc.status !== 'success' &&
tc.status !== 'error' &&
tc.status !== 'cancelled';
if (tc.request.name === ASK_USER_TOOL_NAME && isInProgress) {
const displayName = tc.tool?.displayName ?? tc.request.name;
let hasResultDisplay = false;
if (
tc.status === CoreToolCallStatus.Success ||
tc.status === CoreToolCallStatus.Error ||
tc.status === CoreToolCallStatus.Cancelled
) {
hasResultDisplay = !!tc.response?.resultDisplay;
} else if (tc.status === CoreToolCallStatus.Executing) {
hasResultDisplay = !!tc.liveOutput;
}
// AskUser tools and Plan Mode write/edit are handled by this logic
if (
shouldHideToolCall({
displayName,
status: tc.status,
approvalMode: tc.approvalMode,
hasResultDisplay,
parentCallId: tc.request.parentCallId,
})
) {
return false;
}
// ToolGroupMessage explicitly hides Confirming tools because they are
// rendered in the interactive ToolConfirmationQueue instead.
const displayStatus = mapCoreStatusToDisplayStatus(tc.status);
if (displayStatus === ToolCallStatus.Confirming) {
return false;
}
// ToolGroupMessage now shows all non-canceled tools, so they are visible
// in pending and we need to draw the closing border for them.
return true;
@@ -25,7 +25,7 @@ const mockUIState = {
dialogsVisible: false,
streamingState: StreamingState.Idle,
isBackgroundShellListOpen: false,
mainControlsRef: { current: null },
mainControlsRef: vi.fn(),
customDialog: null,
historyManager: { addItem: vi.fn() },
history: [],
+1 -1
View File
@@ -1,6 +1,6 @@
{
"name": "@google/gemini-cli-core",
"version": "0.36.0-nightly.20260317.2f90b4653",
"version": "0.36.0-preview.7",
"description": "Gemini CLI Core",
"license": "Apache-2.0",
"repository": {
@@ -128,7 +128,10 @@ describe('A2AClientManager', () => {
describe('getInstance / dispatcher initialization', () => {
it('should use UndiciAgent when no proxy is configured', async () => {
await manager.loadAgent('TestAgent', 'http://test.agent/card');
await manager.loadAgent('TestAgent', {
type: 'url',
url: 'http://test.agent/card',
});
const resolverOptions = vi.mocked(DefaultAgentCardResolver).mock
.calls[0][0];
@@ -153,7 +156,10 @@ describe('A2AClientManager', () => {
} as Config;
manager = new A2AClientManager(mockConfigWithProxy);
await manager.loadAgent('TestProxyAgent', 'http://test.proxy.agent/card');
await manager.loadAgent('TestProxyAgent', {
type: 'url',
url: 'http://test.proxy.agent/card',
});
const resolverOptions = vi.mocked(DefaultAgentCardResolver).mock
.calls[0][0];
@@ -172,28 +178,40 @@ describe('A2AClientManager', () => {
describe('loadAgent', () => {
it('should create and cache an A2AClient', async () => {
const agentCard = await manager.loadAgent(
'TestAgent',
'http://test.agent/card',
);
const agentCard = await manager.loadAgent('TestAgent', {
type: 'url',
url: 'http://test.agent/card',
});
expect(manager.getAgentCard('TestAgent')).toBe(agentCard);
expect(manager.getClient('TestAgent')).toBeDefined();
});
it('should configure ClientFactory with REST, JSON-RPC, and gRPC transports', async () => {
await manager.loadAgent('TestAgent', 'http://test.agent/card');
await manager.loadAgent('TestAgent', {
type: 'url',
url: 'http://test.agent/card',
});
expect(ClientFactoryOptions.createFrom).toHaveBeenCalled();
});
it('should throw an error if an agent with the same name is already loaded', async () => {
await manager.loadAgent('TestAgent', 'http://test.agent/card');
await manager.loadAgent('TestAgent', {
type: 'url',
url: 'http://test.agent/card',
});
await expect(
manager.loadAgent('TestAgent', 'http://test.agent/card'),
manager.loadAgent('TestAgent', {
type: 'url',
url: 'http://test.agent/card',
}),
).rejects.toThrow("Agent with name 'TestAgent' is already loaded.");
});
it('should use native fetch by default', async () => {
await manager.loadAgent('TestAgent', 'http://test.agent/card');
await manager.loadAgent('TestAgent', {
type: 'url',
url: 'http://test.agent/card',
});
expect(createAuthenticatingFetchWithRetry).not.toHaveBeenCalled();
});
@@ -204,7 +222,7 @@ describe('A2AClientManager', () => {
};
await manager.loadAgent(
'TestAgent',
'http://test.agent/card',
{ type: 'url', url: 'http://test.agent/card' },
customAuthHandler as unknown as AuthenticationHandler,
);
@@ -221,7 +239,7 @@ describe('A2AClientManager', () => {
};
await manager.loadAgent(
'AuthCardAgent',
'http://authcard.agent/card',
{ type: 'url', url: 'http://authcard.agent/card' },
customAuthHandler as unknown as AuthenticationHandler,
);
@@ -252,7 +270,7 @@ describe('A2AClientManager', () => {
await manager.loadAgent(
'AuthCardAgent401',
'http://authcard.agent/card',
{ type: 'url', url: 'http://authcard.agent/card' },
customAuthHandler as unknown as AuthenticationHandler,
);
@@ -267,19 +285,65 @@ describe('A2AClientManager', () => {
});
it('should log a debug message upon loading an agent', async () => {
await manager.loadAgent('TestAgent', 'http://test.agent/card');
await manager.loadAgent('TestAgent', {
type: 'url',
url: 'http://test.agent/card',
});
expect(debugLogger.debug).toHaveBeenCalledWith(
expect.stringContaining("Loaded agent 'TestAgent'"),
);
});
it('should clear the cache', async () => {
await manager.loadAgent('TestAgent', 'http://test.agent/card');
await manager.loadAgent('TestAgent', {
type: 'url',
url: 'http://test.agent/card',
});
manager.clearCache();
expect(manager.getAgentCard('TestAgent')).toBeUndefined();
expect(manager.getClient('TestAgent')).toBeUndefined();
});
it('should load an agent from inline JSON without calling resolver', async () => {
const inlineJson = JSON.stringify(mockAgentCard);
const agentCard = await manager.loadAgent('JsonAgent', {
type: 'json',
json: inlineJson,
});
expect(agentCard).toBeDefined();
expect(agentCard.name).toBe('test-agent');
expect(manager.getAgentCard('JsonAgent')).toBe(agentCard);
expect(manager.getClient('JsonAgent')).toBeDefined();
// Resolver should not have been called for inline JSON
const resolverInstance = vi.mocked(DefaultAgentCardResolver).mock
.results[0]?.value;
if (resolverInstance) {
expect(resolverInstance.resolve).not.toHaveBeenCalled();
}
});
it('should throw a descriptive error for invalid inline JSON', async () => {
await expect(
manager.loadAgent('BadJsonAgent', {
type: 'json',
json: 'not valid json {{',
}),
).rejects.toThrow(
/Failed to parse inline agent card JSON for agent 'BadJsonAgent'/,
);
});
it('should log "inline JSON" for JSON-loaded agents', async () => {
const inlineJson = JSON.stringify(mockAgentCard);
await manager.loadAgent('JsonLogAgent', {
type: 'json',
json: inlineJson,
});
expect(debugLogger.debug).toHaveBeenCalledWith(
expect.stringContaining('inline JSON'),
);
});
it('should throw if resolveAgentCard fails', async () => {
const resolverInstance = {
resolve: vi.fn().mockRejectedValue(new Error('Resolution failed')),
@@ -289,7 +353,10 @@ describe('A2AClientManager', () => {
);
await expect(
manager.loadAgent('FailAgent', 'http://fail.agent'),
manager.loadAgent('FailAgent', {
type: 'url',
url: 'http://fail.agent',
}),
).rejects.toThrow('Resolution failed');
});
@@ -304,7 +371,10 @@ describe('A2AClientManager', () => {
);
await expect(
manager.loadAgent('FailAgent', 'http://fail.agent'),
manager.loadAgent('FailAgent', {
type: 'url',
url: 'http://fail.agent',
}),
).rejects.toThrow('Factory failed');
});
});
@@ -318,7 +388,10 @@ describe('A2AClientManager', () => {
describe('sendMessageStream', () => {
beforeEach(async () => {
await manager.loadAgent('TestAgent', 'http://test.agent/card');
await manager.loadAgent('TestAgent', {
type: 'url',
url: 'http://test.agent/card',
});
});
it('should send a message and return a stream', async () => {
@@ -433,7 +506,10 @@ describe('A2AClientManager', () => {
describe('getTask', () => {
beforeEach(async () => {
await manager.loadAgent('TestAgent', 'http://test.agent/card');
await manager.loadAgent('TestAgent', {
type: 'url',
url: 'http://test.agent/card',
});
});
it('should get a task from the correct agent', async () => {
@@ -462,7 +538,10 @@ describe('A2AClientManager', () => {
describe('cancelTask', () => {
beforeEach(async () => {
await manager.loadAgent('TestAgent', 'http://test.agent/card');
await manager.loadAgent('TestAgent', {
type: 'url',
url: 'http://test.agent/card',
});
});
it('should cancel a task on the correct agent', async () => {
+22 -4
View File
@@ -26,6 +26,7 @@ import * as grpc from '@grpc/grpc-js';
import { v4 as uuidv4 } from 'uuid';
import { Agent as UndiciAgent, ProxyAgent } from 'undici';
import { normalizeAgentCard } from './a2aUtils.js';
import type { AgentCardLoadOptions } from './types.js';
import type { Config } from '../config/config.js';
import { debugLogger } from '../utils/debugLogger.js';
import { classifyAgentError } from './a2a-errors.js';
@@ -85,7 +86,7 @@ export class A2AClientManager {
*/
async loadAgent(
name: string,
agentCardUrl: string,
options: AgentCardLoadOptions,
authHandler?: AuthenticationHandler,
): Promise<AgentCard> {
if (this.clients.has(name) && this.agentCards.has(name)) {
@@ -119,7 +120,24 @@ export class A2AClientManager {
};
const resolver = new DefaultAgentCardResolver({ fetchImpl: cardFetch });
const rawCard = await resolver.resolve(agentCardUrl, '');
let rawCard: unknown;
let urlIdentifier = 'inline JSON';
if (options.type === 'json') {
try {
rawCard = JSON.parse(options.json);
} catch (error) {
const msg = error instanceof Error ? error.message : String(error);
throw new Error(
`Failed to parse inline agent card JSON for agent '${name}': ${msg}`,
);
}
} else {
urlIdentifier = options.url;
rawCard = await resolver.resolve(options.url, '');
}
// TODO: Remove normalizeAgentCard once @a2a-js/sdk handles
// proto field name aliases (supportedInterfaces → additionalInterfaces,
// protocolBinding → transport).
@@ -153,12 +171,12 @@ export class A2AClientManager {
this.agentCards.set(name, agentCard);
debugLogger.debug(
`[A2AClientManager] Loaded agent '${name}' from ${agentCardUrl}`,
`[A2AClientManager] Loaded agent '${name}' from ${urlIdentifier}`,
);
return agentCard;
} catch (error: unknown) {
throw classifyAgentError(name, agentCardUrl, error);
throw classifyAgentError(name, urlIdentifier, error);
}
}
@@ -19,6 +19,9 @@ import {
DEFAULT_MAX_TIME_MINUTES,
DEFAULT_MAX_TURNS,
type LocalAgentDefinition,
type RemoteAgentDefinition,
getAgentCardLoadOptions,
getRemoteAgentTargetUrl,
} from './types.js';
describe('loader', () => {
@@ -232,6 +235,75 @@ agent_card_url: https://example.com/card
});
});
it('should parse a remote agent with agent_card_json', async () => {
const cardJson = JSON.stringify({
name: 'json-agent',
url: 'https://example.com/agent',
version: '1.0',
});
const filePath = await writeAgentMarkdown(`---
kind: remote
name: json-remote
description: A JSON-based remote agent
agent_card_json: '${cardJson}'
---
`);
const result = await parseAgentMarkdown(filePath);
expect(result).toHaveLength(1);
expect(result[0]).toMatchObject({
kind: 'remote',
name: 'json-remote',
description: 'A JSON-based remote agent',
agent_card_json: cardJson,
});
// Should NOT have agent_card_url
expect(result[0]).not.toHaveProperty('agent_card_url');
});
it('should reject agent_card_json that is not valid JSON', async () => {
const filePath = await writeAgentMarkdown(`---
kind: remote
name: invalid-json-remote
agent_card_json: "not valid json {{"
---
`);
await expect(parseAgentMarkdown(filePath)).rejects.toThrow(
/agent_card_json must be valid JSON/,
);
});
it('should reject a remote agent with both agent_card_url and agent_card_json', async () => {
const filePath = await writeAgentMarkdown(`---
kind: remote
name: both-fields
agent_card_url: https://example.com/card
agent_card_json: '{"name":"test"}'
---
`);
await expect(parseAgentMarkdown(filePath)).rejects.toThrow(
/Validation failed/,
);
});
it('should infer remote kind from agent_card_json', async () => {
const cardJson = JSON.stringify({
name: 'test',
url: 'https://example.com',
});
const filePath = await writeAgentMarkdown(`---
name: inferred-json-remote
agent_card_json: '${cardJson}'
---
`);
const result = await parseAgentMarkdown(filePath);
expect(result).toHaveLength(1);
expect(result[0]).toMatchObject({
kind: 'remote',
name: 'inferred-json-remote',
agent_card_json: cardJson,
});
});
it('should throw AgentLoadError if agent name is not a valid slug', async () => {
const filePath = await writeAgentMarkdown(`---
name: Invalid Name With Spaces
@@ -242,6 +314,99 @@ Body`);
/Name must be a valid slug/,
);
});
describe('error formatting and kind inference', () => {
it('should only show local agent errors when kind is inferred as local (via kind field)', async () => {
const filePath = await writeAgentMarkdown(`---
kind: local
name: invalid-local
# missing description
---
Body`);
const error = await parseAgentMarkdown(filePath).catch((e) => e);
expect(error).toBeInstanceOf(AgentLoadError);
expect(error.message).toContain('Validation failed');
expect(error.message).toContain('description: Required');
expect(error.message).not.toContain('Remote Agent');
});
it('should only show local agent errors when kind is inferred as local (via local-specific keys)', async () => {
const filePath = await writeAgentMarkdown(`---
name: invalid-local
# missing description
tools:
- run_shell_command
---
Body`);
const error = await parseAgentMarkdown(filePath).catch((e) => e);
expect(error).toBeInstanceOf(AgentLoadError);
expect(error.message).toContain('Validation failed');
expect(error.message).toContain('description: Required');
expect(error.message).not.toContain('Remote Agent');
});
it('should only show remote agent errors when kind is inferred as remote (via kind field)', async () => {
const filePath = await writeAgentMarkdown(`---
kind: remote
name: invalid-remote
# missing agent_card_url
---
Body`);
const error = await parseAgentMarkdown(filePath).catch((e) => e);
expect(error).toBeInstanceOf(AgentLoadError);
expect(error.message).toContain('Validation failed');
expect(error.message).toContain('agent_card_url: Required');
expect(error.message).not.toContain('Local Agent');
});
it('should only show remote agent errors when kind is inferred as remote (via remote-specific keys)', async () => {
const filePath = await writeAgentMarkdown(`---
name: invalid-remote
auth:
type: apiKey
key: my_key
# missing agent_card_url
---
Body`);
const error = await parseAgentMarkdown(filePath).catch((e) => e);
expect(error).toBeInstanceOf(AgentLoadError);
expect(error.message).toContain('Validation failed');
expect(error.message).toContain('agent_card_url: Required');
expect(error.message).not.toContain('Local Agent');
});
it('should show errors for both types when kind cannot be inferred', async () => {
const filePath = await writeAgentMarkdown(`---
name: invalid-unknown
# missing description and missing agent_card_url, no specific keys
---
Body`);
const error = await parseAgentMarkdown(filePath).catch((e) => e);
expect(error).toBeInstanceOf(AgentLoadError);
expect(error.message).toContain('Validation failed');
expect(error.message).toContain('(Local Agent)');
expect(error.message).toContain('(Remote Agent)');
expect(error.message).toContain('description: Required');
expect(error.message).toContain('agent_card_url: Required');
});
it('should format errors without a stray colon when the path is empty (e.g. strict object with unknown keys)', async () => {
const filePath = await writeAgentMarkdown(`---
kind: local
name: my-agent
description: test
unknown_field: true
---
Body`);
const error = await parseAgentMarkdown(filePath).catch((e) => e);
expect(error).toBeInstanceOf(AgentLoadError);
expect(error.message).toContain(
"Unrecognized key(s) in object: 'unknown_field'",
);
expect(error.message).not.toContain(': Unrecognized key(s)');
expect(error.message).not.toContain('Required');
});
});
});
describe('markdownToAgentDefinition', () => {
@@ -372,6 +537,40 @@ Body`);
},
});
});
it('should convert remote agent definition with agent_card_json', () => {
const cardJson = JSON.stringify({
name: 'json-agent',
url: 'https://example.com/agent',
});
const markdown = {
kind: 'remote' as const,
name: 'json-remote',
description: 'A JSON remote agent',
agent_card_json: cardJson,
};
const result = markdownToAgentDefinition(
markdown,
) as RemoteAgentDefinition;
expect(result.kind).toBe('remote');
expect(result.name).toBe('json-remote');
expect(result.agentCardJson).toBe(cardJson);
expect(result.agentCardUrl).toBeUndefined();
});
it('should throw for remote agent with neither agent_card_url nor agent_card_json', () => {
// Cast to bypass compile-time check — this tests the runtime guard
const markdown = {
kind: 'remote' as const,
name: 'no-card-agent',
description: 'Missing card info',
} as Parameters<typeof markdownToAgentDefinition>[0];
expect(() => markdownToAgentDefinition(markdown)).toThrow(
/neither agent_card_json nor agent_card_url/,
);
});
});
describe('loadAgentsFromDirectory', () => {
@@ -744,5 +943,103 @@ auth:
},
});
});
it('should throw an error for an unknown auth type in markdownToAgentDefinition', () => {
const markdown = {
kind: 'remote' as const,
name: 'unknown-auth-agent',
agent_card_url: 'https://example.com/card',
auth: {
type: 'apiKey' as const,
key: 'some-key',
},
};
// Mutate the object at runtime to bypass TypeScript compile-time checks cleanly
Object.assign(markdown.auth, { type: 'some-unknown-type' });
expect(() => markdownToAgentDefinition(markdown)).toThrow(
/Unknown auth type: some-unknown-type/,
);
});
});
describe('getAgentCardLoadOptions', () => {
it('should return json options when agentCardJson is present', () => {
const def = {
name: 'test',
agentCardJson: '{"url":"http://x"}',
} as RemoteAgentDefinition;
const opts = getAgentCardLoadOptions(def);
expect(opts).toEqual({ type: 'json', json: '{"url":"http://x"}' });
});
it('should return url options when agentCardUrl is present', () => {
const def = {
name: 'test',
agentCardUrl: 'http://x/card',
} as RemoteAgentDefinition;
const opts = getAgentCardLoadOptions(def);
expect(opts).toEqual({ type: 'url', url: 'http://x/card' });
});
it('should prefer agentCardJson over agentCardUrl when both present', () => {
const def = {
name: 'test',
agentCardJson: '{"url":"http://x"}',
agentCardUrl: 'http://x/card',
} as RemoteAgentDefinition;
const opts = getAgentCardLoadOptions(def);
expect(opts.type).toBe('json');
});
it('should throw when neither is present', () => {
const def = { name: 'orphan' } as RemoteAgentDefinition;
expect(() => getAgentCardLoadOptions(def)).toThrow(
/Remote agent 'orphan' has neither agentCardUrl nor agentCardJson/,
);
});
});
describe('getRemoteAgentTargetUrl', () => {
it('should return agentCardUrl when present', () => {
const def = {
name: 'test',
agentCardUrl: 'http://x/card',
} as RemoteAgentDefinition;
expect(getRemoteAgentTargetUrl(def)).toBe('http://x/card');
});
it('should extract url from agentCardJson when agentCardUrl is absent', () => {
const def = {
name: 'test',
agentCardJson: JSON.stringify({
name: 'agent',
url: 'https://example.com/agent',
}),
} as RemoteAgentDefinition;
expect(getRemoteAgentTargetUrl(def)).toBe('https://example.com/agent');
});
it('should return undefined when JSON has no url field', () => {
const def = {
name: 'test',
agentCardJson: JSON.stringify({ name: 'agent' }),
} as RemoteAgentDefinition;
expect(getRemoteAgentTargetUrl(def)).toBeUndefined();
});
it('should return undefined when agentCardJson is invalid JSON', () => {
const def = {
name: 'test',
agentCardJson: 'not json',
} as RemoteAgentDefinition;
expect(getRemoteAgentTargetUrl(def)).toBeUndefined();
});
it('should return undefined when neither field is present', () => {
const def = { name: 'test' } as RemoteAgentDefinition;
expect(getRemoteAgentTargetUrl(def)).toBeUndefined();
});
});
});
+165 -194
View File
@@ -12,6 +12,7 @@ import * as crypto from 'node:crypto';
import { z } from 'zod';
import {
type AgentDefinition,
type RemoteAgentDefinition,
DEFAULT_MAX_TURNS,
DEFAULT_MAX_TIME_MINUTES,
} from './types.js';
@@ -21,79 +22,6 @@ import { isValidToolName } from '../tools/tool-names.js';
import { FRONTMATTER_REGEX } from '../skills/skillLoader.js';
import { getErrorMessage } from '../utils/errors.js';
/**
* DTO for Markdown parsing - represents the structure from frontmatter.
*/
interface FrontmatterBaseAgentDefinition {
name: string;
display_name?: string;
}
interface FrontmatterMCPServerConfig {
command?: string;
args?: string[];
env?: Record<string, string>;
cwd?: string;
url?: string;
http_url?: string;
headers?: Record<string, string>;
tcp?: string;
type?: 'sse' | 'http';
timeout?: number;
trust?: boolean;
description?: string;
include_tools?: string[];
exclude_tools?: string[];
}
interface FrontmatterLocalAgentDefinition
extends FrontmatterBaseAgentDefinition {
kind: 'local';
description: string;
tools?: string[];
mcp_servers?: Record<string, FrontmatterMCPServerConfig>;
system_prompt: string;
model?: string;
temperature?: number;
max_turns?: number;
timeout_mins?: number;
}
/**
* Authentication configuration for remote agents in frontmatter format.
*/
interface FrontmatterAuthConfig {
type: 'apiKey' | 'http' | 'google-credentials' | 'oauth';
// API Key
key?: string;
name?: string;
// HTTP
scheme?: string;
token?: string;
username?: string;
password?: string;
value?: string;
// Google Credentials
scopes?: string[];
// OAuth2
client_id?: string;
client_secret?: string;
authorization_url?: string;
token_url?: string;
}
interface FrontmatterRemoteAgentDefinition
extends FrontmatterBaseAgentDefinition {
kind: 'remote';
description?: string;
agent_card_url: string;
auth?: FrontmatterAuthConfig;
}
type FrontmatterAgentDefinition =
| FrontmatterLocalAgentDefinition
| FrontmatterRemoteAgentDefinition;
/**
* Error thrown when an agent definition is invalid or cannot be loaded.
*/
@@ -159,15 +87,13 @@ const localAgentSchema = z
})
.strict();
/**
* Base fields shared by all auth configs.
*/
type FrontmatterLocalAgentDefinition = z.infer<typeof localAgentSchema> & {
system_prompt: string;
};
// Base fields shared by all auth configs.
const baseAuthFields = {};
/**
* API Key auth schema.
* Supports sending key in header, query parameter, or cookie.
*/
const apiKeyAuthSchema = z.object({
...baseAuthFields,
type: z.literal('apiKey'),
@@ -175,11 +101,6 @@ const apiKeyAuthSchema = z.object({
name: z.string().optional(),
});
/**
* HTTP auth schema (Bearer or Basic).
* Note: Validation for scheme-specific fields is applied in authConfigSchema
* since discriminatedUnion doesn't support refined schemas directly.
*/
const httpAuthSchema = z.object({
...baseAuthFields,
type: z.literal('http'),
@@ -190,19 +111,12 @@ const httpAuthSchema = z.object({
value: z.string().min(1).optional(),
});
/**
* Google Credentials auth schema.
*/
const googleCredentialsAuthSchema = z.object({
...baseAuthFields,
type: z.literal('google-credentials'),
scopes: z.array(z.string()).optional(),
});
/**
* OAuth2 auth schema.
* authorization_url and token_url can be discovered from the agent card if omitted.
*/
const oauth2AuthSchema = z.object({
...baseAuthFields,
type: z.literal('oauth'),
@@ -222,18 +136,16 @@ const authConfigSchema = z
])
.superRefine((data, ctx) => {
if (data.type === 'http') {
if (data.value) {
// Raw mode - only scheme and value are needed
return;
}
if (data.scheme === 'Bearer' && !data.token) {
ctx.addIssue({
code: z.ZodIssueCode.custom,
message: 'Bearer scheme requires "token"',
path: ['token'],
});
}
if (data.scheme === 'Basic') {
if (data.value) return;
if (data.scheme === 'Bearer') {
if (!data.token) {
ctx.addIssue({
code: z.ZodIssueCode.custom,
message: 'Bearer scheme requires "token"',
path: ['token'],
});
}
} else if (data.scheme === 'Basic') {
if (!data.username) {
ctx.addIssue({
code: z.ZodIssueCode.custom,
@@ -248,55 +160,127 @@ const authConfigSchema = z
path: ['password'],
});
}
} else {
ctx.addIssue({
code: z.ZodIssueCode.custom,
message: `HTTP scheme "${data.scheme}" requires "value"`,
path: ['value'],
});
}
}
});
const remoteAgentSchema = z
.object({
kind: z.literal('remote').optional().default('remote'),
name: nameSchema,
description: z.string().optional(),
display_name: z.string().optional(),
type FrontmatterAuthConfig = z.infer<typeof authConfigSchema>;
const baseRemoteAgentSchema = z.object({
kind: z.literal('remote').optional().default('remote'),
name: nameSchema,
description: z.string().optional(),
display_name: z.string().optional(),
auth: authConfigSchema.optional(),
});
const remoteAgentUrlSchema = baseRemoteAgentSchema
.extend({
agent_card_url: z.string().url(),
auth: authConfigSchema.optional(),
agent_card_json: z.undefined().optional(),
})
.strict();
// Use a Zod union to automatically discriminate between local and remote
// agent types.
const remoteAgentJsonSchema = baseRemoteAgentSchema
.extend({
agent_card_url: z.undefined().optional(),
agent_card_json: z.string().refine(
(val) => {
try {
JSON.parse(val);
return true;
} catch {
return false;
}
},
{ message: 'agent_card_json must be valid JSON' },
),
})
.strict();
const remoteAgentSchema = z.union([
remoteAgentUrlSchema,
remoteAgentJsonSchema,
]);
type FrontmatterRemoteAgentDefinition = z.infer<typeof remoteAgentSchema>;
type FrontmatterAgentDefinition =
| FrontmatterLocalAgentDefinition
| FrontmatterRemoteAgentDefinition;
const agentUnionOptions = [
{ schema: localAgentSchema, label: 'Local Agent' },
{ schema: remoteAgentSchema, label: 'Remote Agent' },
] as const;
{ label: 'Local Agent' },
{ label: 'Remote Agent' },
{ label: 'Remote Agent' },
];
const remoteAgentsListSchema = z.array(remoteAgentSchema);
const markdownFrontmatterSchema = z.union([
agentUnionOptions[0].schema,
agentUnionOptions[1].schema,
localAgentSchema,
remoteAgentUrlSchema,
remoteAgentJsonSchema,
]);
function formatZodError(error: z.ZodError, context: string): string {
const issues = error.issues
.map((i) => {
// Handle union errors specifically to give better context
function guessIntendedKind(rawInput: unknown): 'local' | 'remote' | undefined {
if (typeof rawInput !== 'object' || rawInput === null) return undefined;
const input = rawInput as Partial<FrontmatterLocalAgentDefinition> &
Partial<FrontmatterRemoteAgentDefinition>;
if (input.kind === 'local') return 'local';
if (input.kind === 'remote') return 'remote';
const hasLocalKeys =
'tools' in input ||
'mcp_servers' in input ||
'model' in input ||
'temperature' in input ||
'max_turns' in input ||
'timeout_mins' in input;
const hasRemoteKeys =
'agent_card_url' in input || 'auth' in input || 'agent_card_json' in input;
if (hasLocalKeys && !hasRemoteKeys) return 'local';
if (hasRemoteKeys && !hasLocalKeys) return 'remote';
return undefined;
}
function formatZodError(
error: z.ZodError,
context: string,
rawInput?: unknown,
): string {
const intendedKind = rawInput ? guessIntendedKind(rawInput) : undefined;
const formatIssues = (issues: z.ZodIssue[], unionPrefix?: string): string[] =>
issues.flatMap((i) => {
if (i.code === z.ZodIssueCode.invalid_union) {
return i.unionErrors
.map((unionError, index) => {
const label =
agentUnionOptions[index]?.label ?? `Agent type #${index + 1}`;
const unionIssues = unionError.issues
.map((u) => `${u.path.join('.')}: ${u.message}`)
.join(', ');
return `(${label}) ${unionIssues}`;
})
.join('\n');
return i.unionErrors.flatMap((unionError, index) => {
const label = unionPrefix
? unionPrefix
: ((agentUnionOptions[index] as { label?: string })?.label ??
`Branch #${index + 1}`);
if (intendedKind === 'local' && label === 'Remote Agent') return [];
if (intendedKind === 'remote' && label === 'Local Agent') return [];
return formatIssues(unionError.issues, label);
});
}
return `${i.path.join('.')}: ${i.message}`;
})
.join('\n');
return `${context}:\n${issues}`;
const prefix = unionPrefix ? `(${unionPrefix}) ` : '';
const path = i.path.length > 0 ? `${i.path.join('.')}: ` : '';
return `${prefix}${path}${i.message}`;
});
const formatted = Array.from(new Set(formatIssues(error.issues))).join('\n');
return `${context}:\n${formatted}`;
}
/**
@@ -343,8 +327,7 @@ export async function parseAgentMarkdown(
} catch (error) {
throw new AgentLoadError(
filePath,
// eslint-disable-next-line @typescript-eslint/no-unsafe-type-assertion
`YAML frontmatter parsing failed: ${(error as Error).message}`,
`YAML frontmatter parsing failed: ${getErrorMessage(error)}`,
);
}
@@ -368,7 +351,7 @@ export async function parseAgentMarkdown(
if (!result.success) {
throw new AgentLoadError(
filePath,
`Validation failed: ${formatZodError(result.error, 'Agent Definition')}`,
`Validation failed: ${formatZodError(result.error, 'Agent Definition', rawFrontmatter)}`,
);
}
@@ -383,17 +366,14 @@ export async function parseAgentMarkdown(
];
}
// Local agent validation
// Validate tools
// Construct the local agent definition
const agentDef: FrontmatterLocalAgentDefinition = {
...frontmatter,
kind: 'local',
system_prompt: body.trim(),
};
return [agentDef];
return [
{
...frontmatter,
kind: 'local',
system_prompt: body.trim(),
},
];
}
/**
@@ -403,15 +383,9 @@ export async function parseAgentMarkdown(
function convertFrontmatterAuthToConfig(
frontmatter: FrontmatterAuthConfig,
): A2AAuthConfig {
const base = {};
switch (frontmatter.type) {
case 'apiKey':
if (!frontmatter.key) {
throw new Error('Internal error: API key missing after validation.');
}
return {
...base,
type: 'apiKey',
key: frontmatter.key,
name: frontmatter.name,
@@ -419,20 +393,13 @@ function convertFrontmatterAuthToConfig(
case 'google-credentials':
return {
...base,
type: 'google-credentials',
scopes: frontmatter.scopes,
};
case 'http': {
if (!frontmatter.scheme) {
throw new Error(
'Internal error: HTTP scheme missing after validation.',
);
}
case 'http':
if (frontmatter.value) {
return {
...base,
type: 'http',
scheme: frontmatter.scheme,
value: frontmatter.value,
@@ -440,40 +407,27 @@ function convertFrontmatterAuthToConfig(
}
switch (frontmatter.scheme) {
case 'Bearer':
if (!frontmatter.token) {
throw new Error(
'Internal error: Bearer token missing after validation.',
);
}
// Token is required by schema validation
return {
...base,
type: 'http',
scheme: 'Bearer',
token: frontmatter.token,
token: frontmatter.token!,
};
case 'Basic':
if (!frontmatter.username || !frontmatter.password) {
throw new Error(
'Internal error: Basic auth credentials missing after validation.',
);
}
// Username/password are required by schema validation
return {
...base,
type: 'http',
scheme: 'Basic',
username: frontmatter.username,
password: frontmatter.password,
username: frontmatter.username!,
password: frontmatter.password!,
};
default: {
// Other IANA schemes without a value should not reach here after validation
default:
throw new Error(`Unknown HTTP scheme: ${frontmatter.scheme}`);
}
}
}
case 'oauth':
return {
...base,
type: 'oauth2',
client_id: frontmatter.client_id,
client_secret: frontmatter.client_secret,
@@ -483,8 +437,12 @@ function convertFrontmatterAuthToConfig(
};
default: {
const exhaustive: never = frontmatter.type;
throw new Error(`Unknown auth type: ${exhaustive}`);
const exhaustive: never = frontmatter;
const raw: unknown = exhaustive;
if (typeof raw === 'object' && raw !== null && 'type' in raw) {
throw new Error(`Unknown auth type: ${String(raw['type'])}`);
}
throw new Error('Unknown auth type');
}
}
}
@@ -515,25 +473,41 @@ export function markdownToAgentDefinition(
};
if (markdown.kind === 'remote') {
return {
const base: RemoteAgentDefinition = {
kind: 'remote',
name: markdown.name,
description: markdown.description || '',
displayName: markdown.display_name,
agentCardUrl: markdown.agent_card_url,
auth: markdown.auth
? convertFrontmatterAuthToConfig(markdown.auth)
: undefined,
inputConfig,
metadata,
};
if (
'agent_card_json' in markdown &&
markdown.agent_card_json !== undefined
) {
base.agentCardJson = markdown.agent_card_json;
return base;
}
if ('agent_card_url' in markdown && markdown.agent_card_url !== undefined) {
base.agentCardUrl = markdown.agent_card_url;
return base;
}
throw new AgentLoadError(
metadata?.filePath || 'unknown',
'Unexpected state: neither agent_card_json nor agent_card_url present on remote agent',
);
}
// If a model is specified, use it. Otherwise, inherit
const modelName = markdown.model || 'inherit';
const mcpServers: Record<string, MCPServerConfig> = {};
if (markdown.kind === 'local' && markdown.mcp_servers) {
if (markdown.mcp_servers) {
for (const [name, config] of Object.entries(markdown.mcp_servers)) {
mcpServers[name] = new MCPServerConfig(
config.command,
@@ -606,15 +580,13 @@ export async function loadAgentsFromDirectory(
dirEntries = await fs.readdir(dir, { withFileTypes: true });
} catch (error) {
// If directory doesn't exist, just return empty
// eslint-disable-next-line @typescript-eslint/no-unsafe-type-assertion
if ((error as NodeJS.ErrnoException).code === 'ENOENT') {
if (error instanceof Error && 'code' in error && error.code === 'ENOENT') {
return result;
}
result.errors.push(
new AgentLoadError(
dir,
// eslint-disable-next-line @typescript-eslint/no-unsafe-type-assertion
`Could not list directory: ${(error as Error).message}`,
`Could not list directory: ${getErrorMessage(error)}`,
),
);
return result;
@@ -644,8 +616,7 @@ export async function loadAgentsFromDirectory(
result.errors.push(
new AgentLoadError(
filePath,
// eslint-disable-next-line @typescript-eslint/no-unsafe-type-assertion
`Unexpected error: ${(error as Error).message}`,
`Unexpected error: ${getErrorMessage(error)}`,
),
);
}
@@ -48,6 +48,7 @@ import { debugLogger } from '../utils/debugLogger.js';
import { LocalAgentExecutor, type ActivityCallback } from './local-executor.js';
import { makeFakeConfig } from '../test-utils/config.js';
import { ToolRegistry } from '../tools/tool-registry.js';
import type { MessageBus } from '../confirmation-bus/message-bus.js';
import { PromptRegistry } from '../prompts/prompt-registry.js';
import { ResourceRegistry } from '../resources/resource-registry.js';
import { DiscoveredMCPTool } from '../tools/mcp-tool.js';
@@ -344,10 +345,9 @@ describe('LocalAgentExecutor', () => {
get: () => 'test-prompt-id',
configurable: true,
});
parentToolRegistry = new ToolRegistry(mockConfig, mockConfig.messageBus);
parentToolRegistry.registerTool(
new LSTool(mockConfig, mockConfig.messageBus),
);
const { messageBus } = mockConfig as unknown as { messageBus: MessageBus };
parentToolRegistry = new ToolRegistry(mockConfig, messageBus);
parentToolRegistry.registerTool(new LSTool(mockConfig, messageBus));
parentToolRegistry.registerTool(
new MockTool({ name: READ_FILE_TOOL_NAME }),
);
@@ -687,6 +687,25 @@ describe('LocalAgentExecutor', () => {
// Assert that there is exactly ONE schema for this tool
expect(foundSchemas).toHaveLength(1);
});
it('should provide tools to the model when toolConfig is OMITTED (default to all tools)', async () => {
const fullDefinition = createTestDefinition();
const { toolConfig: _, ...definition } = fullDefinition;
const executor = await LocalAgentExecutor.create(
definition as LocalAgentDefinition,
mockConfig,
onActivity,
);
const toolsList = (
executor as unknown as { prepareToolsList: () => FunctionDeclaration[] }
).prepareToolsList();
// Verify that LS_TOOL_NAME is in the list (since LS was registered in beforeEach)
const toolNames = toolsList.map((t) => t.name);
expect(toolNames).toContain(LS_TOOL_NAME);
});
});
describe('run (Execution Loop and Logic)', () => {
+6 -2
View File
@@ -1335,9 +1335,13 @@ export class LocalAgentExecutor<TOutput extends z.ZodTypeAny> {
toolsList.push(toolRef);
}
}
// Add schemas from tools that were explicitly registered by name, wildcard, or instance.
toolsList.push(...this.toolRegistry.getFunctionDeclarations());
}
// Add schemas from tools that were explicitly registered by name, wildcard, or instance.
toolsList.push(
...this.toolRegistry.getFunctionDeclarations(
this.definition.modelConfig.model,
),
);
// Always inject complete_task.
// Configure its schema based on whether output is expected.
+1 -1
View File
@@ -596,7 +596,7 @@ describe('AgentRegistry', () => {
});
expect(loadAgentSpy).toHaveBeenCalledWith(
'RemoteAgentWithAuth',
'https://example.com/card',
{ type: 'url', url: 'https://example.com/card' },
mockHandler,
);
expect(registry.getDefinition('RemoteAgentWithAuth')).toEqual(
+14 -4
View File
@@ -4,10 +4,12 @@
* SPDX-License-Identifier: Apache-2.0
*/
import * as crypto from 'node:crypto';
import { Storage } from '../config/storage.js';
import { CoreEvent, coreEvents } from '../utils/events.js';
import type { AgentOverride, Config } from '../config/config.js';
import type { AgentDefinition, LocalAgentDefinition } from './types.js';
import { getAgentCardLoadOptions, getRemoteAgentTargetUrl } from './types.js';
import { loadAgentsFromDirectory } from './agentLoader.js';
import { CodebaseInvestigatorAgent } from './codebase-investigator.js';
import { CliHelpAgent } from './cli-help-agent.js';
@@ -162,7 +164,14 @@ export class AgentRegistry {
if (!agent.metadata) {
agent.metadata = {};
}
agent.metadata.hash = agent.agentCardUrl;
agent.metadata.hash =
agent.agentCardUrl ??
(agent.agentCardJson
? crypto
.createHash('sha256')
.update(agent.agentCardJson)
.digest('hex')
: undefined);
}
if (!agent.metadata?.hash) {
@@ -443,12 +452,13 @@ export class AgentRegistry {
);
return;
}
const targetUrl = getRemoteAgentTargetUrl(remoteDef);
let authHandler: AuthenticationHandler | undefined;
if (definition.auth) {
const provider = await A2AAuthProviderFactory.create({
authConfig: definition.auth,
agentName: definition.name,
targetUrl: definition.agentCardUrl,
targetUrl,
agentCardUrl: remoteDef.agentCardUrl,
});
if (!provider) {
@@ -461,7 +471,7 @@ export class AgentRegistry {
const agentCard = await clientManager.loadAgent(
remoteDef.name,
remoteDef.agentCardUrl,
getAgentCardLoadOptions(remoteDef),
authHandler,
);
@@ -515,7 +525,7 @@ export class AgentRegistry {
if (this.config.getDebugMode()) {
debugLogger.log(
`[AgentRegistry] Registered remote agent '${definition.name}' with card: ${definition.agentCardUrl}`,
`[AgentRegistry] Registered remote agent '${definition.name}' with card: ${definition.agentCardUrl ?? 'inline JSON'}`,
);
}
this.agents.set(definition.name, definition);
@@ -189,7 +189,7 @@ describe('RemoteAgentInvocation', () => {
expect(mockClientManager.loadAgent).toHaveBeenCalledWith(
'test-agent',
'http://test-agent/card',
{ type: 'url', url: 'http://test-agent/card' },
undefined,
);
});
@@ -240,7 +240,7 @@ describe('RemoteAgentInvocation', () => {
});
expect(mockClientManager.loadAgent).toHaveBeenCalledWith(
'test-agent',
'http://test-agent/card',
{ type: 'url', url: 'http://test-agent/card' },
mockHandler,
);
});
@@ -266,11 +266,10 @@ describe('RemoteAgentInvocation', () => {
);
const result = await invocation.execute(new AbortController().signal);
expect(result.returnDisplay).toMatchObject({
result: expect.stringContaining(
"Failed to create auth provider for agent 'test-agent'",
),
});
expect(result.returnDisplay).toMatchObject({ state: 'error' });
expect((result.returnDisplay as SubagentProgress).result).toContain(
"Failed to create auth provider for agent 'test-agent'",
);
});
it('should not load the agent if already present', async () => {
@@ -16,6 +16,8 @@ import {
type RemoteAgentDefinition,
type AgentInputs,
type SubagentProgress,
getAgentCardLoadOptions,
getRemoteAgentTargetUrl,
} from './types.js';
import { type AgentLoopContext } from '../config/agent-loop-context.js';
import type { MessageBus } from '../confirmation-bus/message-bus.js';
@@ -92,10 +94,11 @@ export class RemoteAgentInvocation extends BaseToolInvocation<
}
if (this.definition.auth) {
const targetUrl = getRemoteAgentTargetUrl(this.definition);
const provider = await A2AAuthProviderFactory.create({
authConfig: this.definition.auth,
agentName: this.definition.name,
targetUrl: this.definition.agentCardUrl,
targetUrl,
agentCardUrl: this.definition.agentCardUrl,
});
if (!provider) {
@@ -162,7 +165,7 @@ export class RemoteAgentInvocation extends BaseToolInvocation<
if (!this.clientManager.getClient(this.definition.name)) {
await this.clientManager.loadAgent(
this.definition.name,
this.definition.agentCardUrl,
getAgentCardLoadOptions(this.definition),
authHandler,
);
}
+65 -2
View File
@@ -13,6 +13,7 @@ import type { AnyDeclarativeTool } from '../tools/tools.js';
import { type z } from 'zod';
import type { ModelConfig } from '../services/modelConfigService.js';
import type { AnySchema } from 'ajv';
import type { AgentCard } from '@a2a-js/sdk';
import type { A2AAuthConfig } from './auth-provider/types.js';
import type { MCPServerConfig } from '../config/config.js';
@@ -128,6 +129,62 @@ export function isToolActivityError(data: unknown): boolean {
* The base definition for an agent.
* @template TOutput The specific Zod schema for the agent's final output object.
*/
export type AgentCardLoadOptions =
| { type: 'url'; url: string }
| { type: 'json'; json: string };
/** Minimal shape needed by helper functions, avoids generic TOutput constraints. */
interface RemoteAgentRef {
name: string;
agentCardUrl?: string;
agentCardJson?: string;
}
/**
* Derives the AgentCardLoadOptions from a RemoteAgentDefinition.
* Throws if neither agentCardUrl nor agentCardJson is present.
*/
export function getAgentCardLoadOptions(
def: RemoteAgentRef,
): AgentCardLoadOptions {
if (def.agentCardJson) {
return { type: 'json', json: def.agentCardJson };
}
if (def.agentCardUrl) {
return { type: 'url', url: def.agentCardUrl };
}
throw new Error(
`Remote agent '${def.name}' has neither agentCardUrl nor agentCardJson`,
);
}
/**
* Extracts a target URL for auth providers from a RemoteAgentDefinition.
* For URL-based agents, returns the agentCardUrl.
* For JSON-based agents, attempts to parse the URL from the inline card JSON.
* Returns undefined if no URL can be determined.
*/
export function getRemoteAgentTargetUrl(
def: RemoteAgentRef,
): string | undefined {
if (def.agentCardUrl) {
return def.agentCardUrl;
}
if (def.agentCardJson) {
try {
const parsed: unknown = JSON.parse(def.agentCardJson);
// eslint-disable-next-line @typescript-eslint/no-unsafe-type-assertion
const card = parsed as AgentCard;
if (card.url) {
return card.url;
}
} catch {
// JSON parse will fail properly later in loadAgent
}
}
return undefined;
}
export interface BaseAgentDefinition<
TOutput extends z.ZodTypeAny = z.ZodUnknown,
> {
@@ -172,11 +229,10 @@ export interface LocalAgentDefinition<
processOutput?: (output: z.infer<TOutput>) => string;
}
export interface RemoteAgentDefinition<
export interface BaseRemoteAgentDefinition<
TOutput extends z.ZodTypeAny = z.ZodUnknown,
> extends BaseAgentDefinition<TOutput> {
kind: 'remote';
agentCardUrl: string;
/** The user-provided description, before any remote card merging. */
originalDescription?: string;
/**
@@ -187,6 +243,13 @@ export interface RemoteAgentDefinition<
auth?: A2AAuthConfig;
}
export interface RemoteAgentDefinition<
TOutput extends z.ZodTypeAny = z.ZodUnknown,
> extends BaseRemoteAgentDefinition<TOutput> {
agentCardUrl?: string;
agentCardJson?: string;
}
export type AgentDefinition<TOutput extends z.ZodTypeAny = z.ZodUnknown> =
| LocalAgentDefinition<TOutput>
| RemoteAgentDefinition<TOutput>;
@@ -28,6 +28,7 @@ describe('policyCatalog', () => {
const chain = getModelPolicyChain({
previewEnabled: true,
useGemini31: true,
useGemini31FlashLite: false,
});
expect(chain[0]?.model).toBe(PREVIEW_GEMINI_3_1_MODEL);
expect(chain).toHaveLength(2);
@@ -38,6 +39,7 @@ describe('policyCatalog', () => {
const chain = getModelPolicyChain({
previewEnabled: true,
useGemini31: true,
useGemini31FlashLite: false,
useCustomToolModel: true,
});
expect(chain[0]?.model).toBe(PREVIEW_GEMINI_3_1_CUSTOM_TOOLS_MODEL);
@@ -30,6 +30,7 @@ export interface ModelPolicyOptions {
previewEnabled: boolean;
userTier?: UserTierId;
useGemini31?: boolean;
useGemini31FlashLite?: boolean;
useCustomToolModel?: boolean;
}
@@ -85,6 +86,7 @@ export function getModelPolicyChain(
const previewModel = resolveModel(
PREVIEW_GEMINI_MODEL,
options.useGemini31,
options.useGemini31FlashLite,
options.useCustomToolModel,
);
return [
@@ -27,6 +27,7 @@ const createMockConfig = (overrides: Partial<Config> = {}): Config => {
getUserTier: () => undefined,
getModel: () => 'gemini-2.5-pro',
getGemini31LaunchedSync: () => false,
getGemini31FlashLiteLaunchedSync: () => false,
getUseCustomToolModelSync: () => {
const useGemini31 = config.getGemini31LaunchedSync();
const authType = config.getContentGeneratorConfig().authType;
@@ -203,6 +204,7 @@ describe('policyHelpers', () => {
getExperimentalDynamicModelConfiguration: () => dynamic,
getModel: () => model,
getGemini31LaunchedSync: () => useGemini31 ?? false,
getGemini31FlashLiteLaunchedSync: () => false,
getHasAccessToPreviewModel: () => hasAccess ?? true,
getContentGeneratorConfig: () => ({ authType }),
modelConfigService: new ModelConfigService(DEFAULT_MODEL_CONFIGS),
@@ -45,12 +45,15 @@ export function resolvePolicyChain(
let chain;
const useGemini31 = config.getGemini31LaunchedSync?.() ?? false;
const useGemini31FlashLite =
config.getGemini31FlashLiteLaunchedSync?.() ?? false;
const useCustomToolModel = config.getUseCustomToolModelSync?.() ?? false;
const hasAccessToPreview = config.getHasAccessToPreviewModel?.() ?? true;
const resolvedModel = resolveModel(
modelFromConfig,
useGemini31,
useGemini31FlashLite,
useCustomToolModel,
hasAccessToPreview,
config,
@@ -64,6 +67,7 @@ export function resolvePolicyChain(
if (config.getExperimentalDynamicModelConfiguration?.() === true) {
const context = {
useGemini3_1: useGemini31,
useGemini3_1FlashLite: useGemini31FlashLite,
useCustomTools: useCustomToolModel,
};
@@ -120,6 +124,7 @@ export function resolvePolicyChain(
previewEnabled,
userTier: config.getUserTier(),
useGemini31,
useGemini31FlashLite,
useCustomToolModel,
});
} else {
@@ -129,6 +134,7 @@ export function resolvePolicyChain(
previewEnabled: false,
userTier: config.getUserTier(),
useGemini31,
useGemini31FlashLite,
useCustomToolModel,
});
}
@@ -18,6 +18,7 @@ export const ExperimentFlags = {
MASKING_PROTECT_LATEST_TURN: 45758819,
GEMINI_3_1_PRO_LAUNCHED: 45760185,
PRO_MODEL_NO_ACCESS: 45768879,
GEMINI_3_1_FLASH_LITE_LAUNCHED: 45771641,
} as const;
export type ExperimentFlagName =
+44 -2
View File
@@ -1027,7 +1027,7 @@ export class Config implements McpContext, AgentLoopContext {
this.model = params.model;
this.disableLoopDetection = params.disableLoopDetection ?? false;
this._activeModel = params.model;
this.enableAgents = params.enableAgents ?? false;
this.enableAgents = params.enableAgents ?? true;
this.agents = params.agents ?? {};
this.disableLLMCorrection = params.disableLLMCorrection ?? true;
this.planEnabled = params.plan ?? true;
@@ -1821,6 +1821,10 @@ export class Config implements McpContext, AgentLoopContext {
const primaryModel = resolveModel(
this.getModel(),
this.getGemini31LaunchedSync(),
this.getGemini31FlashLiteLaunchedSync(),
this.getUseCustomToolModelSync(),
this.getHasAccessToPreviewModel(),
this,
);
return this.modelQuotas.get(primaryModel)?.remaining;
}
@@ -1833,6 +1837,10 @@ export class Config implements McpContext, AgentLoopContext {
const primaryModel = resolveModel(
this.getModel(),
this.getGemini31LaunchedSync(),
this.getGemini31FlashLiteLaunchedSync(),
this.getUseCustomToolModelSync(),
this.getHasAccessToPreviewModel(),
this,
);
return this.modelQuotas.get(primaryModel)?.limit;
}
@@ -1845,6 +1853,10 @@ export class Config implements McpContext, AgentLoopContext {
const primaryModel = resolveModel(
this.getModel(),
this.getGemini31LaunchedSync(),
this.getGemini31FlashLiteLaunchedSync(),
this.getUseCustomToolModelSync(),
this.getHasAccessToPreviewModel(),
this,
);
return this.modelQuotas.get(primaryModel)?.resetTime;
}
@@ -2911,7 +2923,7 @@ export class Config implements McpContext, AgentLoopContext {
}
/**
* Returns whether Gemini 3.1 has been launched.
* Returns whether Gemini 3.1 Pro has been launched.
* This method is async and ensures that experiments are loaded before returning the result.
*/
async getGemini31Launched(): Promise<boolean> {
@@ -2919,6 +2931,15 @@ export class Config implements McpContext, AgentLoopContext {
return this.getGemini31LaunchedSync();
}
/**
* Returns whether Gemini 3.1 Flash Lite has been launched.
* This method is async and ensures that experiments are loaded before returning the result.
*/
async getGemini31FlashLiteLaunched(): Promise<boolean> {
await this.ensureExperimentsLoaded();
return this.getGemini31FlashLiteLaunchedSync();
}
/**
* Returns whether the custom tool model should be used.
*/
@@ -2960,6 +2981,27 @@ export class Config implements McpContext, AgentLoopContext {
);
}
/**
* Returns whether Gemini 3.1 Flash Lite has been launched.
*
* Note: This method should only be called after startup, once experiments have been loaded.
* If you need to call this during startup or from an async context, use
* getGemini31FlashLiteLaunched instead.
*/
getGemini31FlashLiteLaunchedSync(): boolean {
const authType = this.contentGeneratorConfig?.authType;
if (
authType === AuthType.USE_GEMINI ||
authType === AuthType.USE_VERTEX_AI
) {
return true;
}
return (
this.experiments?.flags[ExperimentFlags.GEMINI_3_1_FLASH_LITE_LAUNCHED]
?.boolValue ?? false
);
}
private async ensureExperimentsLoaded(): Promise<void> {
if (!this.experimentsPromise) {
return;
@@ -218,6 +218,11 @@ export const DEFAULT_MODEL_CONFIGS: ModelConfigServiceConfig = {
model: 'gemini-3-flash-preview',
},
},
'chat-compression-3.1-flash-lite': {
modelConfig: {
model: 'gemini-3.1-flash-lite-preview',
},
},
'chat-compression-2.5-pro': {
modelConfig: {
model: 'gemini-2.5-pro',
@@ -432,6 +437,15 @@ export const DEFAULT_MODEL_CONFIGS: ModelConfigServiceConfig = {
'auto-gemini-2.5': {
default: 'gemini-2.5-pro',
},
'gemini-3.1-flash-lite-preview': {
default: 'gemini-3.1-flash-lite-preview',
contexts: [
{
condition: { useGemini3_1FlashLite: false },
target: 'gemini-2.5-flash-lite',
},
],
},
flash: {
default: 'gemini-3-flash-preview',
contexts: [
@@ -443,6 +457,12 @@ export const DEFAULT_MODEL_CONFIGS: ModelConfigServiceConfig = {
},
'flash-lite': {
default: 'gemini-2.5-flash-lite',
contexts: [
{
condition: { useGemini3_1FlashLite: true },
target: 'gemini-3.1-flash-lite-preview',
},
],
},
},
classifierIdResolutions: {
+94 -29
View File
@@ -21,6 +21,7 @@ import {
supportsMultimodalFunctionResponse,
GEMINI_MODEL_ALIAS_PRO,
GEMINI_MODEL_ALIAS_FLASH,
GEMINI_MODEL_ALIAS_FLASH_LITE,
GEMINI_MODEL_ALIAS_AUTO,
PREVIEW_GEMINI_FLASH_MODEL,
PREVIEW_GEMINI_MODEL_AUTO,
@@ -61,9 +62,26 @@ describe('Dynamic Configuration Parity', () => {
];
const flagCombos = [
{ useGemini3_1: false, useCustomToolModel: false },
{ useGemini3_1: true, useCustomToolModel: false },
{ useGemini3_1: true, useCustomToolModel: true },
{
useGemini3_1: false,
useGemini3_1FlashLite: false,
useCustomToolModel: false,
},
{
useGemini3_1: true,
useGemini3_1FlashLite: false,
useCustomToolModel: false,
},
{
useGemini3_1: true,
useGemini3_1FlashLite: true,
useCustomToolModel: false,
},
{
useGemini3_1: true,
useGemini3_1FlashLite: true,
useCustomToolModel: true,
},
];
it('resolveModel should match legacy behavior when dynamicModelConfiguration flag enabled.', () => {
@@ -84,6 +102,7 @@ describe('Dynamic Configuration Parity', () => {
const legacy = resolveModel(
model,
flags.useGemini3_1,
flags.useGemini3_1FlashLite,
flags.useCustomToolModel,
hasAccess,
mockLegacyConfig,
@@ -91,6 +110,7 @@ describe('Dynamic Configuration Parity', () => {
const dynamic = resolveModel(
model,
flags.useGemini3_1,
flags.useGemini3_1FlashLite,
flags.useCustomToolModel,
hasAccess,
mockDynamicConfig,
@@ -129,6 +149,7 @@ describe('Dynamic Configuration Parity', () => {
anchor,
tier,
flags.useGemini3_1,
flags.useGemini3_1FlashLite,
flags.useCustomToolModel,
hasAccess,
mockLegacyConfig,
@@ -137,6 +158,7 @@ describe('Dynamic Configuration Parity', () => {
anchor,
tier,
flags.useGemini3_1,
flags.useGemini3_1FlashLite,
flags.useCustomToolModel,
hasAccess,
mockDynamicConfig,
@@ -369,7 +391,7 @@ describe('resolveModel', () => {
});
it('should return Gemini 3.1 Pro Custom Tools when auto-gemini-3 is requested, useGemini3_1 is true, and useCustomToolModel is true', () => {
const model = resolveModel(PREVIEW_GEMINI_MODEL_AUTO, true, true);
const model = resolveModel(PREVIEW_GEMINI_MODEL_AUTO, true, false, true);
expect(model).toBe(PREVIEW_GEMINI_3_1_CUSTOM_TOOLS_MODEL);
});
@@ -378,6 +400,16 @@ describe('resolveModel', () => {
expect(model).toBe(DEFAULT_GEMINI_MODEL);
});
it('should return the Default Flash-Lite model when flash-lite is requested', () => {
const model = resolveModel(GEMINI_MODEL_ALIAS_FLASH_LITE);
expect(model).toBe(DEFAULT_GEMINI_FLASH_LITE_MODEL);
});
it('should return the Preview Flash-Lite model when flash-lite is requested and useGemini3_1FlashLite is true', () => {
const model = resolveModel(GEMINI_MODEL_ALIAS_FLASH_LITE, false, true);
expect(model).toBe(PREVIEW_GEMINI_3_1_FLASH_LITE_MODEL);
});
it('should return the requested model as-is for explicit specific models', () => {
expect(resolveModel(DEFAULT_GEMINI_MODEL)).toBe(DEFAULT_GEMINI_MODEL);
expect(resolveModel(DEFAULT_GEMINI_FLASH_MODEL)).toBe(
@@ -397,39 +429,45 @@ describe('resolveModel', () => {
describe('hasAccessToPreview logic', () => {
it('should return default model when access to preview is false and preview model is requested', () => {
expect(resolveModel(PREVIEW_GEMINI_MODEL, false, false, false)).toBe(
DEFAULT_GEMINI_MODEL,
);
expect(
resolveModel(PREVIEW_GEMINI_MODEL, false, false, false, false),
).toBe(DEFAULT_GEMINI_MODEL);
});
it('should return default flash model when access to preview is false and preview flash model is requested', () => {
expect(
resolveModel(PREVIEW_GEMINI_FLASH_MODEL, false, false, false),
resolveModel(PREVIEW_GEMINI_FLASH_MODEL, false, false, false, false),
).toBe(DEFAULT_GEMINI_FLASH_MODEL);
});
it('should return default flash lite model when access to preview is false and preview flash lite model is requested', () => {
expect(
resolveModel(PREVIEW_GEMINI_3_1_FLASH_LITE_MODEL, false, false, false),
resolveModel(
PREVIEW_GEMINI_3_1_FLASH_LITE_MODEL,
false,
false,
false,
false,
),
).toBe(DEFAULT_GEMINI_FLASH_LITE_MODEL);
});
it('should return default model when access to preview is false and auto-gemini-3 is requested', () => {
expect(resolveModel(PREVIEW_GEMINI_MODEL_AUTO, false, false, false)).toBe(
DEFAULT_GEMINI_MODEL,
);
expect(
resolveModel(PREVIEW_GEMINI_MODEL_AUTO, false, false, false, false),
).toBe(DEFAULT_GEMINI_MODEL);
});
it('should return default model when access to preview is false and Gemini 3.1 is requested', () => {
expect(resolveModel(PREVIEW_GEMINI_MODEL_AUTO, true, false, false)).toBe(
DEFAULT_GEMINI_MODEL,
);
expect(
resolveModel(PREVIEW_GEMINI_MODEL_AUTO, true, false, false, false),
).toBe(DEFAULT_GEMINI_MODEL);
});
it('should still return default model when access to preview is false and auto-gemini-2.5 is requested', () => {
expect(resolveModel(DEFAULT_GEMINI_MODEL_AUTO, false, false, false)).toBe(
DEFAULT_GEMINI_MODEL,
);
expect(
resolveModel(DEFAULT_GEMINI_MODEL_AUTO, false, false, false, false),
).toBe(DEFAULT_GEMINI_MODEL);
});
});
});
@@ -521,6 +559,7 @@ describe('resolveClassifierModel', () => {
PREVIEW_GEMINI_MODEL_AUTO,
GEMINI_MODEL_ALIAS_PRO,
true,
false,
true,
),
).toBe(PREVIEW_GEMINI_3_1_CUSTOM_TOOLS_MODEL);
@@ -532,7 +571,11 @@ describe('isActiveModel', () => {
expect(isActiveModel(DEFAULT_GEMINI_MODEL)).toBe(true);
expect(isActiveModel(PREVIEW_GEMINI_MODEL)).toBe(true);
expect(isActiveModel(DEFAULT_GEMINI_FLASH_MODEL)).toBe(true);
expect(isActiveModel(PREVIEW_GEMINI_3_1_FLASH_LITE_MODEL)).toBe(true);
});
it('should return false for Gemini 3.1 models when Gemini 3.1 is not launched', () => {
expect(isActiveModel(PREVIEW_GEMINI_3_1_MODEL)).toBe(false);
expect(isActiveModel(PREVIEW_GEMINI_3_1_FLASH_LITE_MODEL)).toBe(false);
});
it('should return true for unknown models and aliases', () => {
@@ -546,31 +589,53 @@ describe('isActiveModel', () => {
it('should return true for other valid models when useGemini3_1 is true', () => {
expect(isActiveModel(DEFAULT_GEMINI_MODEL, true)).toBe(true);
expect(isActiveModel(PREVIEW_GEMINI_3_1_FLASH_LITE_MODEL, true)).toBe(true);
});
it('should return true for PREVIEW_GEMINI_3_1_FLASH_LITE_MODEL only when useGemini3_1FlashLite is true', () => {
expect(
isActiveModel(PREVIEW_GEMINI_3_1_FLASH_LITE_MODEL, false, true),
).toBe(true);
expect(isActiveModel(PREVIEW_GEMINI_3_1_FLASH_LITE_MODEL, true, true)).toBe(
true,
);
expect(
isActiveModel(PREVIEW_GEMINI_3_1_FLASH_LITE_MODEL, true, false),
).toBe(false);
});
it('should correctly filter Gemini 3.1 models based on useCustomToolModel when useGemini3_1 is true', () => {
// When custom tools are preferred, standard 3.1 should be inactive
expect(isActiveModel(PREVIEW_GEMINI_3_1_MODEL, true, true)).toBe(false);
expect(isActiveModel(PREVIEW_GEMINI_3_1_MODEL, true, false, true)).toBe(
false,
);
expect(
isActiveModel(PREVIEW_GEMINI_3_1_CUSTOM_TOOLS_MODEL, true, true),
isActiveModel(PREVIEW_GEMINI_3_1_CUSTOM_TOOLS_MODEL, true, false, true),
).toBe(true);
// When custom tools are NOT preferred, custom tools 3.1 should be inactive
expect(isActiveModel(PREVIEW_GEMINI_3_1_MODEL, true, false)).toBe(true);
expect(isActiveModel(PREVIEW_GEMINI_3_1_MODEL, true, false, false)).toBe(
true,
);
expect(
isActiveModel(PREVIEW_GEMINI_3_1_CUSTOM_TOOLS_MODEL, true, false),
isActiveModel(PREVIEW_GEMINI_3_1_CUSTOM_TOOLS_MODEL, true, false, false),
).toBe(false);
});
it('should return false for both Gemini 3.1 models when useGemini3_1 is false', () => {
expect(isActiveModel(PREVIEW_GEMINI_3_1_MODEL, false, true)).toBe(false);
expect(isActiveModel(PREVIEW_GEMINI_3_1_MODEL, false, false)).toBe(false);
it('should return false for Gemini 3.1 models when useGemini3_1 and useGemini3_1FlashLite are false', () => {
expect(isActiveModel(PREVIEW_GEMINI_3_1_MODEL, false, false, true)).toBe(
false,
);
expect(isActiveModel(PREVIEW_GEMINI_3_1_MODEL, false, false, false)).toBe(
false,
);
expect(
isActiveModel(PREVIEW_GEMINI_3_1_CUSTOM_TOOLS_MODEL, false, true),
isActiveModel(PREVIEW_GEMINI_3_1_CUSTOM_TOOLS_MODEL, false, false, true),
).toBe(false);
expect(
isActiveModel(PREVIEW_GEMINI_3_1_CUSTOM_TOOLS_MODEL, false, false),
isActiveModel(PREVIEW_GEMINI_3_1_CUSTOM_TOOLS_MODEL, false, false, false),
).toBe(false);
expect(
isActiveModel(PREVIEW_GEMINI_3_1_FLASH_LITE_MODEL, false, false),
).toBe(false);
});
});
+23 -3
View File
@@ -6,6 +6,7 @@
export interface ModelResolutionContext {
useGemini3_1?: boolean;
useGemini3_1FlashLite?: boolean;
useCustomTools?: boolean;
hasAccessToPreview?: boolean;
requestedModel?: string;
@@ -97,6 +98,7 @@ export const DEFAULT_THINKING_MODE = 8192;
export function resolveModel(
requestedModel: string,
useGemini3_1: boolean = false,
useGemini3_1FlashLite: boolean = false,
useCustomToolModel: boolean = false,
hasAccessToPreview: boolean = true,
config?: ModelCapabilityContext,
@@ -104,6 +106,7 @@ export function resolveModel(
if (config?.getExperimentalDynamicModelConfiguration?.() === true) {
const resolved = config.modelConfigService.resolveModelId(requestedModel, {
useGemini3_1,
useGemini3_1FlashLite,
useCustomTools: useCustomToolModel,
hasAccessToPreview,
});
@@ -146,7 +149,9 @@ export function resolveModel(
break;
}
case GEMINI_MODEL_ALIAS_FLASH_LITE: {
resolved = DEFAULT_GEMINI_FLASH_LITE_MODEL;
resolved = useGemini3_1FlashLite
? PREVIEW_GEMINI_3_1_FLASH_LITE_MODEL
: DEFAULT_GEMINI_FLASH_LITE_MODEL;
break;
}
default: {
@@ -160,6 +165,8 @@ export function resolveModel(
switch (resolved) {
case PREVIEW_GEMINI_FLASH_MODEL:
return DEFAULT_GEMINI_FLASH_MODEL;
case PREVIEW_GEMINI_3_1_FLASH_LITE_MODEL:
return DEFAULT_GEMINI_FLASH_LITE_MODEL;
case PREVIEW_GEMINI_MODEL:
case PREVIEW_GEMINI_3_1_MODEL:
case PREVIEW_GEMINI_3_1_CUSTOM_TOOLS_MODEL:
@@ -193,6 +200,7 @@ export function resolveClassifierModel(
requestedModel: string,
modelAlias: string,
useGemini3_1: boolean = false,
useGemini3_1FlashLite: boolean = false,
useCustomToolModel: boolean = false,
hasAccessToPreview: boolean = true,
config?: ModelCapabilityContext,
@@ -203,6 +211,7 @@ export function resolveClassifierModel(
requestedModel,
{
useGemini3_1,
useGemini3_1FlashLite,
useCustomTools: useCustomToolModel,
hasAccessToPreview,
},
@@ -224,7 +233,12 @@ export function resolveClassifierModel(
}
return resolveModel(GEMINI_MODEL_ALIAS_FLASH);
}
return resolveModel(requestedModel, useGemini3_1, useCustomToolModel);
return resolveModel(
requestedModel,
useGemini3_1,
useGemini3_1FlashLite,
useCustomToolModel,
);
}
export function getDisplayString(
@@ -249,6 +263,8 @@ export function getDisplayString(
return PREVIEW_GEMINI_FLASH_MODEL;
case PREVIEW_GEMINI_3_1_CUSTOM_TOOLS_MODEL:
return PREVIEW_GEMINI_3_1_MODEL;
case PREVIEW_GEMINI_3_1_FLASH_LITE_MODEL:
return PREVIEW_GEMINI_3_1_FLASH_LITE_MODEL;
default:
return model;
}
@@ -347,7 +363,7 @@ export function isCustomModel(
config?: ModelCapabilityContext,
): boolean {
if (config?.getExperimentalDynamicModelConfiguration?.() === true) {
const resolved = resolveModel(model, false, false, true, config);
const resolved = resolveModel(model, false, false, false, true, config);
return (
config.modelConfigService.getModelDefinition(resolved)?.tier ===
'custom' || !resolved.startsWith('gemini-')
@@ -420,11 +436,15 @@ export function supportsMultimodalFunctionResponse(
export function isActiveModel(
model: string,
useGemini3_1: boolean = false,
useGemini3_1FlashLite: boolean = false,
useCustomToolModel: boolean = false,
): boolean {
if (!VALID_GEMINI_MODELS.has(model)) {
return false;
}
if (model === PREVIEW_GEMINI_3_1_FLASH_LITE_MODEL) {
return useGemini3_1FlashLite;
}
if (useGemini3_1) {
if (model === PREVIEW_GEMINI_MODEL) {
return false;
+1
View File
@@ -575,6 +575,7 @@ export class GeminiClient {
return resolveModel(
this.config.getActiveModel(),
this.config.getGemini31LaunchedSync?.() ?? false,
this.config.getGemini31FlashLiteLaunchedSync?.() ?? false,
false,
this.config.getHasAccessToPreviewModel?.() ?? true,
this.config,
@@ -180,6 +180,9 @@ export async function createContentGenerator(
config.authType === AuthType.USE_GEMINI ||
config.authType === AuthType.USE_VERTEX_AI ||
((await gcConfig.getGemini31Launched?.()) ?? false),
config.authType === AuthType.USE_GEMINI ||
config.authType === AuthType.USE_VERTEX_AI ||
((await gcConfig.getGemini31FlashLiteLaunched?.()) ?? false),
false,
gcConfig.getHasAccessToPreviewModel?.() ?? true,
gcConfig,
+9 -2
View File
@@ -524,12 +524,18 @@ export class GeminiChat {
const apiCall = async () => {
const useGemini3_1 =
(await this.context.config.getGemini31Launched?.()) ?? false;
const useGemini3_1FlashLite =
(await this.context.config.getGemini31FlashLiteLaunched?.()) ?? false;
const hasAccessToPreview =
this.context.config.getHasAccessToPreviewModel?.() ?? true;
// Default to the last used model (which respects arguments/availability selection)
let modelToUse = resolveModel(
lastModelToUse,
useGemini3_1,
useGemini3_1FlashLite,
false,
this.context.config.getHasAccessToPreviewModel?.() ?? true,
hasAccessToPreview,
this.context.config,
);
@@ -539,8 +545,9 @@ export class GeminiChat {
modelToUse = resolveModel(
this.context.config.getActiveModel(),
useGemini3_1,
useGemini3_1FlashLite,
false,
this.context.config.getHasAccessToPreviewModel?.() ?? true,
hasAccessToPreview,
this.context.config,
);
}
+5 -1
View File
@@ -88,7 +88,11 @@ export * from './utils/approvalModeUtils.js';
export * from './utils/fileDiffUtils.js';
export * from './utils/retry.js';
export * from './utils/shell-utils.js';
export { PolicyDecision, ApprovalMode } from './policy/types.js';
export {
PolicyDecision,
ApprovalMode,
PRIORITY_YOLO_ALLOW_ALL,
} from './policy/types.js';
export * from './utils/tool-utils.js';
export * from './utils/terminalSerializer.js';
export * from './utils/systemEncoding.js';
@@ -110,6 +110,8 @@ priority = 70
modes = ["plan"]
# Allow write_file and replace for .md files in the plans directory (cross-platform)
# We split this into two rules to avoid ReDoS checker issues with nested optional segments.
# This rule handles the case where there is a session ID in the plan file path
[[rule]]
toolName = ["write_file", "replace"]
decision = "allow"
@@ -117,6 +119,14 @@ priority = 70
modes = ["plan"]
argsPattern = "\\x00\"file_path\":\"[^\"]+[\\\\/]+\\.gemini[\\\\/]+tmp[\\\\/]+[\\w-]+[\\\\/]+[\\w-]+[\\\\/]+plans[\\\\/]+[\\w-]+\\.md\"\\x00"
# This rule handles the case where there isn't a session ID in the plan file path
[[rule]]
toolName = ["write_file", "replace"]
decision = "allow"
priority = 70
modes = ["plan"]
argsPattern = "\\x00\"file_path\":\"[^\"]+[\\\\/]+\\.gemini[\\\\/]+tmp[\\\\/]+[\\w-]+[\\\\/]+plans[\\\\/]+[\\w-]+\\.md\"\\x00"
# Explicitly Deny other write operations in Plan mode with a clear message.
[[rule]]
toolName = ["write_file", "replace"]
@@ -702,15 +702,6 @@ export class PolicyEngine {
}
}
// Sandbox Expansion requests MUST always be confirmed by the user,
// even if the base command is otherwise ALLOWED by the policy engine.
if (
decision === PolicyDecision.ALLOW &&
toolCall.args?.['additional_permissions']
) {
decision = PolicyDecision.ASK_USER;
}
return {
decision: this.applyNonInteractiveMode(decision),
rule: matchedRule,
@@ -62,6 +62,7 @@ export class PromptProvider {
const desiredModel = resolveModel(
context.config.getActiveModel(),
context.config.getGemini31LaunchedSync?.() ?? false,
context.config.getGemini31FlashLiteLaunchedSync?.() ?? false,
false,
context.config.getHasAccessToPreviewModel?.() ?? true,
context.config,
@@ -247,6 +248,7 @@ export class PromptProvider {
const desiredModel = resolveModel(
context.config.getActiveModel(),
context.config.getGemini31LaunchedSync?.() ?? false,
context.config.getGemini31FlashLiteLaunchedSync?.() ?? false,
false,
context.config.getHasAccessToPreviewModel?.() ?? true,
context.config,
@@ -59,6 +59,7 @@ describe('ClassifierStrategy', () => {
getModel: vi.fn().mockReturnValue(DEFAULT_GEMINI_MODEL_AUTO),
getNumericalRoutingEnabled: vi.fn().mockResolvedValue(false),
getGemini31Launched: vi.fn().mockResolvedValue(false),
getGemini31FlashLiteLaunched: vi.fn().mockResolvedValue(false),
getUseCustomToolModel: vi.fn().mockImplementation(async () => {
const launched = await mockConfig.getGemini31Launched();
const authType = mockConfig.getContentGeneratorConfig().authType;
@@ -171,14 +171,17 @@ export class ClassifierStrategy implements RoutingStrategy {
const reasoning = routerResponse.reasoning;
const latencyMs = Date.now() - startTime;
const [useGemini3_1, useCustomToolModel] = await Promise.all([
config.getGemini31Launched(),
config.getUseCustomToolModel(),
]);
const [useGemini3_1, useGemini3_1FlashLite, useCustomToolModel] =
await Promise.all([
config.getGemini31Launched(),
config.getGemini31FlashLiteLaunched(),
config.getUseCustomToolModel(),
]);
const selectedModel = resolveClassifierModel(
model,
routerResponse.model_choice,
useGemini3_1,
useGemini3_1FlashLite,
useCustomToolModel,
config.getHasAccessToPreviewModel?.() ?? true,
config,
@@ -26,6 +26,7 @@ export class DefaultStrategy implements TerminalStrategy {
const defaultModel = resolveModel(
config.getModel(),
config.getGemini31LaunchedSync?.() ?? false,
config.getGemini31FlashLiteLaunchedSync?.() ?? false,
false,
config.getHasAccessToPreviewModel?.() ?? true,
config,
@@ -28,6 +28,7 @@ export class FallbackStrategy implements RoutingStrategy {
const resolvedModel = resolveModel(
requestedModel,
config.getGemini31LaunchedSync?.() ?? false,
config.getGemini31FlashLiteLaunchedSync?.() ?? false,
false,
config.getHasAccessToPreviewModel?.() ?? true,
config,
@@ -59,6 +59,7 @@ describe('NumericalClassifierStrategy', () => {
getResolvedClassifierThreshold: vi.fn().mockResolvedValue(90),
getClassifierThreshold: vi.fn().mockResolvedValue(undefined),
getGemini31Launched: vi.fn().mockResolvedValue(false),
getGemini31FlashLiteLaunched: vi.fn().mockResolvedValue(false),
getUseCustomToolModel: vi.fn().mockImplementation(async () => {
const launched = await mockConfig.getGemini31Launched();
const authType = mockConfig.getContentGeneratorConfig().authType;
@@ -147,14 +147,17 @@ export class NumericalClassifierStrategy implements RoutingStrategy {
const { threshold, groupLabel, modelAlias } =
await this.getRoutingDecision(score, config);
const [useGemini3_1, useCustomToolModel] = await Promise.all([
config.getGemini31Launched(),
config.getUseCustomToolModel(),
]);
const [useGemini3_1, useGemini3_1FlashLite, useCustomToolModel] =
await Promise.all([
config.getGemini31Launched(),
config.getGemini31FlashLiteLaunched(),
config.getUseCustomToolModel(),
]);
const selectedModel = resolveClassifierModel(
model,
modelAlias,
useGemini3_1,
useGemini3_1FlashLite,
useCustomToolModel,
config.getHasAccessToPreviewModel?.() ?? true,
config,
@@ -38,6 +38,7 @@ export class OverrideStrategy implements RoutingStrategy {
model: resolveModel(
overrideModel,
config.getGemini31LaunchedSync?.() ?? false,
config.getGemini31FlashLiteLaunchedSync?.() ?? false,
false,
config.getHasAccessToPreviewModel?.() ?? true,
config,
@@ -153,7 +153,10 @@ describe('MacOsSandboxManager', () => {
SAFE_VAR: '1',
GITHUB_TOKEN: 'sensitive',
},
policy: mockPolicy,
policy: {
...mockPolicy,
sanitizationConfig: { enableEnvironmentVariableRedaction: true },
},
});
expect(result.env['SAFE_VAR']).toBe('1');
@@ -30,6 +30,7 @@ import {
PREVIEW_GEMINI_MODEL,
PREVIEW_GEMINI_FLASH_MODEL,
PREVIEW_GEMINI_3_1_MODEL,
PREVIEW_GEMINI_3_1_FLASH_LITE_MODEL,
} from '../config/models.js';
import { PreCompressTrigger } from '../hooks/types.js';
@@ -105,6 +106,8 @@ export function modelStringToModelConfigAlias(model: string): string {
return 'chat-compression-3-pro';
case PREVIEW_GEMINI_FLASH_MODEL:
return 'chat-compression-3-flash';
case PREVIEW_GEMINI_3_1_FLASH_LITE_MODEL:
return 'chat-compression-3.1-flash-lite';
case DEFAULT_GEMINI_MODEL:
return 'chat-compression-2.5-pro';
case DEFAULT_GEMINI_FLASH_MODEL:
@@ -375,9 +375,9 @@ describe('sanitizeEnvironment', () => {
});
describe('getSecureSanitizationConfig', () => {
it('should enable environment variable redaction by default', () => {
it('should default enableEnvironmentVariableRedaction to false', () => {
const config = getSecureSanitizationConfig();
expect(config.enableEnvironmentVariableRedaction).toBe(true);
expect(config.enableEnvironmentVariableRedaction).toBe(false);
});
it('should merge allowed and blocked variables from base and requested configs', () => {
@@ -440,13 +440,13 @@ describe('getSecureSanitizationConfig', () => {
expect(config.blockedEnvironmentVariables).toEqual(['BLOCKED_VAR']);
});
it('should force enableEnvironmentVariableRedaction to true even if requested false', () => {
it('should respect requested enableEnvironmentVariableRedaction value', () => {
const requestedConfig = {
enableEnvironmentVariableRedaction: false,
};
const config = getSecureSanitizationConfig(requestedConfig);
expect(config.enableEnvironmentVariableRedaction).toBe(true);
expect(config.enableEnvironmentVariableRedaction).toBe(false);
});
});
@@ -230,6 +230,9 @@ export function getSecureSanitizationConfig(
allowedEnvironmentVariables: [...new Set(allowed)],
blockedEnvironmentVariables: [...new Set(blocked)],
// Redaction must be enabled for secure configurations
enableEnvironmentVariableRedaction: true,
enableEnvironmentVariableRedaction:
requestedConfig.enableEnvironmentVariableRedaction ??
baseConfig?.enableEnvironmentVariableRedaction ??
false,
};
}
@@ -90,6 +90,7 @@ export interface ModelResolution {
/** The actual state of the current session. */
export interface ResolutionContext {
useGemini3_1?: boolean;
useGemini3_1FlashLite?: boolean;
useCustomTools?: boolean;
hasAccessToPreview?: boolean;
requestedModel?: string;
@@ -98,6 +99,7 @@ export interface ResolutionContext {
/** The requirements defined in the registry. */
export interface ResolutionCondition {
useGemini3_1?: boolean;
useGemini3_1FlashLite?: boolean;
useCustomTools?: boolean;
hasAccessToPreview?: boolean;
/** Matches if the current model is in this list. */
@@ -165,6 +167,8 @@ export class ModelConfigService {
switch (key) {
case 'useGemini3_1':
return value === context.useGemini3_1;
case 'useGemini3_1FlashLite':
return value === context.useGemini3_1FlashLite;
case 'useCustomTools':
return value === context.useCustomTools;
case 'hasAccessToPreview':
@@ -58,6 +58,11 @@ describe('NoopSandboxManager', () => {
MY_SECRET: 'super-secret',
SAFE_VAR: 'is-safe',
},
policy: {
sanitizationConfig: {
enableEnvironmentVariableRedaction: true,
},
},
};
const result = await sandboxManager.prepareCommand(req);
@@ -68,7 +73,7 @@ describe('NoopSandboxManager', () => {
expect(result.env['MY_SECRET']).toBeUndefined();
});
it('should NOT allow disabling environment variable redaction if requested in config (vulnerability fix)', async () => {
it('should allow disabling environment variable redaction if requested in config', async () => {
const req = {
command: 'echo',
args: ['hello'],
@@ -85,8 +90,8 @@ describe('NoopSandboxManager', () => {
const result = await sandboxManager.prepareCommand(req);
// API_KEY should be redacted because SandboxManager forces redaction and API_KEY matches NEVER_ALLOWED_NAME_PATTERNS
expect(result.env['API_KEY']).toBeUndefined();
// API_KEY should be preserved because redaction was explicitly disabled
expect(result.env['API_KEY']).toBe('sensitive-key');
});
it('should respect allowedEnvironmentVariables in config but filter sensitive ones', async () => {
@@ -101,6 +106,7 @@ describe('NoopSandboxManager', () => {
policy: {
sanitizationConfig: {
allowedEnvironmentVariables: ['MY_SAFE_VAR', 'MY_TOKEN'],
enableEnvironmentVariableRedaction: true,
},
},
};
@@ -124,6 +130,7 @@ describe('NoopSandboxManager', () => {
policy: {
sanitizationConfig: {
blockedEnvironmentVariables: ['BLOCKED_VAR'],
enableEnvironmentVariableRedaction: true,
},
},
};
@@ -237,6 +237,10 @@
"model": "gemini-3-flash-preview",
"generateContentConfig": {}
},
"chat-compression-3.1-flash-lite": {
"model": "gemini-3.1-flash-lite-preview",
"generateContentConfig": {}
},
"chat-compression-2.5-pro": {
"model": "gemini-2.5-pro",
"generateContentConfig": {}
@@ -237,6 +237,10 @@
"model": "gemini-3-flash-preview",
"generateContentConfig": {}
},
"chat-compression-3.1-flash-lite": {
"model": "gemini-3.1-flash-lite-preview",
"generateContentConfig": {}
},
"chat-compression-2.5-pro": {
"model": "gemini-2.5-pro",
"generateContentConfig": {}
@@ -233,13 +233,19 @@ export {
export function getShellDefinition(
enableInteractiveShell: boolean,
enableEfficiency: boolean,
enableToolSandboxing: boolean = false,
): ToolDefinition {
return {
base: getShellDeclaration(enableInteractiveShell, enableEfficiency),
base: getShellDeclaration(
enableInteractiveShell,
enableEfficiency,
enableToolSandboxing,
),
overrides: (modelId) =>
getToolSet(modelId).run_shell_command(
enableInteractiveShell,
enableEfficiency,
enableToolSandboxing,
),
};
}
@@ -69,7 +69,7 @@ describe('coreTools snapshots for specific models', () => {
{ name: 'list_directory', definition: LS_DEFINITION },
{
name: 'run_shell_command',
definition: getShellDefinition(true, true),
definition: getShellDefinition(true, true, true),
},
{ name: 'replace', definition: EDIT_DEFINITION },
{ name: 'google_web_search', definition: WEB_SEARCH_DEFINITION },
@@ -81,6 +81,7 @@ export function getCommandDescription(): string {
export function getShellDeclaration(
enableInteractiveShell: boolean,
enableEfficiency: boolean,
enableToolSandboxing: boolean = false,
): FunctionDeclaration {
return {
name: SHELL_TOOL_NAME,
@@ -110,35 +111,39 @@ export function getShellDeclaration(
description:
'Set to true if this command should be run in the background (e.g. for long-running servers or watchers). The command will be started, allowed to run for a brief moment to check for immediate errors, and then moved to the background.',
},
[PARAM_ADDITIONAL_PERMISSIONS]: {
type: 'object',
description:
'Sandbox permissions for the command. Use this to request additional sandboxed filesystem or network permissions if a previous command failed with "Operation not permitted".',
properties: {
network: {
type: 'boolean',
description:
'Set to true to enable network access for this command.',
},
fileSystem: {
type: 'object',
properties: {
read: {
type: 'array',
items: { type: 'string' },
description:
'List of additional absolute paths to allow reading.',
},
write: {
type: 'array',
items: { type: 'string' },
description:
'List of additional absolute paths to allow writing.',
...(enableToolSandboxing
? {
[PARAM_ADDITIONAL_PERMISSIONS]: {
type: 'object',
description:
'Sandbox permissions for the command. Use this to request additional sandboxed filesystem or network permissions if a previous command failed with "Operation not permitted".',
properties: {
network: {
type: 'boolean',
description:
'Set to true to enable network access for this command.',
},
fileSystem: {
type: 'object',
properties: {
read: {
type: 'array',
items: { type: 'string' },
description:
'List of additional absolute paths to allow reading.',
},
write: {
type: 'array',
items: { type: 'string' },
description:
'List of additional absolute paths to allow writing.',
},
},
},
},
},
},
},
},
}
: {}),
},
required: [SHELL_PARAM_COMMAND],
},
@@ -332,8 +332,16 @@ export const DEFAULT_LEGACY_SET: CoreToolSet = {
},
},
run_shell_command: (enableInteractiveShell, enableEfficiency) =>
getShellDeclaration(enableInteractiveShell, enableEfficiency),
run_shell_command: (
enableInteractiveShell,
enableEfficiency,
enableToolSandboxing,
) =>
getShellDeclaration(
enableInteractiveShell,
enableEfficiency,
enableToolSandboxing,
),
replace: {
name: EDIT_TOOL_NAME,
@@ -338,8 +338,16 @@ export const GEMINI_3_SET: CoreToolSet = {
},
},
run_shell_command: (enableInteractiveShell, enableEfficiency) =>
getShellDeclaration(enableInteractiveShell, enableEfficiency),
run_shell_command: (
enableInteractiveShell,
enableEfficiency,
enableToolSandboxing,
) =>
getShellDeclaration(
enableInteractiveShell,
enableEfficiency,
enableToolSandboxing,
),
replace: {
name: EDIT_TOOL_NAME,
@@ -37,6 +37,7 @@ export interface CoreToolSet {
run_shell_command: (
enableInteractiveShell: boolean,
enableEfficiency: boolean,
enableToolSandboxing: boolean,
) => FunctionDeclaration;
replace: FunctionDeclaration;
google_web_search: FunctionDeclaration;
+1
View File
@@ -137,6 +137,7 @@ describe('ShellTool', () => {
getShellToolInactivityTimeout: vi.fn().mockReturnValue(1000),
getEnableInteractiveShell: vi.fn().mockReturnValue(false),
getEnableShellOutputEfficiency: vi.fn().mockReturnValue(true),
getSandboxEnabled: vi.fn().mockReturnValue(false),
sanitizationConfig: {},
sandboxManager: new NoopSandboxManager(),
} as unknown as Config;
+2
View File
@@ -696,6 +696,7 @@ export class ShellTool extends BaseDeclarativeTool<
const definition = getShellDefinition(
context.config.getEnableInteractiveShell(),
context.config.getEnableShellOutputEfficiency(),
context.config.getSandboxEnabled(),
);
super(
ShellTool.Name,
@@ -745,6 +746,7 @@ export class ShellTool extends BaseDeclarativeTool<
const definition = getShellDefinition(
this.context.config.getEnableInteractiveShell(),
this.context.config.getEnableShellOutputEfficiency(),
this.context.config.getSandboxEnabled(),
);
return resolveToolDeclaration(definition, modelId);
}
+1 -1
View File
@@ -1,6 +1,6 @@
{
"name": "@google/gemini-cli-devtools",
"version": "0.36.0-nightly.20260317.2f90b4653",
"version": "0.36.0-preview.7",
"license": "Apache-2.0",
"type": "module",
"main": "dist/src/index.js",
+1 -1
View File
@@ -1,6 +1,6 @@
{
"name": "@google/gemini-cli-sdk",
"version": "0.36.0-nightly.20260317.2f90b4653",
"version": "0.36.0-preview.7",
"description": "Gemini CLI SDK",
"license": "Apache-2.0",
"repository": {
+1 -1
View File
@@ -1,6 +1,6 @@
{
"name": "@google/gemini-cli-test-utils",
"version": "0.36.0-nightly.20260317.2f90b4653",
"version": "0.36.0-preview.7",
"private": true,
"main": "src/index.ts",
"license": "Apache-2.0",
+1 -1
View File
@@ -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": "0.36.0-nightly.20260317.2f90b4653",
"version": "0.36.0-preview.7",
"publisher": "google",
"icon": "assets/icon.png",
"repository": {
File diff suppressed because one or more lines are too long