6c94c4d9ca
feat(prompt-optimization): implement multi-objective evaluation metrics
...
Established a Pareto-ready evaluation foundation for the Genetic-Pareto (GEPA)
optimizer, supporting simultaneous optimization of accuracy and density.
Key improvements:
- Core Architecture: Defined standardized `MetricResult` and `OptimizationDirection`
types in `packages/core/src/evals/types.ts` to support multi-objective fitness.
- Centralized Config: Implemented `packages/core/src/evals/config.ts` with tunable
weights and detailed documentation for scoring gradients.
- Tool Alignment Metric: Created `metrics/toolAlignment.ts` to measure functional
accuracy, argument precision, and explicit shell avoidance.
- Token Frugality Metric: Created `metrics/tokenFrugality.ts` to measure and
penalize conversational noise ("chatter") using a configurable threshold.
- Verification Suite: Added comprehensive unit tests for all metrics, achieving
100% coverage of scoring logic and gradient steps.
- Project Integration: Relocated `schema.ts` to the core package for build safety,
updated the data validator, and extended project-wide lint/format scripts.
2026-03-04 10:08:14 -08:00
717660997d
feat(sandbox): add experimental LXC container sandbox support ( #20735 )
2026-03-04 17:44:33 +00:00
bc89b05f01
feat(core): improve @file autocomplete to prioritize filenames ( #21064 )
2026-03-04 17:24:34 +00:00
352fb0c976
fix(mcp): Notifications/tools/list_changed support not working ( #21050 )
...
Co-authored-by: Bryan Morgan <bryanmorgan@google.com >
2026-03-04 09:46:17 -05:00
bbcfff5cf1
chore/release: bump version to 0.34.0-nightly.20260304.28af4e127 ( #21054 )
2026-03-04 05:06:31 +00:00
5bd21a240d
fix: model persistence for all scenarios ( #21051 )
2026-03-04 03:28:12 +00:00
a63c76522a
feat(core): Add tracker CRUD tools & visualization ( #19489 )
...
Co-authored-by: Jerop Kipruto <jerop@google.com >
2026-03-04 00:42:48 +00:00
28af4e127f
fix: merge duplicate imports in packages/core (3/4) ( #20928 )
2026-03-04 00:12:59 +00:00
d25088956d
feat(ui): standardize semantic focus colors and enhance history visibility ( #20745 )
...
Co-authored-by: jacob314 <jacob314@gmail.com >
2026-03-04 00:10:09 +00:00
75737c1b44
chore(release): bump version to 0.34.0-nightly.20260303.34f0c1538 ( #21034 )
2026-03-04 00:08:34 +00:00
066b0cc15f
fix(core): increase A2A agent timeout to 30 minutes ( #21028 )
2026-03-03 22:59:53 +00:00
f3bbe6e77a
fix(core): send shell output to model on cancel ( #20501 )
2026-03-03 22:10:16 +00:00
28e79831ac
fix(core): sanitize and length-check MCP tool qualified names ( #20987 )
2026-03-03 21:38:52 +00:00
f15bcaf499
feat(agents): directly indicate auth required state ( #20986 )
2026-03-03 17:10:12 +00:00
5f2f60bed6
feat(plan): enable built-in research subagents in plan mode ( #20972 )
2026-03-03 16:52:27 +00:00
c332d1e636
feat(extensions): add support for plan directory in extension manifest ( #20354 )
...
Co-authored-by: Mahima Shanware <mshanware@google.com >
Co-authored-by: Jerop Kipruto <jerop@google.com >
2026-03-03 16:50:18 +00:00
aa158e18d3
fix(core): enforce optionality for API response fields in code_assist ( #20714 )
2026-03-03 16:17:34 +00:00
fca29b0bd8
fix(plan): clean up session directories and plans on deletion ( #20914 )
2026-03-03 14:11:25 +00:00
0d69f9f7fa
Build binary ( #18933 )
...
Co-authored-by: Gal Zahavi <38544478+galz10@users.noreply.github.com >
2026-03-03 01:02:19 +00:00
2e7722d6a3
fix(core): restrict "System: Please continue" invalid stream retry to Gemini 2 models ( #20897 )
2026-03-02 23:21:13 +00:00
69e15a50d1
fix(core): skip telemetry logging for AbortError exceptions ( #19477 )
...
Co-authored-by: Yuna Seol <yunaseol@google.com >
2026-03-02 23:14:31 +00:00
3f7ef816f1
fix(core): increase default headers timeout to 5 minutes ( #20890 )
2026-03-02 22:36:58 +00:00
d05ba11a31
refactor(core): replace manual syncPlanModeTools with declarative policy rules ( #20596 )
2026-03-02 22:30:50 +00:00
bb6d1a2775
feat(core): add tool name validation in TOML policy files ( #19281 )
...
Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>
2026-03-02 21:47:21 +00:00
dd9ccc9807
Adding MCPOAuthProvider implementing the MCPSDK OAuthClientProvider ( #20121 )
2026-03-02 21:37:44 +00:00
18d0375a7f
feat(core): support authenticated A2A agent card discovery ( #20622 )
...
Co-authored-by: Adam Weidman <adamfweidman@google.com >
Co-authored-by: Adam Weidman <65992621+adamfweidman@users.noreply.github.com >
2026-03-02 21:29:31 +00:00
b7a8f0d1f9
fix(core): ensure subagents use qualified MCP tool names ( #20801 )
2026-03-02 21:12:13 +00:00
7ca3a33f8b
Subagent activity UX. ( #17570 )
2026-03-02 21:04:31 +00:00
ce5a2d0760
feat(core): truncate large MCP tool output ( #19365 )
2026-03-02 21:01:49 +00:00
659301ff83
feat(core): centralize read_file limits and update gemini-3 description ( #20619 )
2026-03-02 20:11:58 +00:00
446a4316c4
feat(core): implement HTTP authentication support for A2A remote agents ( #20510 )
...
Co-authored-by: Adam Weidman <adamfweidman@google.com >
2026-03-02 19:59:48 +00:00
2e1efaebe4
fix(plan): deflake plan mode integration tests ( #20477 )
2026-03-02 19:51:44 +00:00
7c9fceba7f
fix(core): reduce LLM-based loop detection false positives ( #20701 )
2026-03-02 19:08:15 +00:00
740efa2ac2
Merge User and Agent Card Descriptions #20849 ( #20850 )
2026-03-02 17:59:29 +00:00
a153ff587b
refactor(core): Extract tool parameter names as constants ( #20460 )
2026-02-28 21:27:54 +00:00
cd3a8c3f07
fix(cli): reset themeManager between tests to ensure isolation ( #20598 )
2026-02-28 19:45:31 +00:00
b2214a6676
fix: acp/zed race condition between MCP initialisation and prompt ( #20205 )
...
Signed-off-by: Kartik Angiras <angiraskartik@gmail.com >
2026-02-28 17:33:08 +00:00
fb6ff847dd
chore/release: bump version to 0.33.0-nightly.20260228.1ca5c05d0 ( #20644 )
2026-02-28 02:13:48 +00:00
0c6c9c6a62
chore(release): bump version to 0.33.0-nightly.20260227.ba149afa0 ( #20637 )
2026-02-28 00:51:22 +00:00
a1367e9cdd
fix(core): parse raw ASCII buffer strings in Gaxios errors ( #20626 )
2026-02-27 23:57:32 +00:00
ba149afa0b
fix: merge duplicate imports in a2a-server package (2/4) ( #19781 )
2026-02-27 21:13:30 +00:00
966b9059d0
feat(core): enable contiguous parallel admission for Kind.Agent tools ( #20583 )
2026-02-27 21:08:10 +00:00
20d884da2f
fix(core): reduce intrusive MCP errors and deduplicate diagnostics ( #20232 )
2026-02-27 20:04:36 +00:00
ea48bd9414
feat: better error messages ( #20577 )
...
Co-authored-by: Gal Zahavi <38544478+galz10@users.noreply.github.com >
2026-02-27 18:18:16 +00:00
b2d6844f9b
feat(billing): implement G1 AI credits overage flow with billing telemetry ( #18590 )
2026-02-27 18:15:06 +00:00
fdd844b405
fix(core): disable retries for code assist streaming requests ( #20561 )
2026-02-27 18:04:43 +00:00
23905bcd77
fix(plan): prevent agent from using ask_user for shell command confirmation ( #20504 )
2026-02-27 17:51:47 +00:00
e709789067
fix(core): handle optional response fields from code assist API ( #20345 )
2026-02-27 16:52:37 +00:00
32e777f838
fix(core): revert auto-save of policies to user space ( #20531 )
2026-02-27 16:03:36 +00:00
d7320f5425
refactor(core,cli): useAlternateBuffer read from config ( #20346 )
...
Co-authored-by: Jacob Richman <jacob314@gmail.com >
2026-02-27 15:55:02 +00:00
Abhijit Balaji