What's Changed

• feat : add Anthropic thinking support by @fengju0213 in #4033

Full Changelog: v0.2.91a3...v0.2.91a4

What's Changed

• feat: Support DeepSeek V4 thinking tool calls by @fengju0213 in #4026
• chore: support GPT 5.5, update doc by @Wendong-Fan in #4029

Full Changelog: v0.2.91a2...v0.2.91a3

What's Changed

• feat(toolkits): add Querit search API integration to SearchToolkit by @littledoglily in #4000
• fix: pre commit error by @Wendong-Fan in #4021
• add new claude models by @YunfeiZHAO in #4023
• fix: keep planning worktree state until remove succeeds by @shaun0927 in #4017
• enhance: PlanningWorktreeToolkit clears worktree state before remove … by @Wendong-Fan in #4024
• fix: handle empty xAI streams without crashing by @shaun0927 in #4015
• fix: preserve response-mode session scoping without Langfuse by @shaun0927 in #4016
• fix: simplify anthropic structured output flow by @fengju0213 in #4003

New Contributors

• @littledoglily made their first contribution in #4000
• @shaun0927 made their first contribution in #4017

Full Changelog: v0.2.91a1...v0.2.91a2

What's Changed

• feat: Support xAI native client and openai compatible response client by @Wendong-Fan in #4010

Full Changelog: v0.2.91a0...v0.2.91a1

What's Changed

• fix: preserve missing parameters in ChatAgent.clone() by @majiayu000 in #3967
• fix: ChatAgent clone stream accumulate passing by @Wendong-Fan in #3970
• fix: extract usage information from Anthropic response and messages by @Zephyroam in #4006

New Contributors

• @majiayu000 made their first contribution in #3967

Full Changelog: v0.2.90...v0.2.91a0

What's Changed

• chore: adjust workforce quality check threadhold to 60 by @Wendong-Fan in #3839
• chore: fix pre-commit error by @Wendong-Fan in #3840
• feat: Interleaved thinking for kimi k2 and glm by @Zephyroam in #3737
• feat: support minimax m2.5 and glm5 by @Wendong-Fan in #3841
• fix: correct typo 'seperated' to 'separated' by @thecaptain789 in #3791
• chore: update contributing guideline and PR template for AI code rule by @Wendong-Fan in #3844
• enhance: throw runtime error on absent ffmpeg binary by @a7m-1st in #3850
• fix: capture rich output from e2b execution results by @yaowubarbara in #3847
• feat: add skill filter by @fengju0213 in #3843
• fix: mistral and sambanova model tool calling by @Wendong-Fan in #3862
• chore : Update skill prompt in workforce by @fengju0213 in #3856
• fix: ffmpg dependency issue in cicd test by @Wendong-Fan in #3865
• refactor: model log system include tool schema and support multi-ling… by @Wendong-Fan in #3866
• fix: anthropic max_tokens limitation by @Wendong-Fan in #3874
• feat(oceanbase): add advanced ANN query options (#3649) by @YixinZ-NUS in #3758
• Bump pyobvector version to 0.2.22 by @zhangjiongai-debug in #3757
• Add Eigent as first ecosystem highlight by @Copilot in #3798
• feat: add Avian as a new LLM provider by @avianion in #3881
• fix: fix function param type by @Benjmmi in #3499
• feat:add_request_level_token_callback by @fengju0213 in #3872
• release: v0.2.90a5 by @Wendong-Fan in #3896
• chore: remove default 180s timeout threshold by @nitpicker55555 in #3898
• feat: add headless browser search toolkit & enhance browser stealth by @nitpicker55555 in #3895
• feat: add gemini 3.1 to model enum list by @Wendong-Fan in #3903
• feat: support response api by @fengju0213 in #3814
• feat: Extend Prompt Caching Support to More Providers by @Zephyroam in #3901
• enhance: TerminalToolkit safe-mode sanitization by @Wendong-Fan in #3873
• Add gpt-5.4 by @zechengz in #3907
• fix: use partition_html for URL parsing in UnstructuredIO by @Wendong-Fan in #3908
• fix: add botocore to mypy ignore list by @Wendong-Fan in #3909
• feat(workforce): unify streaming response handling and callback integration by @coolbeevip in #3744
• fix(commons): use correct env var for API key URL in api_keys_required error by @jnMetaCode in #3913
• fix(github_toolkit): pass repo_name in recursive github_get_all_file_paths call by @jnMetaCode in #3912
• chore: fix pre commit issue by @Wendong-Fan in #3926
• chore: update qianfan_model schema by @fengju0213 in #3925
• feat: auto-coerce dict args to Pydantic models in FunctionTool by @Wendong-Fan in #3916
• Async support for aws bedwork converse api by @Zephyroam in #3922
• release: v0.2.90a6 by @Wendong-Fan in #3932
• fix: SSE fallback + robust connect for Harbor-spec HTTP MCP sidecars by @nitpicker55555 in #3934
• Check terminal toolkit initial environment setting by @Abishek-Newar in #3818
• chore: fix precommit by @fengju0213 in #3943
• feat(models): add MiniMax-M2.7 and M2.7-highspeed model support by @octo-patch in #3948
• docs: update camel doc and build cicd by @fengju0213 in #3885
• fix: merge consecutive toolResult blocks into single user message by @Wendong-Fan in #3957

New Contributors

• @thecaptain789 made their first contribution in #3791
• @yaowubarbara made their first contribution in #3847
• @YixinZ-NUS made their first contribution in #3758
• @zhangjiongai-debug made their first contribution in #3757
• @Copilot made their first contribution in #3798
• @avianion made their first contribution in #3881
• @Benjmmi made their first contribution in #3499
• @jnMetaCode made their first contribution in #3913
• @Abishek-Newar made their first contribution in #3818
• @octo-patch made their first contribution in #3948

Full Changelog: v0.2.89...v0.2.90

What's Changed

• chore: remove default 180s timeout threshold by @nitpicker55555 in #3898
• feat: add headless browser search toolkit & enhance browser stealth by @nitpicker55555 in #3895
• feat: add gemini 3.1 to model enum list by @Wendong-Fan in #3903
• feat: support response api by @fengju0213 in #3814
• feat: Extend Prompt Caching Support to More Providers by @Zephyroam in #3901
• enhance: TerminalToolkit safe-mode sanitization by @Wendong-Fan in #3873
• Add gpt-5.4 by @zechengz in #3907
• fix: use partition_html for URL parsing in UnstructuredIO by @Wendong-Fan in #3908
• fix: add botocore to mypy ignore list by @Wendong-Fan in #3909
• feat(workforce): unify streaming response handling and callback integration by @coolbeevip in #3744
• fix(commons): use correct env var for API key URL in api_keys_required error by @jnMetaCode in #3913
• fix(github_toolkit): pass repo_name in recursive github_get_all_file_paths call by @jnMetaCode in #3912
• chore: fix pre commit issue by @Wendong-Fan in #3926
• chore: update qianfan_model schema by @fengju0213 in #3925
• feat: auto-coerce dict args to Pydantic models in FunctionTool by @Wendong-Fan in #3916
• Async support for aws bedwork converse api by @Zephyroam in #3922
• release: v0.2.90a6 by @Wendong-Fan in #3932

New Contributors

• @jnMetaCode made their first contribution in #3913

Full Changelog: v0.2.90a5...v0.2.90a6

What's Changed

• feat(oceanbase): add advanced ANN query options (#3649) by @YixinZ-NUS in #3758
• Bump pyobvector version to 0.2.22 by @zhangjiongai-debug in #3757
• Add Eigent as first ecosystem highlight by @Copilot in #3798
• feat: add Avian as a new LLM provider by @avianion in #3881
• fix: fix function param type by @Benjmmi in #3499
• feat:add_request_level_token_callback by @fengju0213 in #3872
• release: v0.2.90a5 by @Wendong-Fan in #3896

New Contributors

• @YixinZ-NUS made their first contribution in #3758
• @zhangjiongai-debug made their first contribution in #3757
• @Copilot made their first contribution in #3798
• @avianion made their first contribution in #3881
• @Benjmmi made their first contribution in #3499

Full Changelog: v0.2.90a4...v0.2.90a5

What's Changed

• fix: anthropic max_tokens limitation by @Wendong-Fan in #3874

Full Changelog: v0.2.90a3...v0.2.90a4

What's Changed

• refactor: model log system include tool schema and support multi-ling… by @Wendong-Fan in #3866

Full Changelog: v0.2.90a2...v0.2.90a3