04cb and Xubin Ren
e3df310309
fix(heartbeat): skip when HEARTBEAT.md has no tasks and fail closed on delivery ( #4111 )
2026-05-31 15:15:37 +08:00
04cb and Xubin Ren
e29c9c3906
fix(agent): acquire per-session lock in process_direct ( #4080 )
2026-05-30 23:45:37 +08:00
04cb and Xubin Ren
13ab092cea
feat(dream): add enabled toggle to skip Dream job registration ( #3885 )
2026-05-30 15:35:04 +08:00
04cb and Xubin Ren
5fe57f8afa
fix(providers): coerce typeless Anthropic content blocks to text ( #3993 )
2026-05-30 15:35:04 +08:00
04cb and Xubin Ren
9d3fe7c34b
fix(providers): surface clear arrearage warning on quota/billing errors ( #3006 )
2026-05-29 15:31:17 +08:00
04cb and Xubin Ren
7a6cc657db
feat(spawn): allow per-subagent sampling temperature ( #3969 )
2026-05-24 13:54:37 +08:00
04cb and Xubin Ren
ef2ef4f789
fix(transcription): normalize chat-style apiBase to audio endpoint ( #3637 )
2026-05-23 17:32:59 +08:00
04cb and Xubin Ren
5b71f61f55
fix(exec): uncap config exec timeout; 0 means no limit ( #3595 )
2026-05-23 17:32:59 +08:00
04cb and Xubin Ren
bd0ba745dd
fix(wecom): preserve real filename from SDK when payload omits name ( #3737 )
2026-05-12 10:27:32 +08:00
04cb and Xubin Ren
9d6afd86b5
fix(provider): backfill DeepSeek reasoning_content instead of dropping history ( #3554 , #3584 )
2026-05-04 12:14:38 +08:00
04cb and Xubin Ren
c15d816d9c
fix(cli): intercept _retry_wait so provider retry messages don't garble interactive output ( #3600 )
2026-05-03 22:59:08 +08:00
04cb and Xubin Ren
651b6b933f
fix(feishu): streaming card and tool hint respect reply_to_message in groups
2026-04-30 12:51:08 +08:00
04cb and Xubin Ren
9239429a00
fix(anthropic): omit temperature for opus-4-7 ( #3417 )
2026-04-24 15:33:59 +08:00
04cb and Xubin Ren
c27b4d07c4
fix(utils): recurse into PPTX groups and tables when extracting text ( #3250 )
2026-04-18 12:30:42 +08:00
04cb and Xubin Ren
eacc9fbb5f
refactor(providers): drop unreachable GenerationSettings fallback
2026-04-15 23:52:38 +08:00
04cb and Xubin Ren
54f7ad3752
fix(providers): guard chat_with_retry against explicit None max_tokens ( #3102 )
2026-04-15 23:52:38 +08:00
04cb and Xubin Ren
3f59bd1443
fix(shell): reject LLM-supplied working_dir outside workspace ( #2826 )
2026-04-12 16:38:55 +08:00
04cb and Xubin Ren
00fb491bc9
fix(shell): block exec writes to history.jsonl and cursor files ( #2989 )
2026-04-12 16:38:55 +08:00
04cb
4cd4ed8ada
fix(agent): preserve tool results on fatal error to prevent orphan tool_calls ( #2943 )
2026-04-11 21:50:44 +08:00
04cb and Xubin Ren
e392c27f7e
fix(utils): anchor unclosed think-tag regex to string start ( #3004 )
2026-04-11 13:46:15 +08:00
04cb and Xubin Ren
f4904c4bdf
fix(cron): add optional name parameter to separate job label from message ( #2680 )
2026-04-07 13:22:20 +08:00
04cb and Xubin Ren
5f08d61d8f
fix(security): add ssrfWhitelist config to unblock Tailscale/CGNAT ( #2669 )
2026-04-04 19:43:18 +08:00
04cb and Xubin Ren
929ee09499
fix(utils): ensure reasoning_content present with thinking_blocks ( #2579 )
2026-03-31 11:49:23 +08:00
04cb and Xubin Ren
3f21e83af8
fix(tools): clarify cron message param as agent instruction ( #2566 )
2026-03-31 11:49:23 +08:00
04cb and Xubin Ren
8682b017e2
fix(tools): add Accept header for MCP SSE connections ( #2651 )
2026-03-31 11:49:23 +08:00
04cb
e381057356
Fix tool_call_id length error for GitHub Copilot provider
...
GitHub Copilot and some other providers have a 64-character limit on
tool_call_id. When switching from providers that generate longer IDs
(such as OpenAI Codex), this caused validation errors.
This fix truncates tool_call_id to 64 characters by preserving the first
32 and last 32 characters to maintain uniqueness while respecting the
provider's limit.
Fixes #1554
2026-03-07 08:31:15 +08:00
04cbandXubin Ren