Commit Graph
122 Commits
Author SHA1 Message Date
dd3xp 5887450e6a Merge upstream main into desktop main 2026-07-14 00:47:12 -04:00
Liang Jiaqing 8553e98723 feat: scale tool-output limits with context_win
Add maxlen_multiplier derived from context_win (capped 3x); drives per-tool
truncation caps and history compression interval. WebScan uses half-increment
growth to avoid saturating its 35000 base. Docs updated for the new coupling.
2026-07-13 09:55:52 +08:00
dd3xp dd174cfd01 Merge origin/main into main 2026-07-07 10:37:48 -04:00
Liang Jiaqing b636a8df61 Refine UltraPlan orchestration workflow 2026-06-30 17:22:16 +08:00
Liang Jiaqing 1e4ca151a7 feat: support --history arg and in-memory intervene/keyinfo injection; add export_history; fix GenericAgent typo 2026-06-27 17:10:33 +08:00
Liang Jiaqing 5de55ea872 feat(ultraplan): add phase/parallel/pipeline orchestrator with phase-tree HTML monitor
- assets/ga_ultraplan.py: 3 primitives (phase/parallel/pipeline), file-first
  message passing, runtime f-string prompt fill, subagent via subprocess,
  nested phase tree + random-port auto-popup HTML panel (1h idle self-destruct)
- agentmain.py: --nolog flag
- llmcore.py: honor log suppression
2026-06-26 15:16:11 +08:00
Liang Jiaqing 5bc4945dcb feat: --func pure function mode for subagent; merge bg launch; refine SOP 2026-06-26 12:03:36 +08:00
Liang Jiaqing 99954324bc feat: generic extra_sys_prompts slot replaces hardcoded commit_signature module 2026-06-20 21:03:38 +08:00
Liang Jiaqing df025ab34f feat: loop mode + idle auto-action in stapp, cleanup launch.pyw
- stapp.py: add loop mode (auto-inject prompt after each reply)
- stapp.py: replace DOM-based idle monitor with st.fragment timer
- stapp.py: i18n for all new UI text and LLM prompts
- launch.pyw: remove inject/get_last_reply_time/idle_monitor, keep only paste hook
- agentmain/ga: raise max_turns 80->180, DANGER thresholds accordingly
- remove obsolete tool_usable_history.json
2026-06-19 19:00:44 +08:00
Liang Jiaqing 21158fff43 Update agent docs and task help 2026-06-16 11:27:31 +08:00
Liang Jiaqing 63128dd5bb Fix reflect loop throttling 2026-06-15 10:01:08 +08:00
Jiaqing Liang 0107f6c6ce Update LLM user agent handling and incubator notes 2026-06-08 15:05:47 +08:00
Liang Jiaqing 8ec764a87e Update agent hints and UI SOP note 2026-06-08 09:28:05 +08:00
Liang Jiaqing 3acd58b36e core: per-instance print injection (no_print), EXIT sentinel for run loop, log name collision fix, checklist poll hard cap 2026-06-04 11:09:54 +08:00
Liang Jiaqing f7c6dba679 feat: bbs key模式2天超时自关; reflect首次立即check; worker交付规范; vision_sop修正 2026-05-31 21:53:09 +08:00
Jiaqing Liang db979aafc8 feat: force non-stream mode for reflect/task, fix BaseSession.ask() return type, broadcast stream/read_timeout in MixinSession 2026-05-28 16:15:41 +08:00
Jiaqing Liang 57b71e995c feat: externalize long user prompts (>1500 chars) to temp file 2026-05-28 15:22:34 +08:00
Jiaqing Liang 795d2d51a2 fix: ensure mykey.py is always importable from llmcore
- Add sys.path.append for llmcore's own directory as fallback
- Remove dead variable _imp_err
- Distinguish missing mykey vs import failure with clear error messages
2026-05-21 18:17:53 +08:00
cv70 e0e7487409 feat: add agent lifecycle hook trigger mechanism 2026-05-21 14:33:56 +08:00
Shen Hao 7298758b18 feat(cli): one-line startup banner showing model name (#400)
Display a concise startup banner with model name when running in interactive TTY mode.
2026-05-16 20:30:49 +08:00
Liang Jiaqing 72d28500a3 fix: cancel→abort, cancel_requested guard, wechatapp use turn/outputs instead of _turn_parts 2026-05-16 18:03:51 +08:00
Liang Jiaqing cebd6a37c9 agentloop yield turn dict marker; agentmain intercepts and tracks per-turn outputs 2026-05-16 17:13:42 +08:00
Jiaqing Liang 5f54d4570f feat(reflect): add init() for runtime args with fallback to env/config 2026-05-12 12:59:30 +08:00
Liang Jiaqing 3ddf4cb6f9 Add lightweight goal mode reflect workflow 2026-05-08 17:28:16 +08:00
Liang Jiaqing 0a2be47c00 refactor: simplify LLM client init with resolve_client; add fast_ask; auto-fold old tasks in TUI 2026-05-07 21:06:14 +08:00
Liang Jiaqing 6fd7e2d7a0 feat: add Textual TUI frontend + minor fixes to agentmain/llmcore/stapp 2026-05-07 17:38:58 +08:00
Jiaqing Liang 27428395e2 refactor: rename GeneraticAgent, default bg for task, add fork history support 2026-05-07 15:12:46 +08:00
Liang Jiaqing b3f1b380f3 feat: agent BBS server + team worker + timeout bump 2026-05-06 18:55:18 +08:00
Jiaqing Liang f50b6507e3 Improve English UI defaults and task timeout 2026-05-06 13:23:33 +08:00
Liang Jiaqing c2918a6e87 feat: add contentSettings cmd to CDP bridge (bypass multi-download dialog); minor agentmain cleanup 2026-05-05 11:57:40 +08:00
Liang Jiaqing 1bf2a530b9 /resume prompt优化(v5) + wechatapp去or空占位 + summary截断80 2026-05-03 16:36:48 +08:00
Liang Jiaqing 2eb69983dd feat: peer_hint mechanism & history folding compression 2026-05-02 20:30:33 +08:00
Liang Jiaqing 2e4775f98a refactor: extract _stream_with_retry, MixinSession failover on stream abort 2026-04-28 17:54:37 +08:00
Liang Jiaqing 86ac99c9dc fix: /resume normalize literal \n to real newlines before regex matching 2026-04-28 09:38:04 +08:00
Liang Jiaqing c82df6e347 refactor: 提取script_dir为模块级变量,清理procmem_sop涉敏感代码 2026-04-27 17:46:31 +08:00
Liang Jiaqing e07c1b046b hotreload mykeys, fix SSE residual block, unify context_win, cleanup 2026-04-26 17:46:24 +08:00
VonEquinox 9da2fb8979 fix(cli): 修复中文退格显示残影 2026-04-24 17:21:04 +08:00
Liang Jiaqing 6e68fa41cf refactor: unify system prompt injection into _openai_stream; disable gpt done_hooks 2026-04-23 20:07:20 +08:00
Jiaqing Liang a5aca5988e fix: add requests to pip install & tune max_turns/prompt 2026-04-23 17:21:22 +08:00
Liang Jiaqing 4c1268d821 fix: guard all id/call_id fields with or "" to prevent None leaking into payload 2026-04-23 11:00:04 +08:00
Liang Jiaqing 19057fa9eb improve GPT done_hook: require informative summary on completion 2026-04-22 10:53:29 +08:00
Jiaqing Liang 27203ee5e6 refactor: plugins dir + opt-in langfuse via __getattr__ guard
- mv langfuse_tracing.py -> plugins/langfuse_tracing.py
- llmcore: load plugin lazily inside __getattr__ when langfuse_config present
  (PEP 562 module __getattr__ naturally fires only once after globals().update)
- llmcore: extract _record_usage() from 4 scattered [Cache] print sites
- agentmain: /resume scans only latest 10 files
2026-04-20 15:56:06 +08:00
Liang Jiaqing 777c517949 refactor: update agent loop and tool extraction 2026-04-19 23:21:37 +08:00
Liang Jiaqing 72e1b38877 improve /resume: use regex to extract history blocks instead of naive tail reading 2026-04-19 17:15:38 +08:00
Liang Jiaqing 450edc285c fix: get_llm_name case-insensitive match & refactor to accept optional backend param 2026-04-18 14:46:05 +08:00
Jiaqing Liang 34639c195d refactor: extract slash cmd handler; fix inline_eval param name; support extra_sys_prompt 2026-04-17 15:39:51 +08:00
Liang Jiaqing ca76daddd3 refactor: unify docstrings, add /resume command, simplify error msg 2026-04-16 21:04:04 +08:00
Liang Jiaqing a1d8304457 feat: i18n support - auto-detect system language for zh/en prompts 2026-04-16 18:47:40 +08:00
Liang Jiaqing 16329ffdc1 fix: empty file ZeroDivisionError in file_read; remove SiderLLMSession; init global_mem with header 2026-04-16 16:04:18 +08:00
Jiaqing Liang dc2abca38f preserve history across llm switch 2026-04-15 14:34:06 +08:00