Skip to content
Open
Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
292 commits
Select commit Hold shift + click to select a range
cfc6401
docs(swarm): require leader-written swarmflow scripts under team work…
Shuo93 Jul 6, 2026
7e12bd2
feat(swarm): unify external CLI inbound rendering to XML and relocate…
Shuo93 Jul 6, 2026
2a2b0e9
feat(swarm): unify member roster and make team sections static
Shuo93 Jul 6, 2026
7bfcf2c
feat(swarm): wake teammate only on claimable growth; emit TASK_RELEASED
Shuo93 Jul 7, 2026
203dfaf
feat(swarm): make stale-task nudge self-only, minimal, and appended
Shuo93 Jul 7, 2026
0a19ca0
fix(swarm): retry abnormally-dead rounds and add stall watchdogs
Shuo93 Jul 7, 2026
16779cd
feat(swarm): reassign via TASK_REVOKED event and cap one active claim…
Shuo93 Jul 7, 2026
fba98e2
test(swarm): cover one-active-claim guard and revoke-based reassignment
Shuo93 Jul 7, 2026
e4ea618
docs(swarm): archive F_54 and sync S_03/S_08 for revoke-based reassign
Shuo93 Jul 7, 2026
91e2386
docs(swarm): require reading S_/F_ docs before designing, not just at…
Shuo93 Jul 8, 2026
c7e4a46
feat(swarm): atomic create_task graph + narrow one-active-claim probe
Shuo93 Jul 8, 2026
a405683
feat(swarm): rank swarmflow triggers by priority, force on keyword
Shuo93 Jul 8, 2026
94d5277
perf(swarm): skip team-wide cycle scan when a mutation adds no edges
Shuo93 Jul 8, 2026
107b84c
feat(swarm): atomic reassign and task-level cancel/edit signals repla…
Shuo93 Jul 8, 2026
543edef
test(swarm): cover atomic reassign, targeted cancel/edit signals, and…
Shuo93 Jul 8, 2026
f598910
docs(swarm): archive F_56 and sync S_03/S_08 for task-level signals
Shuo93 Jul 8, 2026
4cdca57
feat(swarm): condition-named task state machine and dispatch-mode too…
Shuo93 Jul 9, 2026
9388e07
test(swarm): cover condition-named state machine and tool variants
Shuo93 Jul 9, 2026
bc7c269
docs(swarm): archive F_57/F_59, sync S_08/S_12 for the state machine
Shuo93 Jul 9, 2026
8ea846e
feat(swarm): add the verify gate — reviewer column and verify_task
Shuo93 Jul 9, 2026
6254425
test(swarm): cover the verify gate, reviewer column and migration
Shuo93 Jul 9, 2026
f25e23e
docs(swarm): finalize F_59 Phase 2, sync S_08/S_12 for the verify gate
Shuo93 Jul 9, 2026
e05492a
feat(swarm): pause/abort at inner react iteration boundaries, add resume
Shuo93 Jul 9, 2026
51b0dab
test(swarm): cover phase-aware pause/abort, warm resume, continuation
Shuo93 Jul 9, 2026
65813f1
docs(swarm): archive F_60, sync S_18 and AGENTS for pause/abort/resume
Shuo93 Jul 9, 2026
59ab99e
feat(swarm): cold-resume a paused round across stop/start
Shuo93 Jul 9, 2026
48932f5
test(swarm): cover pending-resume marker, cold resume and its query
Shuo93 Jul 9, 2026
0d3033c
docs(swarm): archive F_61, sync S_18 and AGENTS for cold resume
Shuo93 Jul 9, 2026
1dbff53
feat(swarm): scheduled dispatch runtime with review voting
Shuo93 Jul 10, 2026
deb813c
test(swarm): cover the scheduler, review voting and dispatch config
Shuo93 Jul 10, 2026
dbe188a
docs(swarm): archive F_62, add S_22 and sync specs for scheduled disp…
Shuo93 Jul 10, 2026
f8dd124
refactor(swarm): resolve static-check findings in locales, rail and DAOs
Shuo93 Jul 14, 2026
71a8787
feat(swarm): render scheduler handoffs from templates at delivery time
Shuo93 Jul 14, 2026
9f20d61
test(swarm): cover message templating, meta storage and delivery expa…
Shuo93 Jul 14, 2026
d5a1951
docs(swarm): archive F_63 and sync S_12/S_22 for message templating
Shuo93 Jul 14, 2026
86cc38d
fix(swarm): let the leader shut down human members like any teammate
Shuo93 Jul 14, 2026
6a774d8
feat(monitor): forward new task lifecycle events to team channels
Shuo93 Jul 15, 2026
6324592
feat(swarm): route messages by content shape and cap send_message con…
Shuo93 Jul 16, 2026
593833f
feat(swarm): detect member stalls from a runtime idle clock
Shuo93 Jul 16, 2026
60208ba
fix(swarm): resume a paused round from kernel.start, not notify_team_…
Shuo93 Jul 16, 2026
73d53aa
fix(evolution): skill evolution score update and two-stage optimizer …
Jul 16, 2026
1d54561
feat(agent_teams): use Claude Agent SDK backend
ZhCoding Jul 13, 2026
74789a9
refactor(agent_teams): remove legacy Claude CLI adapter
ZhCoding Jul 13, 2026
adda0c6
refactor(agent_teams): remove generic ssh transport
ZhCoding Jul 13, 2026
191666c
refactor(agent_teams): expose cli runtime base
ZhCoding Jul 13, 2026
56ae078
fix(agent_teams): recognize sdk cli backends
ZhCoding Jul 14, 2026
72b482f
fix(agent_teams): avoid claude ssh read close race
ZhCoding Jul 14, 2026
5dd2c23
fix(agent_teams): stop external runtime on spawn cancel
ZhCoding Jul 14, 2026
5937679
feat(agent_teams): resume claude external backend sessions
ZhCoding Jul 14, 2026
b8477c5
test(agent_teams): add external claude pause resume e2e
ZhCoding Jul 14, 2026
958ca23
fix(agent_teams): clean external cli backend sessions
ZhCoding Jul 15, 2026
561bc22
fix(agent_teams): close messager on coordination stop
ZhCoding Jul 16, 2026
e18a44d
fix(agent_teams): align external CLI startup contracts
ZhCoding Jul 16, 2026
d13c543
test(agent_teams): fix Claude SDK runtime database config
ZhCoding Jul 16, 2026
e582ab2
fix(agent_teams): address external CLI quality warnings
ZhCoding Jul 16, 2026
8ce2336
!1952 fix(agent_teams): recognize sdk cli backends
ZhCoding Jul 17, 2026
13ec360
feat(harness): adding spec-driven deepagent
maggiezhangshere Jul 17, 2026
2675958
fix: isolate sandbox_id from different agent
MarcoPoloPie Jul 17, 2026
9c39e5d
fix(context):remove chinese query translator
yan-zj722 Jul 18, 2026
fed34fb
fix(fs_operation):There is a memory leak issue with the file read/wri…
Jul 20, 2026
d34a9d4
fix(agent_evolving): validate skill package install paths
z2tng Jul 20, 2026
ce52bbe
fix(evolving): refine skill creation and evolution triggers
z2tng Jul 20, 2026
ba0f9f7
feat(harness): Session-Scoped Supports Goal Execution
Li-111 Jul 20, 2026
6c30418
fix(skill_rail): persist session baseline and track runtime changes
lindada11 Jul 20, 2026
5aba739
feat(agent_teams): sync external member team context
ZhCoding Jul 17, 2026
eb4843f
feat(agent_teams): resolve external claude cwd
ZhCoding Jul 18, 2026
9e443f8
feat(agent_teams): align external teammate workspace context
ZhCoding Jul 20, 2026
ae90106
feat(agent_teams): expose workspace meta to external members
ZhCoding Jul 20, 2026
18b80c3
fix(agent_teams): add empty schema for async task list tool
ZhCoding Jul 20, 2026
1498bba
fix(evolving): stabilize skill evolution runtime flow
z2tng Jul 21, 2026
f906c2f
fix(harness):1.Fix concurrent edit overwrite issue. 2.Fix chunked ful…
yan-zj722 Jul 21, 2026
940b5ed
fix(memory): resolve BM25 retrieval tokenizer defects
zhouziliang1015 Jul 21, 2026
be4a93b
[mirror] openJiuwen-ai/agent-core#71: fix(browser subagent): fix brow…
Jul 21, 2026
82542b4
fix(agent): main agent and subagent share artifact directory
Cyucong Jul 22, 2026
1f22e17
fix(operator): close rw lock probe before deleting database
Jul 22, 2026
5b0ed53
feat(swarmflow): SwarmFlow human-in-the-loop and run-scoped reply rou…
chenchunzhou1995 Jul 22, 2026
3027726
fix(ascend affinity): integrate AscendAffinity KV Cache affinity mana…
JuneHM Jul 22, 2026
4d3299d
fix(swarm): declare an object schema for async_tasks_list
Shuo93 Jul 17, 2026
72b7710
feat(swarm): enforce swarmflow token budget from real model usage
Shuo93 Jul 17, 2026
f686145
test(swarm): cover the swarmflow token budget end to end
Shuo93 Jul 17, 2026
9392229
docs(swarm): archive F_66 and sync the budget contract
Shuo93 Jul 17, 2026
3f267aa
chore: ignore harness workspace scaffolding written by unit tests
Shuo93 Jul 17, 2026
beb8efa
feat(swarm): add intelli_router model allocation strategy
Shuo93 Jul 17, 2026
a50210e
test(swarm): unify system-test model endpoints behind one config
Shuo93 Jul 17, 2026
bc740a1
test(swarm): move the swarmflow E2E suite into its own directory
Shuo93 Jul 17, 2026
1cd0dc1
fix:使用浏览器工具检索之后,chrome窗口没有自动关闭,手动关闭导致后台僵尸进程
Jul 23, 2026
96f0a33
fix(team):Add checkpoint to the dispose method of team_harness
wch1135361471 Jul 23, 2026
b7341ac
feat(agent-teams): integrate Codex Python SDK backend
Mavs-32 Jul 23, 2026
d17a91b
fix(context_engine): Add exception handling as a fallback in the infi…
Li-111 Jul 24, 2026
65f6f18
fix(security): allow SSL verification without custom cert
Jul 24, 2026
9850da1
fix(harness): Optimize the effectiveness of the goal function
Li-111 Jul 24, 2026
63f6359
fix(agent-teams): add public TeamMonitor.get_task(task_id) query API
chenchunzhou1995 Jul 24, 2026
472c63a
fix(filesystem): avoid duplicate PIL decode/base64 encode in _read_image
chenran0522 Jul 24, 2026
95f333c
docs(openai-auth): add OpenAIAccount OAuth provider guide and API docs
chenran0522 Jul 24, 2026
999c1e0
fix(skill): add checker to SkillTool's relative_file_path
Jul 24, 2026
28f9627
feat(agent-teams): publish hybrid events via team.mq.publish
Jul 25, 2026
f9fcca5
fix(browser subagent): add reset_active_browser_runtimes function
ShowHsiang Jul 25, 2026
16bfc0b
fix(agent-teams): separate task failures from runtime crashes
tangfei0306 Jul 25, 2026
055176e
fix(context): When canceled, save the message history
Li-111 Jul 25, 2026
f74d8e3
fix(agent_teams): retry on pool-exhaustion timeout in retry_on_locked
ZhCoding Jul 25, 2026
63a6d21
fix(tool): default wake_offset_seconds to 0 when omitted
Li-111 Jul 27, 2026
7ded9cf
fix(harness): remove unsupported regex lookarounds from worktree schema
LasVie Jul 27, 2026
026acd8
fix(context): Continuously invoking the same tool triggers compressio…
Li-111 Jul 27, 2026
b7b65f9
fix(llm): The model enhances pooling capability and reduces call latency
Li-111 Jul 27, 2026
66fbb4f
fix(trace): remove changing llm inputs&outputs to str in otel_rail
huaweizyl Jul 27, 2026
3475c9d
fix(context): switch ContextProcessorRail presets to forked processors
lindada11 Jul 28, 2026
3bb4750
[mirror] openJiuwen-ai/agent-core#127: fix(openai-auth): retry transi…
LasVie Jul 28, 2026
3604368
fix(agent_teams): support assigned autonomous tasks
ZhCoding Jul 28, 2026
73de83b
fix(browser): restore tool result windowing in browser subagent
lindada11 Jul 28, 2026
0093eb0
fix(observability): reasoning span duration + reasoning/cache token c…
hwxiajing Jul 28, 2026
204c8f0
fix(agent-team): rollback dependency agent-protocol way
tangfei0306 Jul 28, 2026
f40aa05
fix(llm): All think information of the MiniMax-M3 model is displayed …
Jul 28, 2026
ad0d331
fix(agent_teams): ignore shutdown members in unread completion gate
ZhCoding Jul 28, 2026
cbfeb9c
fix(agent-team): rollback dependency agent-protocol way
tangfei0306 Jul 28, 2026
30c45fe
fix(sub_agent): The description of input parameters misleads the model
Li-111 Jul 28, 2026
3bb4ae5
fix(goal): Persistently add time data to the goal
Li-111 Jul 28, 2026
b5736c5
fix(swarm): correct send_message comms policy for teammate vs leader
Shuo93 Jul 23, 2026
3cbde10
fix(swarm): harden teammate ban on conveying messages via plain text
Shuo93 Jul 24, 2026
28013cd
fix(swarm): make the user role explicit to every team member
Shuo93 Jul 28, 2026
69f9aef
fix(code): fix code tool description
wenliH Jul 28, 2026
1627d34
[mirror] openJiuwen-ai/agent-core#119: mango merge
LYA2333 Jul 28, 2026
2d9c1a3
[mirror] openJiuwen-ai/agent-core#91: fix(tool_call): The call does n…
yuxiaogou Jul 28, 2026
04e2a9a
fix: add support for yuanrong sandbox
MarcoPoloPie Jul 28, 2026
f9262ae
fix(agent_teams): start human-assigned ready tasks
ZhCoding Jul 28, 2026
5d82d89
feat(context): add asynchronous session memory processor
lindada11 Jul 28, 2026
a3518ff
fix(swarm): skip observability rail when the extra is not installed
Shuo93 Jul 28, 2026
e3a0b47
feat(swarm): move member identity out of the shared prompt prefix
Shuo93 Jul 28, 2026
35fc949
feat(swarm): cover assigned-but-unstarted tasks in autonomous nudging
Shuo93 Jul 28, 2026
15f834c
test(swarm): cover startup board survey and backlog stall sweep
Shuo93 Jul 28, 2026
9556530
docs(swarm): archive F_69, sync S_03 / S_12 and coordination AGENTS
Shuo93 Jul 28, 2026
59389c7
docs(swarm): document handler decision policy in handlers/AGENTS.md
Shuo93 Jul 28, 2026
7fadda3
feat(swarm): separate cwd, workspace and project root
Shuo93 Jul 28, 2026
bf8a3b8
test(swarm): cover cwd/workspace separation for members and workers
Shuo93 Jul 28, 2026
adbbf1c
docs(swarm): archive F_69 and rewrite the workspace/worktree invariants
Shuo93 Jul 28, 2026
94dcf4e
feat: Backport the security guardrail features from the enterprise br…
yyuse Jul 29, 2026
e009361
feat(symphony): Add new Symphony module and README.md
xiao-yu-lin Jul 29, 2026
e81a86c
feat(agent_teams): normalize external tool call chunks
ZhCoding Jul 29, 2026
f0adc83
[mirror] openJiuwen-ai/agent-core#152: feat(example):merge perstream
Jul 29, 2026
e6fd940
feat(examples): add adaptive multi-agent collaboration example
Jul 29, 2026
aaad809
fix(harness): add timeout for modality probe
ubaidalih Jul 29, 2026
532499a
feat(agent_teams): preserve claude tool result names
ZhCoding Jul 29, 2026
0f08038
fix(prompts): add type= mapping in teammate communication protocol
hwxiajing Jul 29, 2026
3e91d10
fix(team): skip foreign session worktree recovery
Jul 29, 2026
f43d827
fix:openai preserve real finish_reason
Jul 29, 2026
a5a0a60
fix(harness): switch default processor preset to forked and enable co…
lindada11 Jul 30, 2026
79d43ac
fix(core): enable skill_tool to return directory tree for agent sub-s…
wenliH Jul 30, 2026
53426e0
feat(agent-teams): Reproducibility test case supplement
Jul 30, 2026
a8a231e
feat(memory): auto-load today/yesterday daily_memory into system prom…
HuskyJs Jul 30, 2026
40fceda
fix(tools): Refine todo prompts to prevent over-planning and enforce …
Li-111 Jul 30, 2026
b8e3929
fix(prompts): guard human_agent shutdown against active tasks
hwxiajing Jul 30, 2026
448c9b9
doc(operator): Reproducibility test case supplement
Jul 30, 2026
c07b849
fix(observability): add gen_ai.prompt.N.tool_calls attrs for assistan…
hwxiajing Jul 30, 2026
e471f47
fix(agent_teams): add team existence check in spawn_member
Little-Shepherd Jul 30, 2026
fdd8c3d
fix(swarm): access team_name through blueprint.spec in scheduler acti…
Little-Shepherd Jul 30, 2026
00d373a
feat(session): Add session_id field to tracer
Li-111 Jul 30, 2026
013adce
fix(mcp): Resolve the issue where, in SSE mode, if the TCP connection…
yuxiaogou Jul 30, 2026
967757d
fix(swarm): echo summary in send_message map_result to prevent duplic…
hwxiajing Jul 30, 2026
842aa0b
feat(agent_teams): use in-process MCP for Claude
ZhCoding Jul 30, 2026
ec86797
docs(skill-evolution): align API references and navigation
z2tng Jul 31, 2026
683f28f
feat(agent-teams): add Codex OpenTelemetry observability
Mavs-32 Jul 31, 2026
6bddde3
update: 更新文件 README.zh.md
tlk1997 Jul 31, 2026
5ac38dd
update: 更新文件 README.md
tlk1997 Jul 31, 2026
40ce1c7
fix(context): save cleaned context to session state on cancel
lindada11 Jul 31, 2026
7c73dbc
fix(agent_teams): reject leader messages to user
ZhCoding Jul 31, 2026
de55bc7
feat(tiny-agent): make SecurityRail opt-in
Jul 31, 2026
425b07d
feat: change reviewer startup logic to use temporary harness, and upd…
Aug 1, 2026
613acc5
feat(swarm): deliver team context into member conversations
Shuo93 Jul 29, 2026
1c49ffd
test(swarm): cover team context delivery and roster deltas
Shuo93 Jul 29, 2026
9ffd139
docs(swarm): archive F_70 and revise the team context contract
Shuo93 Jul 29, 2026
d488397
feat(core): add ON_USER_MESSAGE rail hook for consumed inputs
Shuo93 Jul 29, 2026
a494e7d
fix(swarm): deliver team context on the admitted input
Shuo93 Jul 29, 2026
dddf446
test(swarm): cover the user-message hook and team context delivery
Shuo93 Jul 29, 2026
0ec2884
docs(swarm): record the input-hook delivery contract in F_70 and S_09
Shuo93 Jul 29, 2026
f827729
fix(swarm): tell each member its own workspace path
Shuo93 Jul 29, 2026
a4a4ef1
feat(swarm): hand rails the input batch and drop superseded boards
Shuo93 Jul 31, 2026
38bc610
test(swarm): cover input batches and superseded board dropping
Shuo93 Jul 31, 2026
1bdc222
docs(swarm): archive F_71 and revise the input batch contracts
Shuo93 Jul 31, 2026
300a3dc
feat(swarm): nest team notes inside the block they annotate
Shuo93 Jul 31, 2026
bdb74a0
fix(harness): keep prompt sections and rail tool ownership from clobb…
Shuo93 Aug 1, 2026
ff5afbb
perf(harness): stop paying for probes, locks and log payloads on hot …
Shuo93 Aug 1, 2026
b6d42af
chore(core): report slow rail chains and invoke preparation
Shuo93 Aug 1, 2026
891843e
refactor(harness): order rail init by priority too
Shuo93 Aug 1, 2026
e90baed
chore(core): gate per-chunk stream tracing behind a switch
Shuo93 Aug 1, 2026
f3fe55e
feat(harness): time rail init through a single entry point
Shuo93 Aug 1, 2026
f49671d
perf(harness): cache context files by their on-disk identity
Shuo93 Aug 1, 2026
66edd8a
chore(core): report a per-callback split of every rail chain
Shuo93 Aug 1, 2026
17d53c0
doc(operator): Operator documentation supplement
Aug 3, 2026
754ec3a
docs: update docs based on v0.1.16.post1
yzqcjdtc Aug 3, 2026
be9fc03
fix(agent-team): humaninbox can auto_start_member when target is none
tangfei0306 Aug 3, 2026
22f64cd
fix(LSP): avoid blocking LSP rail initialization
yau703 Aug 3, 2026
70b619e
feat(evolution): JiuwenSwarm online RL optimize for AgentOS
Aug 3, 2026
319f8a0
feat(agent_teams): add claude otel bridge
ZhCoding Aug 3, 2026
c389bd3
fix(agent_teams): bind external cli context to member backend
ZhCoding Aug 3, 2026
b29a790
fix(harness): use builtin list annotations
ZhCoding Aug 3, 2026
602b8a7
fix(evolving): docs(evolving): Supplement the self-evolving online RL…
Aug 3, 2026
ba8cfd1
feat(harness): support the creation, loading, and execution of templa…
maggiezhangshere Aug 3, 2026
a9f5fa1
[mirror] openJiuwen-ai/agent-core#255: fix(context): stop counting ba…
Aug 4, 2026
6127f5a
feat(team): JSON string tool parameter missing
Aug 4, 2026
0278ce7
fix(context): refresh built-in model context window table
lindada11 Aug 4, 2026
da31164
feat(rsi): add recursive self-improvement framework
ggg66 Aug 4, 2026
db6825d
fix(agent-team): add hitt、trace、permissions docs
tangfei0306 Aug 4, 2026
1e715c1
fix(browser subagent): improve browser subagent tools and action speed
ShowHsiang Aug 4, 2026
e54e238
fix(context): invalidate stale usage token baselines
lindada11 Aug 4, 2026
523ebd9
[mirror] openJiuwen-ai/agent-core#288: fix(llm): drop top_p when temp…
Aug 4, 2026
19b1053
fix: preserve boolean substrings in condition variables
tangfei0306 Aug 4, 2026
5d139c5
fix(agent-evolving): tighten skill evolution contracts
z2tng Aug 4, 2026
84e4d43
fix(tiny-agent): require structured output for schema runs
Aug 4, 2026
cd3f640
feat(symphony): add symphony module
xiao-yu-lin Aug 4, 2026
eb8a9e5
docs(context_engine): document the default forked processor chain
lindada11 Aug 5, 2026
7c41449
fix(memory): support jiuwen memory(mem2.0) external provider
zhouziliang1015 Aug 5, 2026
44f9689
feat(symphony): add experience module for skill retrieval
yanmenxue Aug 5, 2026
46bb3da
feat(swarmflow): add budget/tokens and nested workflow progress obser…
chenchunzhou1995 Aug 5, 2026
b46dd82
chore(llm): Add anthropic dependency
Li-111 Aug 5, 2026
905f0ee
docs: add missing English translations for comprehensive zh/en alignment
yzqcjdtc Aug 5, 2026
d9bbf68
feat(symphony): add capability fingerprint evaluation
layShang264 Aug 5, 2026
ddd6011
refactor(symphony): refactor symphony
xiao-yu-lin Aug 6, 2026
89738f1
[mirror] openJiuwen-ai/agent-core#273: feat(docs): browser subagent docs
Aug 6, 2026
ce4b06d
[mirror] openJiuwen-ai/agent-core#291: fix (mcp): stop dropping tool …
Aug 6, 2026
36c4230
feat(symphony): symphony move to retrieval
doujzc Aug 6, 2026
e98a0f9
fix: Apply prompt improvements and incomplete-read guidance
yan-zj722 Aug 6, 2026
c4e0470
refactor(rsi): streamline integration code
ggg66 Aug 6, 2026
305463d
feat(rl): add online RL gateway client and validation guide
Aug 6, 2026
e3fb7c5
fix(llm): use safe _extract_reasoning_tokens in IntelliRouterModelClient
hwxiajing Aug 6, 2026
11bb65b
fix(worktree): initialize worktree session before task scheduler
Aug 6, 2026
346cddd
fix(swarm): separate an avatar's controller channel from team user
Shuo93 Aug 5, 2026
b2d7a1a
feat(swarm): emit a debounced team-idle marker from the leader stream
Shuo93 Aug 5, 2026
c433e08
fix(observability): resolve the trace root span outside its ContextVar
Shuo93 Aug 5, 2026
f9edf7d
docs(swarm): spell out broadcast cost in send_message docs and prompts
Shuo93 Aug 6, 2026
7db2b31
fix(harness): pass the parent sys_operation to injected sub-agents
Shuo93 Aug 6, 2026
9a5ad54
feat(swarm): expose cli_agent on MemberInfo
Shuo93 Aug 6, 2026
e600738
fix(observability): nest a sub-agent under the tool call that dispatc…
Shuo93 Aug 6, 2026
8c9dcf7
fix(observability): correlate llm.call spans by request id
Shuo93 Aug 6, 2026
4567f9d
fix(observability): trace every sub-agent, and nest rounds under thei…
Shuo93 Aug 6, 2026
d9e4053
fix(observability): stop double-counting cached and reasoning tokens
Shuo93 Aug 6, 2026
074c105
fix(observability): stop reporting unmeasured reasoning as instant
Shuo93 Aug 6, 2026
5e4c9ab
fix(harness): honor blank endpoint overrides for paid search
Shuo93 Aug 6, 2026
dbe4677
fix(swarm): hide swarmflow from the leader when the tool is gated out
Shuo93 Aug 6, 2026
d0d9a28
fix(team): clarify team_desc is context, not a directive for members
ZhCoding Aug 7, 2026
f44869d
feat(agent_teams): support ssh exec toggle
ZhCoding Aug 7, 2026
f83c9d8
feat(agent-teams): add fork with checkpoint and compaction
Little-Shepherd Aug 7, 2026
65ef36e
fix(goal): Change prompt injection to dynamic injection
Li-111 Aug 7, 2026
8ef9567
fix(tiny_agent): make tiny-agent structured output retries model-comp…
Aug 7, 2026
376d922
feat(team): rework & replan at review fail + add roles for reviewer
Aug 7, 2026
22f1337
fix: persist React context after interrupted runs
lindada11 Aug 8, 2026
6eac310
feat(evolution): support reviewer feedback driven skill evolution
Mavs-32 Aug 10, 2026
f5fa6e8
fix(symphony): 图谱构建时,单个SKILL异常不影响整体图谱构建
xiao-yu-lin Aug 10, 2026
550da6b
fix(symphony): 修复Windows系统SKILL目录扫描失败的问题
xiao-yu-lin Aug 10, 2026
65d5023
fix(symphony): 修复模型不可用异常未返回给swarm图谱构建的问题
xiao-yu-lin Aug 10, 2026
5408621
fix(agent_teams): default to autonomous dispatch and disable task ver…
ZhCoding Aug 10, 2026
ff698d8
fix(mcp): surface BaseExceptionGroup from add_tool_server connect
ChenfromChina123 Aug 10, 2026
File filter

Filter by extension

Filter by extension


Conversations
Failed to load comments.
Loading
Jump to
The table of contents is too big for display.
Diff view
Diff view
  •  
  •  
  •  
30 changes: 30 additions & 0 deletions .gitignore
Original file line number Diff line number Diff line change
Expand Up @@ -9,6 +9,10 @@ __pycache__/
# Distribution / packaging
.Python
build/
!openjiuwen/symphony/retrieval/build/
!openjiuwen/symphony/retrieval/build/**
!tests/unit_tests/symphony/retrieval/build/
!tests/unit_tests/symphony/retrieval/build/**
develop-eggs/
dist/
downloads/
Expand Down Expand Up @@ -255,3 +259,29 @@ images/
*.bin
*.sqlite3
tests/system_tests/config_llm_local.yaml

# Harness workspace scaffolding that unit tests write to the repo root when they
# run from it instead of a scratch dir (copies of the templates under
# openjiuwen/harness/prompts/workspace_content/).
# Root-anchored with a leading "/" on purpose: every one of these names is also a
# real tracked path (memory/, skills/, context/, agents/, and the AGENT.md /
# SOUL.md templates themselves), and an unanchored rule matches at any depth --
# it would silently ignore new files added under those source directories.
/AGENT.md
/HEARTBEAT.md
/IDENTITY.md
/SOUL.md
/USER.md
/agents/
/coding_memory/
/context/
/memory/
/messages/
/skills/
/todo/

# context_evolver unit test store, written next to the test
tests/unit_tests/extensions/context_evolver/memory_file/

/.codex/
/.agents/
2 changes: 1 addition & 1 deletion AGENTS.md
Original file line number Diff line number Diff line change
Expand Up @@ -74,4 +74,4 @@ Pass `COMMITS=N` to check recent commits instead.
- Logging: `.claude/rules/logging.md`
- Git workflow: `.claude/rules/git-workflow.md`
- Deep operational guides: `.claude/skills/`
- Permissions and env vars: `.claude/settings.json`
- Permissions and env vars: `.claude/settings.json`
2 changes: 2 additions & 0 deletions README.md
Original file line number Diff line number Diff line change
Expand Up @@ -175,3 +175,5 @@ We welcome all forms of contributions, including but not limited to:
## Open Source License

This project is licensed under the Apache-2.0 License.

This product serves solely as a workflow orchestration tool and does not embed any AI model capabilities. When users integrate AI models for specific business scenarios, they shall bear full responsibility for compliance obligations under the EU AI Act and other relevant regulatory frameworks.
2 changes: 2 additions & 0 deletions README.zh.md
Original file line number Diff line number Diff line change
Expand Up @@ -175,3 +175,5 @@ WorkflowAgent output result >>> 工作流输出文本: 冰箱罢工了,因为
## 开源许可证

本项目依据Apache-2.0许可证授权。

本产品仅作为流程编排工具,不包含 AI 模型能力;用户在连接 AI 模型用于特定业务场景时,需自行承担欧盟 AI 法案等相关合规义务。
Binary file added docs/dev/zh-doc-audit-issues-v0.1.10-16.xlsx
Binary file not shown.
257 changes: 257 additions & 0 deletions docs/dev/zh-doc-audit-v0.1.10-16.md

Large diffs are not rendered by default.

167 changes: 167 additions & 0 deletions docs/dev/zh-doc-fix-summary-v0.1.10-16.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,167 @@
# 文档修复汇总(v0.1.10–v0.1.16,杨泽渠 29 项)

> 基线:tag `v0.1.16.post1`(commit `b28b4f3a6`),分支 `dev_doc_v0.1.16.post1`
> 审计依据:`docs/dev/zh-doc-audit-v0.1.10-16.md`(v3 校正版)
> 修复范围:审计中责任人 杨泽渠 的 29 项(Batch A/B/C/D)
> 原则:仅改文档,不改代码(假定代码正确);存疑即停,不臆造。
> **审阅修正(commit `7c2d7b1ac`,新增提交非 amend)**:依据审阅反馈纠正 4 项——WF-04(HTTPRequestComponent 实为导出,修正导入)、AS-02(回退 agent_evolving 行,仅保留 auto_harness)、AS-11(整体回退以避免冲突)、AS-01(加深为按子包的详细 `.md`,进行中)。详见 §七。

## 一、总体统计

| 批次 | 类别 | 应修 | 已修 | 跳过 | 备注 |
|---|---|---|---|---|---|
| A | Mismatch(内容与源码不符) | 11 | 11 | 0 | 全部完成 |
| B | Missing(向已有文档补内容) | 9 | 7 | 2 | AS-02 审阅回退、AS-15 跳过 |
| C | Missing(新建 API 文档) | 8 | 5 | 3 | AS-06 跳过、AS-11 审阅回退、AS-14 部分(计入跳过列) |
| D | N/A(非真实缺失) | 1 | 0 | 1 | AS-16 确认非 gap |
| **合计** | | **29** | **23** | **6** | AS-14 部分应用计入跳过列 |

跳过/回退项均有依据(见 §四、§七)。

## 二、修复明细(按 issue)

### Batch A — Mismatch(11 项,全部完成)

| ID | 严重度 | 文件 | 修改内容 |
|---|---|---|---|
| WF-01 | High | 编排组件.md:113、使用预置组件.md:781 | `Start({})` → `Start()`:源码 `__init__(self)` 无参数,示例不可运行。 |
| WF-02 | High | 使用预置组件.md:1065-1077、:1186-1189 | Memory 组件导入路径修正:`from openjiuwen.core.workflow import MemoryWriteComponent` → `from openjiuwen.core.workflow.components.resource.memory_write_comp import MemoryWriteComponent`(`__init__.__all__` 无 Memory 组件组)。 |
| HAR-01 | High | Agent Skills.md:135 | OperationMode 补全 SANDBOX(源码为 LOCAL + SANDBOX,非仅 LOCAL)。 |
| HAR-02 | High | Agent Skills.md:143 | 移除 `LocalWorkConfig(work_dir=...)` 中不存在的 `work_dir` 字段。 |
| HAR-04 | High | Agent Skills.md:87、:283 | 导入路径修正:`from openjiuwen.core.skills` → `from openjiuwen.core.single_agent.skills`。 |
| MEM-01 | High | 记忆引擎.md:22-23 | 补全 MemoryEngineConfig 遗漏字段。 |
| MEM-02 | High | 记忆引擎.md:354-365 | `add_messages` 示例演示取回返回值 `AddMemResult`(含写入的 memory-unit 列表,非聚合统计)。 |
| MEM-03 | Med | 记忆引擎.md:216-220 | `register_store` 的 `db_store` 参数标注 v0.1.16 后可选(默认 None)。 |
| RET-01 | Med | 知识检索.md:115 | 支持的文件格式列表补 HTML(v0.1.11 新增 html_file_parser)。 |
| RET-02 | Med | 记忆引擎.md:90-95 | ProviderType 枚举补全(与 AS-03 联动,详见 §三)。 |
| RET-03 | Med | 接入大模型.md:52 | 客户端列表补全(原文仅提 OpenAI/SiliconFlow,源码支持 11 个)。 |

### Batch B — Missing(向已有文档补内容,9 项)

| ID | 严重度 | 文件 | 修改内容 | 状态 |
|---|---|---|---|---|
| WF-03 | High | 构建工作流.md:57 | 补 `add_workflow_comp` 的 `max_retries`/`timeout`/`exception_config`(`ExceptionConfig.handle_type`)参数文档。 | ✓ |
| WF-04 | High | 使用预置组件.md | 补 HTTPRequestComponent 节。**审阅修正**:原误判为「未在 `__init__` 导出,需直引模块路径」;实证 `_HTTP_RELATED_COMPONENTS`(`workflow/__init__.py:138-151`)含 HTTPRequestComponent + 11 个 HTTP 符号并合并入 `__all__`,正确导入为 `from openjiuwen.core.workflow import HTTPRequestComponent, HttpComponentConfig`(§七)。 | ✓(已修正) |
| WF-05 | Med | 组件支持的能力类型.md:7-19 | 能力类型表补 `MemoryWriteComponent`、`HTTPRequestComponent` 两行。 | ✓ |
| CTX-01 | High | 上下文引擎.md | 补上下文处理器系统章节:10 个处理器(6 压缩器 + 4 offloader)。 | ✓ |
| RET-04 | Med | 知识检索.md | 补 DashscopeReranker 概念覆盖。 | ✓ |
| RET-05 | High | 记忆引擎.md / 知识检索.md | 补 LakeBaseMemoryProvider(Lakeon 湖仓记忆)概念覆盖。 | ✓ |
| RET-06 | Med | 插件开发-存储后端.md | 补消息队列模块概念文档(Pulsar 后端依赖模型;导入为可选依赖,失败需说明)。 | ✓ |
| AS-02 | Med | API文档/README.md:5-11 | 原索引表补 `agent_evolving` 行。**审阅修正**:回退 `agent_evolving` 行(保留 `auto_harness` 行);总索引由同事后续统一。 | 回退 |
| AS-15 | Low | extensions/tracer_otel/ | tracer_otel 部分覆盖需扩展。 | 跳过 |

### Batch C — Missing(新建 API 文档,8 项)

| ID | 严重度 | 新建/扩展文件 | 修改内容 | 状态 |
|---|---|---|---|---|
| AS-01 | High | `openjiuwen.auto_harness.README.md`(新建索引)+ `openjiuwen.auto_harness/*.md`(15 个详细子包文档,进行中) | 初版子包索引(13 子目录、约 90 个类/函数一句话说明,AST 抽取)。**审阅加深**:按子包输出详细 `.md`(类签名 / `__init__` 参数 / 方法 / 返回值 / 示例,参照 `llm.md` 深度),写入 `openjiuwen.auto_harness/` 子目录,README 更新为链接索引(§七)。 | 进行中 |
| AS-03 | High | `llm.md`(扩展,+167 行) | 补 6 个缺失 ModelClient(Anthropic/DeepSeek/AscendAffinity/InferenceAffinity/OpenRouter/IntelliRouter)+ `IntelliRouterClientConfig`;更新尾部 ProviderType 说明。 | ✓ |
| AS-04 | High | `foundation/store/memory_index.md`(新建) | BaseMemoryIndex ABC 逐方法 + MemoryDoc + StorageCodec Protocol + SimpleMemoryIndex(含 DEPRECATED 标注)。 | ✓ |
| AS-05 | Med | `common/background_tasks.md`(新建) | 协程任务管理 API 文档(旧 task_manager.md 仅覆盖 task_manager.py)。 | ✓ |
| AS-06 | High | — | LakeBaseMemoryProvider 零 API 文档。 | 跳过 |
| AS-11 | High | `extensions/context_evolver/README.md`(原新建) | 原子包索引:ACE/Cognition/ReasoningBank/ReMe 四算法 retrieve+summary+service+ReAct agent。**审阅修正**:整体回退(删除文件、移除 SUMMARY 与 extensions.README 行),避免冲突(§七)。 | 回退 |
| AS-12 | Med | `extensions/sys_operation/sandbox/providers/jiuwenbox.md` + `sys_operation/README.md`(新建) | 3 provider + 5 helper + mixin 实现细节 + 6-provider 索引表。 | ✓ |
| AS-14 | Med | `extensions/vendor_specific/README.md`(新建) | `AliyunReranker` 弃用别名文档;`harness`/`common` 跳过。 | 部分 |

### Batch D — N/A(1 项)

| ID | 严重度 | 说明 | 状态 |
|---|---|---|---|
| AS-16 | Low | `deepagents/` 仅含 `.pyc` 字节码(52 个),无活跃 `.py` 源文件;DeepAgent 已在 `harness/deep_agent.md` 覆盖。非真实缺失(与附录 A.2 同条)。 | 跳过(已复核确认) |

## 三、AS-03 / RET-02 联动说明

`RET-02`(记忆引擎.md 的 ProviderType 说明)与 `AS-03`(llm.md 客户端层)同源同因,合并处理:

- **ProviderType 枚举**(`foundation/llm/schema/config.py:13-24`,共 10 值):OpenAI / OpenAIAccount / OpenRouter / Anthropic / SiliconFlow / DashScope / DeepSeek / InferenceAffinity / AscendAffinity / IntelliRouter(`intelli_router`)。
- **llm.md** 补 6 个原缺失客户端 + `IntelliRouterClientConfig`(896 → 1063 行):Anthropic(进程级 AsyncAnthropic 缓存池)、DeepSeek(`reasoning_content` 重写)、AscendAffinity(KV-cache 亲和)、InferenceAffinity(vLLM `release()`)、OpenRouter(attribution 头保护 + 提示词缓存)、IntelliRouter(ReliableRouter 封装 + 委托 DashScope)。
- **记忆引擎.md** ProviderType 说明处同步补全枚举值。

## 四、跳过项依据

| ID | 跳过原因 |
|---|---|
| AS-06 | 前提错误:`memory/external/lakebase_memory_provider.md`(512 行)已存在,与 RET-05 交叉。非真实缺失。 |
| AS-15 | 用户决定跳过(Low;tracer_otel 团队级可观测性扩展属多智能体团队 TE-04 范畴)。 |
| AS-14(harness 子集) | `extensions/harness/__init__.py` 为 0 字节空命名空间,无任何公共 API。 |
| AS-14(common 子集) | `extensions/common/` 仅含 `__pycache/` 下 `.pyc` 字节码(configs/、log/),无 `.py` 源文件,与 AS-16 同类遗留。 |
| AS-16 | `deepagents/` 仅 `.pyc`,无活跃 `.py`;DeepAgent 已在 `harness/deep_agent.md` 覆盖。审计已标注「非真实缺失」。 |

> AS-14 的 `harness`/`common` 跳过依据已写入 `vendor_specific/README.md` 的「备注」节,保持可追溯。

## 五、文件清单

### 新建文档(6 个,已提交)

| 文件 | 对应 issue |
|---|---|
| `docs/zh/2.开发指南/API文档/openjiuwen.auto_harness.README.md` | AS-01(索引;详细子包文档见下「进行中」) |
| `docs/zh/2.开发指南/API文档/openjiuwen.core/foundation/store/memory_index.md` | AS-04 |
| `docs/zh/2.开发指南/API文档/openjiuwen.core/common/background_tasks.md` | AS-05 |
| `docs/zh/2.开发指南/API文档/openjiuwen.extensions/sys_operation/README.md` | AS-12 |
| `docs/zh/2.开发指南/API文档/openjiuwen.extensions/sys_operation/sandbox/providers/jiuwenbox.md` | AS-12 |
| `docs/zh/2.开发指南/API文档/openjiuwen.extensions/vendor_specific/README.md` | AS-14 |

> AS-11 原新建的 `extensions/context_evolver/README.md` 已随审阅修正删除,不再列入。

### 进行中(AS-01 加深,待提交)

| 文件 | 内容 |
|---|---|
| `docs/zh/2.开发指南/API文档/openjiuwen.auto_harness/orchestrator.md` | 顶层入口:AutoHarnessOrchestrator / create_auto_harness_orchestrator / load_auto_harness_config 等 |
| `docs/zh/2.开发指南/API文档/openjiuwen.auto_harness/schema.md` | ~30 数据类/枚举 + AutoHarnessConfig(~40 字段) |
| `docs/zh/2.开发指南/API文档/openjiuwen.auto_harness/agents.md` | 11 个 agent 工厂函数 |
| `docs/zh/2.开发指南/API文档/openjiuwen.auto_harness/artifacts.md` | ArtifactStore |
| `docs/zh/2.开发指南/API文档/openjiuwen.auto_harness/contexts.md` | TaskRuntime / BaseExecutionContext / SessionContext / TaskContext / task_key |
| `docs/zh/2.开发指南/API文档/openjiuwen.auto_harness/experience.md` | ExperienceStore / ActiveContextSynthesizer |
| `docs/zh/2.开发指南/API文档/openjiuwen.auto_harness/infra.md` | 17 模块:CI 门控 / git+worktree / 运行时扩展 / 解析选择 / PR 模板 / 编辑范围 |
| `docs/zh/2.开发指南/API文档/openjiuwen.auto_harness/pipelines.md` | BasePipeline / MetaEvolve / ExtendedEvolve / PRTask / ExtensionTask |
| `docs/zh/2.开发指南/API文档/openjiuwen.auto_harness/prompts.md` | build_auto_harness_sections |
| `docs/zh/2.开发指南/API文档/openjiuwen.auto_harness/rails.md` | 7 个护栏 |
| `docs/zh/2.开发指南/API文档/openjiuwen.auto_harness/registry.md` | BaseRegistry / Stage/PipelineRegistry + 4 注册函数 |
| `docs/zh/2.开发指南/API文档/openjiuwen.auto_harness/stages.md` | 11 个 stage 模块(base/assess/plan/implement/verify/commit/publish/learnings/activate/merge/select_pipeline) |
| `docs/zh/2.开发指南/API文档/openjiuwen.auto_harness/tools.md` | ExperienceSearchTool / ExperienceSearchMetadataProvider |
| `docs/zh/2.开发指南/API文档/openjiuwen.auto_harness/resources.md` | resources/ 下 2 个 YAML 配置模板参考 |
| `docs/zh/2.开发指南/API文档/openjiuwen.auto_harness/skills.md` | skills/ 下 10 个 SKILL.md 提示词技能参考 |

### 修改文档(18 个)

| 文件 | 对应 issue |
|---|---|
| `docs/zh/SUMMARY.md` | AS-02、AS-04、AS-05、AS-12、AS-01、AS-14(TOC 条目);AS-11 行已随回退移除 |
| `docs/zh/2.开发指南/API文档/README.md` | AS-01(auto_harness 行);AS-02 行已随回退移除 |
| `docs/zh/2.开发指南/API文档/openjiuwen.extensions.README.md` | AS-12、AS-14(MODULE 表行);AS-11 行已随回退移除 |
| `docs/zh/2.开发指南/API文档/openjiuwen.core/common.README.md` | AS-05(索引项) |
| `docs/zh/2.开发指南/API文档/openjiuwen.core/foundation/store.md` | AS-04(intro bullet) |
| `docs/zh/2.开发指南/API文档/openjiuwen.core/foundation/store.README.md` | AS-04(Classes 表 4 行) |
| `docs/zh/2.开发指南/API文档/openjiuwen.core/foundation/llm/llm.md` | AS-03(+167 行) |
| `docs/zh/2.开发指南/工作流/使用组件/编排组件.md` | WF-01 |
| `docs/zh/2.开发指南/工作流/使用组件/使用预置组件.md` | WF-01、WF-02、WF-04(审阅修正导入) |
| `docs/zh/2.开发指南/工作流/构建工作流.md` | WF-03 |
| `docs/zh/2.开发指南/工作流/使用组件/组件支持的能力类型.md` | WF-05 |
| `docs/zh/2.开发指南/智能体/Agent Skills.md` | HAR-01、HAR-02、HAR-04 |
| `docs/zh/2.开发指南/高阶用法/记忆引擎.md` | MEM-01、MEM-02、MEM-03、RET-02、RET-05 |
| `docs/zh/2.开发指南/高阶用法/知识检索.md` | RET-01、RET-04 |
| `docs/zh/2.开发指南/基础功能/接入大模型.md` | RET-03 |
| `docs/zh/2.开发指南/高阶用法/上下文引擎.md` | CTX-01 |
| `docs/zh/2.开发指南/高阶用法/插件开发-存储后端.md` | RET-06 |

## 六、未尽事项

- **AS-01 加深(进行中)**:6 个并行子代理正基于权威 AST 抽取(`auto_harness_api_rich.txt`,1235 行、65 文件、12 子包)撰写 15 个详细子包 `.md`;完成后更新 README 链接索引与 SUMMARY 子条目并提交。
- **AS-15**(tracer_otel 团队级可观测性扩展):用户已决定跳过;如后续需补,属多智能体团队 TE-04 范畴。
- **多智能体团队项**(TE-02/03/04、AS-07/08/09/10、AS-13、AS-17):非杨泽渠职责,不在本次修复范围。
- 本次未改动任何 `.py` 源码;未执行 `make check`/`make type-check`(无 Python 改动,这两者作用于暂存 Python 文件)。

## 七、审阅修正(commit `7c2d7b1ac`)

依据审阅反馈对 4 项做纠正,均以新增提交 `7c2d7b1ac`(非 amend)落地,保留审计轨迹:

| ID | 原处理 | 审阅修正 | 依据 |
|---|---|---|---|
| WF-04 | 原注明「HTTPRequestComponent 未在 `__init__` 导出,需直引模块路径」 | 修正:`_HTTP_RELATED_COMPONENTS`(`workflow/__init__.py:138-151`)含 HTTPRequestComponent + 11 个 HTTP 符号并合并入 `__all__`(line 168+);正确导入为 `from openjiuwen.core.workflow import HTTPRequestComponent, HttpComponentConfig`。 | 源码 `openjiuwen/core/workflow/__init__.py:56-68, 138-151, 168+` 实证 |
| AS-02 | 原 `API文档/README.md` 补 `agent_evolving` 行 | 回退 `agent_evolving` 行(保留 `auto_harness` 行);同事后续统一总索引。 | 审阅反馈:总索引由同事统一处理 |
| AS-11 | 原新建 `context_evolver/README.md` + SUMMARY/extensions 行 | 整体回退:删除文件、移除 SUMMARY 与 extensions.README 行,避免冲突。 | 审阅反馈:避免合并冲突 |
| AS-01 | 原仅子包索引(一句话表) | 加深为按子包的详细 `.md`(类签名 / `__init__` 参数 / 方法 / 返回值 / 示例,参照 `llm.md` 深度);15 个子包文档写入 `openjiuwen.auto_harness/` 子目录,README 更新为链接索引。 | 审阅反馈:用户选择「加深为按子包的详细 .md」 |

> **WF-04 同类错误排查**:WF-02(Memory 组件)经核实确未导出——`_RESOURCE_RELATED_COMPONENTS`(line 132-136)仅含 ComponentKBConfig/KnowledgeRetrievalComponent/config,无 Memory 组件,深路径正确;HAR-04(GitHubTree)位于 `openjiuwen.core.single_agent.skills`(`__init__.py:13,19` 重导出自 `remote_skill_util.py:20`),正确。未发现同类错误。
2 changes: 2 additions & 0 deletions docs/en/2.Development Guide/API Docs/README.md
Original file line number Diff line number Diff line change
Expand Up @@ -8,4 +8,6 @@ This section contains the API documentation for the main openJiuwen packages.
| [`openjiuwen.agent_teams`](./openjiuwen.agent_teams.README.md) | Multi-agent team collaboration framework: Leader/Teammate collaboration, transport layer, and persistent storage. |
| [`openjiuwen.dev_tools`](./openjiuwen.dev_tools.README.md) | Development tools: `prompt_builder`, `tune` (prompt generation/optimization and tuning), and [`agent_builder`](./openjiuwen.dev_tools/agent_builder.README.md) for building Agents. |
| [`openjiuwen.harness`](./openjiuwen.harness.README.md) | Agent harness framework: DeepAgent with built-in planning, context management, sub-agent spawning, rails, tools, and long-term memory. |
| [`openjiuwen.agent_evolving`](./openjiuwen.agent_evolving.README.md) | Self-evolution framework: online RL training, experience store, signal collection, and trajectory management. |
| [`openjiuwen.extensions`](./openjiuwen.extensions.README.md) | Optional extensions, such as `message_queue`. |
| [`openjiuwen.auto_harness`](./openjiuwen.auto_harness.README.md) | Autonomous optimization harness framework: orchestrator, pipelines, stages, rails, and experience store. |
Loading