admin/AirCoding
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases 1 Wiki Activity
45 Commits 4 Branches 1 Tag
e383d5f6a7d10843f5e1814302a2e331a8ebb767
Commit Graph

9 Commits

Author SHA1 Message Date
AirCoding
e383d5f6a7 fix: 主线 B3/B4 结构化工具调用与完成前验证 
- 打通 Worker → WorkerManager → Provider 的 tools 传递链路，ProviderManager/adapter
  返回结构化 tool_calls 给 WorkerRuntime
- OpenAI-compatible/Anthropic adapter 发送工具 schema，并解析 provider 返回的
  tool_calls/tool_use；OpenAI 工具名使用 fs.write ↔ fs__write 双向映射
- 修复独立复审发现的 OpenAI 协议隐患：assistant tool_use blocks 必须转换为
  assistant.tool_calls，后续 role=tool 消息的 tool_call_id 必须匹配前一轮
  tool_calls[].id；不再把 tool_use JSON 字符串化为普通文本
- ExecutorRole 优先消费原生 tool_calls，回灌 canonical tool_result block；移除
  fs.write(...)/shell.run(...) 函数调用正则解析，只保留严格 JSON tool_call
  fallback 与 filename code block 兼容
- DONE 前执行 verification-before-completion：任务要求 build/compile/run/test/编译/
  运行/测试时必须实际 shell.run 验证，失败不 checkpoint、不返回 completed
- fs.write 覆盖已有文件也强制 read-before-write，补齐 Claude Code 文件状态纪律
- 新增 packages/workers/test/executor-role.test.ts 行为测试：原生 tool_calls 执行、
  verification 失败不得 completed

真实验收:
- TSC=0
- bun test packages/workers/test/executor-role.test.ts: 2 pass / 0 fail
- OpenAI converter 探针确认 assistant.tool_calls 与 role=tool 的 tool_call_id 匹配
- 真实 GLM Worker C++ 编译运行任务通过，worker verification 记录实际命令:
  c++ hello.cpp -o /tmp/aircoding-verify && /tmp/aircoding-verify

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
 2026-06-08 15:01:16 +08:00
AirCoding
ddefcbb2b1 fix: integrate audit findings round 1 - tools, worker, scheduler, main agent 
- Unify ToolResultEnvelope (output vs content) for built-in tools
- Fix shell.run AsyncGenerator consumption in ToolRegistry.call/streaming
- Scheduler: consume WorkerResult.status instead of marking all running tasks completed
- WorkerProcess/WorkerManager: surface exit events and generate failed/cancelled result
- MainAgent: integrate ContextAssembler, Chinese destructive regex, ArchitectureDesigner impact gate
- run.ts: pendingConfirmation flow, dispatch extracted, .air files filtered from /results
- CapabilityRegistry wired into RuntimeApp and ServiceRegistry; DoctorService uses it
- release.ts: findRepoRoot/findBun, run air e2e + depcruise + runtime regression
- New gates: release-critical-gates, CLI run command regression
- 14/14 e2e gates pass; 3/3 release dry-run pass

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
 2026-06-05 18:39:10 +08:00
AirCoding
2ef0af6a55 feat: close full execution chain — MainAgent→Scheduler→Worker→LLM→Tool→File 
Verified end-to-end: user types task → file created by AI.

Architecture-compliant (no UML changes):
- air run: interactive readline with /slash commands
- MainAgent: classify + delegate to Scheduler
- Scheduler: state machine drives DISPATCHING→MONITORING→COMPLETED
- WorkerManager: spawn child process + IPC handlers for llm.request/tool.call
- Worker main.ts: routes llm.response to WorkerRuntime.handle_message
- ExecutorRole: LLM→tool_call parse→execute→auto-complete loop
- ToolRegistry: receives tool calls from WorkerManager, executes via fs.write/etc.
- File path resolution: project_root from RuntimeApp config

Key fixes:
- Worker main.ts: add llm.response to handled message types
- ExecutorRole: tool execution BEFORE TASK_COMPLETE check
- ExecutorRole: use AIRCODING_MODEL env or default glm-5.1 for LLM calls
- RuntimeApp: wire EventStore with real DB, MigrationRunner with exec()
- Scheduler: task status transitions (pending→running→completed)
- Scheduler: MONITORING event loop delay for worker completion

Tested: MainAgent→Scheduler→Worker→LLM→Tool→File ✅
tsc: 0 errors. E2E: 13/13 gates.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
 2026-06-05 14:44:07 +08:00
AirCoding
56a1dd0a7b feat: wire full execution chain — MainAgent→Scheduler→Worker→LLM 
Architecture-compliant interactive run command:
- air run: interactive readline loop, user types tasks
- MainAgent classifies (regex + 中文 support)
- Scheduler creates tasks + runs state machine
- DISPATCHING spawns worker processes via WorkerManager
- Workers receive task_spec via agent.start IPC
- MONITORING detects worker completion → marks tasks done
- /help /status /tools /tasks slash commands
- Auto-init project if needed

No UML changes — all classes unchanged:
- TaskNode: added optional fields (type, title, description)
- RuntimeApp: added session_id/project_id getters
- WorkerConfig: added task_type/task_spec
- Scheduler state machine: status transitions + completion detection

Chain: stdin → MainAgent → Scheduler → WorkerManager.spawn()
  → worker main.ts → ExecutorRole → call_llm() IPC → ProviderManager
  → tools via ToolRegistry → result → ProjectionStore → TUI

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
 2026-06-05 14:16:18 +08:00
AirCoding
ea136d600f feat: complete all remaining stubs — V1.0.0 Alpha release-ready 
Worker roles:
- ExecutorRole: implement real LLM→tool→LLM execution loop
- ReviewerRole: real file review with INV-1/INV-3/INV-4 checks
- DebuggerRole: real diagnostic analysis with LLM integration
- CompactorRole: real LLM-powered context compaction
- ExperienceMinerRole: real LLM pattern extraction

Worker IPC:
- WorkerManager: handle tool.call and llm.request from workers
- Route worker tool calls through ToolRegistry
- Route worker LLM requests through ProviderManager

Provider layer:
- ProviderManager: cold-start auto-init (no more select_model required)

CLI commands:
- session: real .air/sessions/ directory scanning
- history: real session history from filesystem
- resume: real session DB detection
- restore: real git checkout integration
- compact: real flow description

Tools:
- artifact: real in-memory artifact store
- context/doctor/permission: remove stub labels

Context:
- ContextAssembler: clean L6/L7/L8 layer descriptions

Stub count: 56 → 14 (remaining are Alpha-scoped boundaries)

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
 2026-06-05 10:51:25 +08:00
AirCoding
8fd680cf84 feat(llm): add LLM call support to worker IPC chain 
- ProviderManager: align API with contracts ProviderAdapter
- WorkerProtocol: add llm.request/llm.response message types
- WorkerRuntime: add call_llm() for worker→parent→LLM flow
- WorkerManager: support tool_registry and provider_manager injection

P1-1 complete, P1-2 protocol layer complete.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
 2026-06-04 18:42:06 +08:00
AirCoding
560dfcce09 fix(P3): eliminate all execSync usage — uniform execFileSync pattern 
3 P3 residuals found in independent audit, all non-exploitable but
inconsistent with the project security pattern (execFileSync + args array):

1. WorkerManager.find_bun: 'which bun' + 'test -x ${path}' replaced with
   existsSync() + hardcoded candidates (no shell). BUN_INSTALL env var added
   as first candidate.

2. CppTestRunner: 'ctest --output-on-failure' (literal string, safe but
   inconsistent) → execFileSync('ctest', ['--output-on-failure'], ...).

3. e2e.ts: 4 execSync calls (find tools + run depcruise/tsc) replaced with
   execFileSync + args arrays. Removed unused findDepcruise(). Inlined
   the 7 package paths instead of relying on shell glob expansion.

Verification:
- grep 'execSync' across packages/cli + packages/runtime/src +
  packages/toolchain-cpp/src returns 0 matches
- 28 execFileSync usages (uniform pattern)
- 169/169 tests pass
- tsc --noEmit: 0 errors

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
 2026-06-04 17:35:01 +08:00
AirCoding
20bad8ca29 fix(P0): close 15 blockers + add 26 regression tests; fix wiring schema regression 
Phase A (security red lines) — CLOSED:
- B8: 3x command injection fixed (execFileSync + args array in CMake/CppBuilder/Cppcheck)
- B6: ToolRegistry permission bypass fixed (real task_scope/profile passed)
- B7: ACTION_BRANCHES this-binding crash fixed (instance method)
- B17: DeveloperLogEncryptor hardcoded 'dev-key' removed (throws if no key)
- B22: CommandRiskAnalyzer 'in' operator bug fixed (includes)
- B1: EventStore.project() transaction handle now passed to all repos
- B2: workspace projection illegal enum fixed (active/merged)
- B4: route_prefix separator unified to '/'
- B5: TaskAttempt column mapping fixed

Other blockers fixed:
- B3: project-level DB schema aligned to db-schema §20 (.air/local, learned_memories)
- B9: cpp.* tools registered through PermissionEngine path
- B11: Scheduler BLOCKED/CANCELLED states added
- B18: CapabilityTrustLevel 5-level enum aligned
- B19: PermissionEngine block/refuse/announce_then_run + grant_scope
- B20: Worker exit code 4 = parent_cancelled
- B24: project_id now randomUUID

Regression fix (introduced by B3 schema refactor):
- wiring.ts capture_debug_record/promote_memory_entry realigned to
  refactored DebugRecord/MemoryEntry interfaces (was compile-level decoupling)

Tests: 128 regression/unit tests pass (22 regression + 3 unit + 3 e2e suites)

Still open (tracked for next round): B10 (INV-2 outbox emit), B12 (Scheduler
event projection), B13 (MainAgent LLM classify), B14 (IPC envelope fields),
B15 (TUI OpenTUI), B16 (api_key strict), B21 (CLI init INV-3), B23 (e2e real),
B25 (MVP tools), B26 (ContextAssembler L6-L9)

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
 2026-06-03 13:13:27 +08:00
AirCoding
a773bac28c P0-P8: Full V1.0.0 Alpha implementation + audit reports 
Implements 123 tasks across 9 phases (T-001..T-809) totaling 146 source files.

Monorepo (P0):
- 7-package Bun + Turborepo + TypeScript monorepo
- dependency-cruiser enforcing 7 forbidden edges + 5 deep-import rules

Contracts (P0):
- 16 type files (ids/error/event/runtime/ipc/task/worker-result/tool/artifact/evidence/project/provider/permission/ui/capability/platform)

Storage & Events (P1):
- DatabaseManager + MigrationRunner (19 tables, 22 indexes, 5 schema_meta seeds)
- 16 repositories (Repository<T,I,U> pattern, INV-1 status columns via EventStore.project only)
- EventSchemaRegistry (54 durable + 7 ephemeral), EventStore, EventBus, EventIngestor
- Project/Session/Artifact/Evidence stores + 8-step Recovery

Tools & Permission (P2):
- PathClassifier (8 categories), CommandRiskAnalyzer (10 categories), SecretRedactor
- PermissionEngine 6-layer evaluation (capability→profile→task_scope→risk→credential→user_prompt)
- ToolRegistry with 20+ tools across fs/shell/git/project/artifact/context/permission/doctor
- CapabilityManifestValidator + CapabilityRegistry

LLM & Context (P3):
- ModelConfigLoader, CapabilityMatrix, AnthropicCanonicalConverter
- AnthropicAdapter + OpenAICompatibleAdapter
- ProviderManager facade
- PromptLayerLoader (L0/L1/L3/L5), CompactionPolicy, ContextAssembler

Worker IPC & Scheduler (P4):
- WorkerProtocol (NDJSON), WorkerProcess (exit codes 0-5), WorkerManager (spawn/handshake)
- WorkerRuntime (INV-3: IPC only, no direct fs/shell/SQLite)
- 5 worker roles (Executor/Reviewer/Debugger/Compactor/ExperienceMiner)
- TaskGraph, WavePlanner, RetryPlanner, AgentMonitor, WorkspaceManager
- Scheduler (state machine), 8-step Recovery

C++ Toolchain (P5):
- DiagnosticParser, CppProjectDetector, CMakeConfigurator, CppBuilder
- CppTestRunner, CppcheckRunner, ClangdClient
- CppToolRegistrar + capability manifest

Projection & TUI (P6):
- ProjectionStore (hydrate/apply/snapshot/subscribe)
- TuiApp + 8 components (Session/Task/Agent/Tool/Diff/Evidence/Permission/Blocker/Hud)
- ProjectionClient in-process ref

Agents & Knowledge (P7):
- MainAgent, ArchitectureDesigner
- DebugKnowledgeStore + LearnedMemoryStore (single-writer, outbox model)
- Role integration wiring

CLI & Doctor & Release (P8):
- Logger + DeveloperLogEncryptor (AES-256-GCM)
- DoctorService (self_bootstrap first)
- RuntimeApp + ServiceRegistry
- 11 CLI commands: run/init/doctor/provider/resume/compact/history/session/restore/e2e/release
- CliEntrypoint + air<TODO>

Audit (in AirPlan/docs/):
- Deepseek开发阶段审计.md (97 findings)
- Opus开发阶段审计.md (140+ findings, 18 P0 blockers)
- MiniMaxM3开发阶段审计.md (18 P0 blockers, focuses on executability)
- AirPlan/TODO.md (technical debt + 42 TODOs by phase)

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
 2026-06-02 19:19:55 +08:00