Commit Graph

27 Commits

Author SHA1 Message Date
xiaoju 43a6600378 feat: ReAct ExtractFn with tool-use
- RoleDefinition.extractMode: "single" | "react"
- reactExtract: multi-turn LLM with cas_get tool for DAG traversal
- Max 10 tool-call rounds, schema validation on final output
- create-workflow routes to reactExtract when extractMode is "react"
- All existing roles set to "single" (no behavior change)
- 162 tests passing

Fixes #44
2026-05-07 13:28:00 +00:00
xiaoju 84de74721d feat: Merkle node format + content → CAS
- MerkleNode type: { type, payload, children } serialized as YAML
- RoleOutput.content → contentHash (CAS hash of Merkle content node)
- Engine stores content in global CAS before writing to .data.jsonl
- create-workflow puts content as Merkle node, merges contentHash into refs
- fork/parse adapted for contentHash format
- buildAgentPrompt now async, reads content from CAS
- Bundle validator allows @uncaged/workflow import
- 150 tests passing

BREAKING: .data.jsonl no longer contains inline content

Fixes #41
2026-05-07 13:14:01 +00:00
xiaoju e95e76c145 feat: workflowAsAgent factory
- workflowAsAgent(name) resolves via registry → bundle → child thread
- System-level depth limit (max 3, constant)
- Returns summary string, errors as string (no throw)
- Integration test with nested workflow execution
- 146 tests passing

Fixes #33
2026-05-07 10:52:26 +00:00
xiaoju 30e4e99908 feat: add refs tracking to RoleStep
- RoleOutput gains refs: string[] for CAS reference tracking
- RoleDefinition gains extractRefs: ((meta) => string[]) | null
- planner: phases.map(p => p.hash), coder: [completedPhase]
- Engine persists refs, fork preserves refs
- Backward compat: missing refs normalized to []
- 137 tests passing

Fixes #31
2026-05-07 10:44:25 +00:00
xiaoju 43f466eb67 style(solve-issue): fix biome formatting in test file
Refs #28
2026-05-07 10:15:10 +00:00
xiaoju fe829d9ae6 feat(workflow): add preparer role as first step in solve-issue workflow
- New package: @uncaged/workflow-role-preparer with PreparerMeta type,
  schema, system prompt, and extract prompt
- Preparer locates/clones repo, detects toolchain and conventions
- Moderator updated: preparer → planner → coder → reviewer → committer
- solve-issue template re-exports preparer types
- Tests updated for new flow (129 pass, 0 fail)

Fixes #28
2026-05-07 10:12:59 +00:00
xiaoju 341ff656dc feat(planner,coder,moderator): integrate CAS for phase tracking
Phase 2 of #23:
- Planner schema compact: {hash, title} only, details stored via CAS CLI
- Planner prompt instructs agent to shell out `cas put` for each phase
- Coder prompt instructs agent to `cas get` for phase details, report hash
- Moderator compares hashes instead of names
- Removed COMPLETED_PHASE_SENTINELS — hash matching eliminates ambiguity

Refs #23
2026-05-07 04:54:25 +00:00
xiaoju 4b44665c7e feat(workflow): add thread-scoped CAS (Content-Addressable Storage)
Phase 1 of #23:
- createThreadCas() core API: put/get/delete/list with XXH64 hashing
- hashString() utility for string → 13-char Crockford Base32
- CLI: uncaged-workflow cas get/put/list/rm subcommands
- thread rm now cleans up .cas/ directory
- 10 new tests for CAS operations

Refs #23
2026-05-07 04:30:19 +00:00
xiaoju 172e9b34cc feat(planner): add hash and title fields to phase schema
Each phase now carries a hash (Crockford Base32 identifier) and a
one-line title alongside the existing name/description/acceptance.
This gives agents immediate semantic context in the prompt without
needing to load full phase details from CAS.

Refs #23
2026-05-07 04:18:42 +00:00
xiaoju 96fc3e220a fix(solve-issue): handle one-shot coder completing all phases at once
The moderator now treats completedPhase matching the last planned phase name
as full completion. Also recognizes sentinel values (all-done, all_done,
complete) for robustness.

Fixes #21

小橘 🍊(NEKO Team)
2026-05-07 03:13:44 +00:00
xiaoju d472de1247 refactor: three-phase context (Moderator/Agent/Extract) + extractPrompt + unified ExtractFn
- ModeratorContext → AgentContext → ExtractContext progressive types
- ThreadContext is now alias for AgentContext
- RoleDefinition adds extractPrompt field
- ExtractFn = (schema, ctx: ExtractContext) => Promise<T>
- createWorkflow takes ExtractFn, engine loop: moderator → agent → extract
- Remove ExtractConfig, extractMetaOrThrow, extract-meta.ts

小橘 <xiaoju@shazhou.work>
2026-05-07 01:05:31 +00:00
xiaoju 2482fb7e62 chore: remove all dryRun infrastructure
dryRun no longer needed — tests use mock agents + mock fetch instead.
Removes isDryRun from WorkflowFnOptions, dryRun from ExtractConfig,
dryRunMeta from RoleDefinition, --dry-run from CLI, and all related
plumbing in engine/worker/fork/extract.

小橘 <xiaoju@shazhou.work>
2026-05-06 14:25:44 +00:00
xiaoju fa9163e462 refactor: all-agentic architecture — roles as pure data, agent binding at runtime
BREAKING: Major architecture change.

- RoleDefinition = pure data (systemPrompt + schema + dryRunMeta)
- AgentFn = (ctx: ThreadContext) => Promise<string>, reads ctx.currentRole
- WorkflowDefinition decoupled from agents, bound via AgentBinding at runtime
- createWorkflow(def, binding, extract) replaces createRoleModerator
- Meta extraction moved into engine loop
- Delete workflow-util-role package (createRole, decorators, extract all gone)
- Role packages become pure data exports
- Agent packages updated to single-arg AgentFn

小橘 <xiaoju@shazhou.work>
2026-05-06 14:14:33 +00:00
xiaoju c9cdfe37db refactor: extract planner and coder into standalone role packages
- New @uncaged/workflow-role-planner (phaseSchema, createPlannerRole)
- New @uncaged/workflow-role-coder (coderMetaSchema, createCoderRole)
- solve-issue template imports from new packages, keeps dry-run defaults

小橘 <xiaoju@shazhou.work>
2026-05-06 11:40:19 +00:00
xiaoju 45bb5af99a feat: per-role agent config + phased planner/coder in solve-issue template
- SolveIssueRolesConfig.agents allows per-role AgentFn overrides
- PlannerMeta now outputs phases (name, description, acceptance)
- CoderMeta reports completedPhase, works one phase at a time
- Moderator routes coder→coder until all phases done, then reviewer

小橘 <xiaoju@shazhou.work>
2026-05-06 11:35:45 +00:00
xiaoju c7b0beb6be refactor: unify RoleDefinition + WorkflowDefinition with description & schema
- Add RoleDefinition<Meta> = { description, run, schema } to core types
- WorkflowDefinition now carries description and RoleDefinition per role
- Add buildDescriptor() in core to derive WorkflowDescriptor from WorkflowDefinition
- Remove buildDescriptorFromRoles / RoleDescriptorInput from workflow-util-role
- Update solve-issue template, examples, and all tests

小橘 <xiaoju@shazhou.work>
2026-05-06 11:19:49 +00:00
xiaoju 196562c82a feat: committer distinguishes recoverable vs unrecoverable failures
CommitterMeta is now a 3-way discriminated union:
- committed: success with branch + commitSha
- recoverable: coder can fix (hook failures, lint, test, conflicts)
- unrecoverable: can't be fixed by code (auth, permissions, disk)

Moderator routes recoverable → coder for retry.
2026-05-06 10:53:17 +00:00
xiaoju aee71fd2e7 refactor: simplify committer — minimal prompt, config simplified
- CommitterGitConfig → CommitterConfig { cwd } (remote dropped)
- threadId from ctx.threadId, not config
- Remove summarizeThreadContext — agent uses uncaged-workflow thread CLI
- Prompt doesn't teach git or require structured output
2026-05-06 10:39:09 +00:00
xiaoju 4b27943871 refactor: simplify reviewer — discriminated union meta, minimal prompt
- ReviewerMeta → discriminated union: approved | rejected (with issues)
- Remove method-heavy prompt — agent has built-in code review capability
- Prompt now just says: project path, threadId for context, approve or reject
- No non-blocking suggestions (they get ignored anyway)
- ReviewerConfig simplified to just { cwd }
2026-05-06 10:29:48 +00:00
xiaoju 513c006ce3 refactor: rename workflow-role-llm → workflow-agent-llm
The package only contains createLlmAdapter (OpenAI chat → AgentFn),
which is an agent adapter, not a role. Aligns with workflow-agent-cursor
and workflow-agent-hermes naming.
2026-05-06 10:14:35 +00:00
xingyue 94fa964b84 fix(workflow): add typecheck script and fix remaining type errors
- Add typecheck script (bunx tsc --build) to package.json
- Update check script to run typecheck before biome
- Fix mock fetch casts in test files (bun-types preconnect)
- Fix RequestInfo → Request | string | URL in llm-extract test
- Fix ThreadContext generic cast in solve-issue-template test
- Fix git-exec.ts missing return and module resolution
- Remove @types/node from workflow-role-committer
- Add exports field to workflow-util-agent/package.json
2026-05-06 18:10:25 +08:00
xiaoju 2c642b1a53 refactor: committer as pure agent role with discriminated union meta (#17)
- Remove hardcoded git commands (git-exec.ts deleted)
- CommitterMeta is now a discriminated union: committed | failed
- Agent handles git operations, committer extracts structured result
- Moderator can route on meta.status for retry logic

Closes #17
2026-05-06 10:06:27 +00:00
xingyue 98b6153070 fix(workflow): resolve type errors across all packages and remove tsbuildinfo from tracking
- Add bun-types and @types/xxhashjs to root devDependencies
- Add types: ['bun-types'] to root and package tsconfigs
- Fix AST Node type narrowing in bundle-validator.ts with AcornNode type and narrowNode helper
- Fix generic ThreadContext variance in create-role-moderator.ts
- Add explicit parameter types in worker.ts
- Fix ChildProcess type in worker-spawn.ts to match spawn() stdio config
- Remove all tsconfig.tsbuildinfo from git tracking
- Add tsconfig.tsbuildinfo to .gitignore
2026-05-06 17:41:11 +08:00
xiaoju c04e7c31af refactor: move llmExtract, extractMeta, buildDescriptor, types to workflow-util-role
workflow-role-llm now only contains LLM-as-agent specifics:
  - createRole (wires agent + extract)
  - createLlmAdapter (OpenAI chat completions agent)

workflow-util-role now provides all role infrastructure:
  - decorators (decorateRole, withDryRun, onFail)
  - llmExtract / extractMetaOrThrow (structured extraction)
  - buildDescriptorFromRoles (zod → JSON Schema)
  - LlmProvider, LlmMessage types
2026-05-06 08:13:27 +00:00
xiaoju 6e62c7458d refactor: remove schemaDefaults, use caller-provided dryRunMeta
Cursor completed removal of schemaDefaults. All dry-run paths now
use explicit dryRunMeta from the caller.
2026-05-06 08:10:40 +00:00
xiaoju 2a71454c10 refactor: extract @uncaged/workflow-util-agent + smart prompt
- New package: spawn-cli + build-agent-prompt shared utils
- Smart prompt: start + meta summaries for middle steps + last step full
- Cursor/Hermes adapters now import from util-agent (no duplicate code)
- 109 tests pass, biome clean

Closes #14
小橘 <xiaoju@shazhou.work>
2026-05-06 07:17:59 +00:00
xiaoju db5cbd49e2 feat: @uncaged/workflow-template-solve-issue — first workflow template
planner → coder → reviewer → committer flow with retry logic.
- createSolveIssueWorkflow factory (agent-agnostic)
- buildSolveIssueDescriptor with zod@4 JSON Schema
- Moderator: reviewer reject → coder retry, maxRounds → END
- 103 tests pass, biome clean

Closes #13
小橘 <xiaoju@shazhou.work>
2026-05-06 07:04:28 +00:00