- Extract core logic into testable performRollback(opts) function
- CLI action now delegates to performRollback and prints output
- Tests cover: dry-run, event export/delete, git checkout, missing
commit error, noop when already at target, dump metadata
- All tests use tmpdir, no dependency on ~/.upulse
- WorkflowCheckpoint caches lastEventId, topic summaries, and per-topic events
- First tick: full read (getAfter(0)), subsequent ticks: incremental (getAfter(lastEventId))
- Newly written role events also update checkpoint in-place
- No interface changes, no new files
- ENGINE_DIR = ~/.upulse/engine for meta coder/reviewer/tester/promoter
- REPO_DIR still used for coding workflow
- Tester accepts buildCmd/testCmd opts, no longer hardcoded to monorepo paths
- Init engine dir with package.json + tsconfig.json + git repo
Tester was failing with ENOENT because cwd pointed to
/home/azureuser/repos/packages/pulse (nonexistent) instead of
/home/azureuser/repos/pulse/packages/pulse.
- cooldownMs: 0 for both coding and meta workflow rules
- logTick wrapper prints timestamp + elapsed ms per tick
- start-daemon.sh loads secrets before starting bun
- pulse-workflow.service systemd unit (enabled + started)
- ArchitectMeta now includes targetFiles, changes (file→description), and verification criteria
- CoderMeta prepPrompt explicitly consumes architect.changes, architect.verification, and reviewer.rejectionReason on retry
- ReviewerMeta now outputs verdict, rejectionReason, and monotonic retryCount (starting from 0)
- Moderator enforces retryCount < 3 limit — forces closer when exceeded
- Updated mock implementations, role factories, and e2e to match new meta structures
- Added test for retry limit (retryCount >= 3 → closer)
- analyst: structured JSON output via tool_choice (score, highlights, issues, roleAnalysis with eventId refs)
- renderer: pure code HTML template — dark theme, gantt bar, role cards, verdict badges
- report-live.ts: e2e test script, reads coding workflow DB → generates HTML report
- renderer-llm.ts: kept as reference (LLM version, not used — timeout issues)
- 2 unit tests pass, live test: 17.9s end-to-end
- council-v2-live.ts: coder/reviewer now use createCoderRole/createReviewerRole
- Removed inline runCursorAgent (moved to agent-executor.ts)
- reviewer verdict via LLM₂ tool_choice instead of includes('rejected')
- Live verified: architect→coder→reviewer(approved)→closer in 135s
- createAgentExecutorRole: prepPrompt → agent.run → LLM₂ tool_choice → meta
- createCursorRunner: extracts Cursor CLI into AgentRunner interface
- Coder role: LLM₂ extracts filesChanged + testsPassed via tool_choice
- Reviewer role: LLM₂ extracts verdict via tool_choice (no more includes hack)
- defaultMeta fallback when LLM₂ fails or returns no tool_call
- 5 new agent-executor tests, 24 total pass
Simplified structure with inline code block for WorkflowType,
streamlined section headings, and tighter prose while retaining
all key topics: START/END automaton, pure roles, Moore-machine
diff-driven ticks, and source file references.
Made-with: Cursor
- Role<Meta> = (chain, topicId, store) => Promise<Meta>
- MetaOf<R> extracts Meta from Role type
- RoleOutput<Roles> = discriminated union { role: K, meta: MetaOf<Roles[K]> }
- Moderator receives RoleOutput, switch on role narrows meta type
- WorkflowType simplified: no projection field, just name + roles + moderator
- Adapter builds per-topic summary internally
- coding-workflow: ArchitectMeta, CoderMeta, ReviewerMeta, CloserMeta
- -538/+326 lines — simpler AND more type-safe
Add 'upulse topic create' to write coding.created events with
auto-generated slugified topicId, and 'upulse topic list' to
display topics grouped by state with --all flag for closed topics.
Made-with: Cursor
xiaoju
xiaomo
xingyue
luming