- `uwf prompt usage` now outputs only the usage skill (was three combined)
- `uwf prompt bootstrap` replaces `setup` with framework-agnostic instructions
- Remove `usage-reference` and `setup` subcommands
- Remove `generateBootstrapReference` from util (moved to cli)
Fixes#99
小橘 🍊（NEKO Team）
Address review nit from PR #97: clientInfo.version should be
agent-hermes's own version for correct identification under
independent versioning.
小橘 🍊（NEKO Team）
- All uwf-* adapter CLIs now support --version / -V
- util VERSION constant reads from package.json at runtime
- agent-hermes ACP clientInfo uses dynamic VERSION
小橘 🍊（NEKO Team）
Tokens (inputTokens, outputTokens) now come from ACP PromptResponse.usage
which is populated synchronously from run_conversation() — no WAL race.
Turns still come from DB before/after snapshot.
Previously both were read from hermes state.db after ACP prompt returned,
but WAL write lag caused incomplete token data (e.g. 235 vs actual 26,080).
Refs #91
When --agent is passed to uwf thread exec, try config.agents[alias]
first (e.g. 'hermes' → config.agents.hermes = {command: 'uwf-hermes'}),
then fall back to parseAgentOverride for raw command names.
Also change eval CLI default --agent from 'hermes' to 'uwf-hermes'
so it works without config alias lookup.
Refs #91
The extract pipeline stores step output as a JSON object in CAS,
but the frontmatter judge only checked for raw markdown strings.
Now accepts both formats: parsed objects check $status directly,
raw strings go through YAML frontmatter extraction.
Fixes eval frontmatter-compliance scoring 0 on valid outputs.
On session resume, turns/inputTokens/outputTokens were cumulative
(entire session history) instead of per-step increments. Now we
snapshot metrics before prompt, compare after, and report the delta.
Changes:
- acp-client: add getSessionId() accessor
- hermes: extract snapshotUsage() + computeUsageDelta() pure functions
- hermes: runPrompt/runHermes/continueHermes use before/after snapshots
- 9 new unit tests for usage delta computation
Refs #68
- Add inputTokens/outputTokens to HermesSessionJson type
- Query input_tokens, output_tokens from sessions table in loadHermesSessionFromDb
- Update test fixture schema with token columns
- runPrompt now reports real token counts from Hermes state.db
Refs #76, #68
All three agent packages had bin pointing to ./src/cli.ts (bun-era
leftover). Node cannot execute .ts files directly, causing
ERR_MODULE_NOT_FOUND when spawning agents.
Closes#78
Deleted:
- acp-client.integration.test.ts (3 cases)
- resume-e2e.integration.test.ts (1 case, already skipped)
These tests spawn a real hermes CLI and hit live LLM,
belonging to the eval layer (#34), not CI.
ACP protocol parsing is already covered by unit test
acp-client.test.ts.
Also removed the --exclude integration/ hack from test:ci.
Fixes#60
Three scenarios testing the full CLI pipeline:
1. Linear workflow (planner → worker → $END): CAS chain integrity
2. Loop workflow (developer ↔ reviewer): moderator routing through cycles
3. Role mismatch detection: agent catches routing bugs
Uses workflow add → thread start → thread exec with uwf-mock,
verifying CAS state, thread lifecycle, and error handling.
Updated assertions to use getThread().status === 'completed'
(aligned with PR #45 unified thread storage).
Refs #33
New package @united-workforce/agent-mock (uwf-mock CLI):
- Reads pre-scripted outputs from a YAML mock data file (--mock-data)
- Counts existing CAS chain steps to determine step index
- Validates expected role matches actual moderator routing
- Stores minimal detail node in CAS for valid step refs
- Zero LLM, instant execution, 100% deterministic
Usage in config.yaml:
agents:
mock:
command: uwf-mock
args: ["--mock-data", "./fixtures/scenario.yaml"]
Refs #33
Instead of hardcoding workflow as empty string for completed/cancelled
threads, use resolveWorkflowFromHead to get the actual workflow hash
from the CAS chain, consistent with active thread handling.
Closes#46
uwf thread resume now supports completed threads:
- Evaluates workflow graph from $START to find first role
- Clears completed state (status → idle, completedAt → null)
- Builds resume prompt with supplement context
- Full CAS chain preserved for rich context
Suspended resume behavior unchanged.
Cancelled/idle threads still rejected.
425 tests pass.
Part of #39, closes#43
- store.ts: all threads in @uwf/thread/* with status tag
- Remove HISTORY_VAR_PREFIX, ThreadHistoryLine, deleteThread
- Add loadActiveThreads, loadHistoryThreads, completeThread
- Add migrateHistoryVarsToThreadVars migration
- thread.ts: replace deleteThread+addHistoryEntry with completeThread
- shared.ts: remove findHistoryEntry fallback
- Update all tests for unified storage model
422 tests pass.
Part of #39, closes#41, closes#42
- ThreadIndexEntry gains status and completedAt fields
- createThreadIndexEntry defaults to idle/null
- normalizeThreadIndexEntry backward-compat defaults
- updateThreadHead resets to idle (衔尾蛇 resume prep)
- markThreadSuspended sets status=suspended
- New markThreadCompleted(entry, status, now) function
- serializeThreadIndexEntry includes new fields
Part of #39, closes#40
xiaoju
xiaomo
xingyue