v0.1.7
Highlights
APScheduler-backed schedule wakes, parallel tool UI fixes, and clearer LLM errors since v0.1.6.
Scheduler
- APScheduler (
AsyncIOScheduler) wakes jobs; SQLite stays source of truth - Interval / cron / one-shot triggers register on create and reload on boot
- Claim-before-fire + 15m safety sweep for missed/stale claims
- Agent
scheduletool and run/pause/resume API unchanged
Chat / tools UI
- Tool results match by
call_id(parallel RO tools no longer stick on “running”) - Parallel delegates and subagent inspector use the same matching
- ATG wave tools get stable
atg:<node>ids - History mid-turn refresh pairs outputs by
call_id
LLM errors
- Provider body/status extracted into readable messages (not blank “LLM request failed:”)
- Hints for timeout, context overflow, rate limit, auth
- Default HTTP timeout raised to 300s (
llm_timeout_secondssetting) for long answers - Empty streams called out as early disconnect / timeout
Install
App (managed):
curl -fsSL https://raw.githubusercontent.com/Alg0rix/tomo/v0.1.7/scripts/install.sh | bashConnector:
curl -fsSL https://raw.githubusercontent.com/Alg0rix/tomo/v0.1.7/scripts/install-connector.sh | bash
# pin: TOMO_CONNECTOR_VERSION=v0.1.7Wheel: tomo-0.1.7-py3-none-any.whl (attached)
What's Changed
- feat: APScheduler wakes, parallel tool UI, clearer LLM errors
Full Changelog: v0.1.6...v0.1.7