loom-ops

Durable runbook ops agent on the Loom stack — ops sibling of loom-run, built on loom-runner with flow-xray traces.

Run нельзя потерять: упал на шаге 12/40 → resume, каждый tool call в SQLite, postmortem через explain + flow-xray.

Loom stack (this repo’s place)

The stack is a pyramid, not five equal frameworks. Tail-call optimization is the primitive, runner is the durable runtime, xray is the microscope, and the apps prove the stack in real workflows.

loom-ops is the ops/runbook proof app. It depends on loom-runner, loom-tailcalls, and flow-xray; none of those lower layers depend on this repo.

Full map: docs/STACK.md. Dev assistant → use loom-run; ops runbooks → loom-ops.

Install

From PyPI (after publish — see releases):

pip install "loom-ops[api,telegram]"
loom-ops supervise "incident: API latency spike" --run-id inc-001 --db ops.sqlite --mock-llm --workspace .

PyPI workflow: .github/workflows/publish.yml on GitHub Release. Requires PYPI_API_TOKEN with upload scope for project loom-ops.

From source (development):

git clone https://github.com/kroq86/loom-ops.git
cd loom-ops
python3.13 -m venv .venv && source .venv/bin/activate
pip install -e ".[dev,api,telegram]"

Demo (mock LLM, no API key)

# Supervisor team: planner → parallel executor + verifier
loom-ops supervise "incident: API latency spike" --run-id inc-001 --db ops.sqlite --mock-llm --workspace .

# Inspect verifier subagent audit trail
loom-ops explain --run-id inc-001:sub:verifier --db ops.sqlite --workspace .

# Single runbook walk
loom-ops runbook "incident: API latency spike" --run-id rb-001 --db ops.sqlite --mock-llm --workspace .

# HITL: executor pauses for approval (see tests/test_supervisor_ops.py)
loom-ops approve --run-id exec-1 --note "LGTM" --role executor --db ops.sqlite --mock-llm --workspace .

# Allowlisted shell (optional)
export LOOM_OPS_ALLOW_SHELL=1
# uses workspace/ops.shell.allowlist.json by default
loom-ops runbook "deploy check" --run-id rb-shell --db ops.sqlite --mock-llm --workspace .

# HTTP SSE
loom-ops serve --db ops.sqlite --mock-llm --workspace .
curl -N -X POST http://127.0.0.1:8766/supervise \
  -H 'Content-Type: application/json' \
  -d '{"message":"incident: API latency spike","run_id":"inc-http","max_steps":20}'
curl http://127.0.0.1:8766/runs/inc-http/explain

# Telegram gateway (optional extra)
export TELEGRAM_BOT_TOKEN=...
export TELEGRAM_ALLOWED_CHAT_IDS=123456789
loom-ops telegram --db ops.sqlite --mock-llm --workspace .

Ops tools

Docs

• Loom stack integration
• Product positioning
• Runbooks
• Environment variables
• Security
• Telegram gateway

Tests

python -m pytest -q

License

MIT