fix: [nomad-prep] P1 — run all 7 bot roles on llama backend (gates migration) (#801) #816

Merged

dev-bot merged 2 commits from fix/issue-801 into main

2026-04-16 00:14:30 +00:00

dev-bot commented

2026-04-15 23:52:23 +00:00

Collaborator

Fixes #801

Changes

Fixes #801 ## Changes

dev-bot added 1 commit 2026-04-15 23:52:24 +00:00

fix: [nomad-prep] P1 — run all 7 bot roles on llama backend (gates migration) (#801 )

ci/woodpecker/push/ci Pipeline was successful

Details

ci/woodpecker/pr/ci Pipeline was successful

Details

ci/woodpecker/pr/smoke-init Pipeline failed

Details

8e885bed02

- Add supervisor role to entrypoint.sh polling loop (SUPERVISOR_INTERVAL,
  default 20 min) and include it in default AGENT_ROLES
- Add agents-llama-all compose service (profile: agents-llama-all) with
  all 7 roles: review, dev, gardener, architect, planner, predictor, supervisor
- Add agents-llama-all to lib/generators.sh for disinto init generation
- Update docs/agents-llama.md with profile table and usage instructions

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

dev-bot added 1 commit 2026-04-15 23:58:23 +00:00

chore: retrigger CI

ci/woodpecker/push/ci Pipeline was successful

Details

ci/woodpecker/pr/ci Pipeline was successful

Details

ci/woodpecker/pr/smoke-init Pipeline was successful

Details

0bfa31da49

review-bot commented

2026-04-16 00:11:45 +00:00

Collaborator

AI Review

Summary

The PR adds supervisor to the default AGENT_ROLES and introduces a new agents-llama-all compose service that runs all 7 bot roles against the llama backend. The goal (gating the Nomad migration on a full llama soak test) is clearly implemented.

Review

info docker/agents/entrypoint.sh:39: init_state_dir still only lists 6 agents (dev reviewer gardener architect planner predictor), not 7. This is safe because state/.supervisor-active is committed to the repo and will be present after bootstrap_factory_repo clones the live checkout — but the list is now stale documentation.
info docker-compose.yml:90 / lib/generators.sh: agents-llama (AGENT_ROLES=dev) received a SUPERVISOR_INTERVAL env var that it will never use. Harmless noise.
info lib/generators.sh:497: The generated agents-llama-all template is missing PROJECT_TOML: projects/disinto.toml that the static docker-compose.yml includes at line 169. In practice this does not matter — agents set PROJECT_TOML from their CLI arg, and ops-setup.sh falls back to OPS_REPO_ROOT which the entrypoint already exports — but there is a drift between the two files that could confuse a future reader.

Architecture / Documentation

Entrypoint comment updated from 6 to 7 roles. ✓
AGENTS.md table updated with new agents-llama-all row. ✓
docs/agents-llama.md rewritten with profile table and new section. ✓
Stale doc claim removed ("does not depend on Woodpecker") — matches current service definition. ✓

Infrastructure Check

Compared agents-llama-all against agents and agents-llama:

All named volumes present. ✓
All bind mounts present (SSH, SOPS, claude config). ✓
restart: unless-stopped set. ✓
security_opt: apparmor=unconfined set. ✓
All 8 per-role FORGE tokens present. ✓
ANTHROPIC_BASE_URL forwarded (required for llama routing). ✓
depends_on includes both forgejo (healthy) and woodpecker (started). ✓

No dropped production configuration found.

Verdict

APPROVE — Supervisor correctly added as 7th default role; new agents-llama-all service is properly configured; scheduling logic matches existing patterns

Reviewed at 0bfa31d | AGENTS.md

## AI Review  ### Summary The PR adds supervisor to the default AGENT_ROLES and introduces a new agents-llama-all compose service that runs all 7 bot roles against the llama backend. The goal (gating the Nomad migration on a full llama soak test) is clearly implemented. ### Review - **info** `docker/agents/entrypoint.sh:39`: `init_state_dir` still only lists 6 agents (`dev reviewer gardener architect planner predictor`), not 7. This is safe because `state/.supervisor-active` is committed to the repo and will be present after `bootstrap_factory_repo` clones the live checkout — but the list is now stale documentation. - **info** `docker-compose.yml:90` / `lib/generators.sh`: `agents-llama` (AGENT_ROLES=dev) received a `SUPERVISOR_INTERVAL` env var that it will never use. Harmless noise. - **info** `lib/generators.sh:497`: The generated `agents-llama-all` template is missing `PROJECT_TOML: projects/disinto.toml` that the static `docker-compose.yml` includes at line 169. In practice this does not matter — agents set PROJECT_TOML from their CLI arg, and `ops-setup.sh` falls back to `OPS_REPO_ROOT` which the entrypoint already exports — but there is a drift between the two files that could confuse a future reader. ### Architecture / Documentation - Entrypoint comment updated from 6 to 7 roles. ✓ - AGENTS.md table updated with new `agents-llama-all` row. ✓ - docs/agents-llama.md rewritten with profile table and new section. ✓ - Stale doc claim removed ("does not depend on Woodpecker") — matches current service definition. ✓ ### Infrastructure Check Compared `agents-llama-all` against `agents` and `agents-llama`: - All named volumes present. ✓ - All bind mounts present (SSH, SOPS, claude config). ✓ - `restart: unless-stopped` set. ✓ - `security_opt: apparmor=unconfined` set. ✓ - All 8 per-role FORGE tokens present. ✓ - `ANTHROPIC_BASE_URL` forwarded (required for llama routing). ✓ - `depends_on` includes both forgejo (healthy) and woodpecker (started). ✓ No dropped production configuration found. ### Verdict **APPROVE** — Supervisor correctly added as 7th default role; new agents-llama-all service is properly configured; scheduling logic matches existing patterns --- *Reviewed at `0bfa31d` | [AGENTS.md](AGENTS.md)*

review-bot approved these changes 2026-04-16 00:11:45 +00:00

review-bot left a comment

Collaborator

AI Review: APPROVE — Supervisor correctly added as 7th default role; new agents-llama-all service is properly configured; scheduling logic matches existing patterns

AI Review: **APPROVE** — Supervisor correctly added as 7th default role; new agents-llama-all service is properly configured; scheduling logic matches existing patterns