admin/orchestrator
1
0
Fork 0
Code Issues Pull Requests 4 Actions Packages Projects Releases Wiki Activity

feat(launcher): drop dead frontmatter model + validate model name (never-break)

Browse Source 
G1: remove the dead `model:` line from all 6 .openclaw/agents/*.md prompts —
launcher never read it; config (agent_model_*) is the single source of truth.

G2: add is_valid_model helper (format check ^claude-…$) applied inside
resolve_agent_model's resolution cascade and at the inline --fallback-model
read in _spawn. An invalid name is logged and skipped to the next valid level
(in the limit: no --model flag), never passed to the CLI, never raises. Format
check chosen over an allowlist for forward-compatibility (ADR-001).

G3 (routing) and G4 (fallback) intentionally NOT enabled — all agents stay on
claude-opus-4-8; agent_fallback_model stays "".

Docs (golden source) updated in the same change: README model/effort table +
validation, CLAUDE.md, .env.example (ORCH_AGENT_MODEL_*/EFFORT_*/FALLBACK_MODEL),
CHANGELOG. Tests: test_agent_frontmatter_no_model.py (G1), extended
test_resolve_agent_model.py (G2 never-break).

Refs: ORCH-074
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
This commit is contained in:
claude-bot
2026-06-08 21:53:09 +03:00
committed by stream
parent 0c240198e4
commit 0873803faa
13 changed files with 288 additions and 17 deletions
Download Patch File Download Diff File Expand all files Collapse all files

38
.env.example
View File

@@ -12,6 +12,44 @@ ORCH_GITEA_WEBHOOK_SECRET=
ORCH_CLAUDE_BIN=/usr/bin/claude
ORCH_REPOS_DIR=/home/slin/repos
ORCH_DB_PATH=/app/data/orchestrator.db
# ── Agent model / effort / fallback (ORCH-41, validation ORCH-74) ─────────────
# Per-agent LLM model + reasoning effort, resolved by launcher.resolve_agent_*.
# Resolution priority (per agent): project-override (projects_json agent_models/
# agent_efforts) > ORCH_AGENT_MODEL_<AGENT> / ORCH_AGENT_EFFORT_<AGENT> >
# ORCH_AGENT_MODEL_DEFAULT / ORCH_AGENT_EFFORT_DEFAULT > CLI default (no flag).
# The frontmatter `model:` in .openclaw/agents/*.md is DESCRIPTIVE only and is NOT
# read — config below is the single source of truth for the model (ORCH-74 G1).
#
# ORCH-74 (G2): a resolved MODEL name is validated (^claude-…$ format check) before
# it reaches --model. A structurally invalid name (typo, gpt-4, empty) is logged and
# the next valid level is used (in the limit: no --model flag). Forward-compatible:
# a future claude-* version passes without editing any allowlist. EFFORT is validated
# against low|medium|high|xhigh|max (ORCH-41); an invalid effort is dropped.
#
# All 6 agents resolve to claude-opus-4-8 (model-routing G3 NOT enabled). Leave the
# per-agent overrides empty to use the default. Do NOT hardcode the model version
# anywhere except ORCH_AGENT_MODEL_DEFAULT.
ORCH_AGENT_MODEL_DEFAULT=claude-opus-4-8
ORCH_AGENT_MODEL_ANALYST=
ORCH_AGENT_MODEL_ARCHITECT=
ORCH_AGENT_MODEL_DEVELOPER=
ORCH_AGENT_MODEL_REVIEWER=
ORCH_AGENT_MODEL_TESTER=
ORCH_AGENT_MODEL_DEPLOYER=
# Effort split: thinking agents (analyst/architect/developer/reviewer) -> high;
# mechanical agents (tester/deployer) -> medium.
ORCH_AGENT_EFFORT_DEFAULT=high
ORCH_AGENT_EFFORT_ANALYST=high
ORCH_AGENT_EFFORT_ARCHITECT=high
ORCH_AGENT_EFFORT_DEVELOPER=high
ORCH_AGENT_EFFORT_REVIEWER=high
ORCH_AGENT_EFFORT_TESTER=medium
ORCH_AGENT_EFFORT_DEPLOYER=medium
# Optional --fallback-model used when the primary is overloaded. Empty -> no flag
# (G4 NOT enabled, ADR-001 ORCH-74: determinism — all agents stay on opus-4-8). A
# non-empty value is validated by the SAME predicate as the model; a typo is dropped.
ORCH_AGENT_FALLBACK_MODEL=
# ORCH-042/ORCH-067: live-tracker mode. bump (DEFAULT since ORCH-067) -> on every
# update the old card is deleted and a fresh one is sent silently to the BOTTOM of
# the chat (deleteMessage + sendMessage + repoint), so the current status is always