drew
6178be3aa7
.opencode/models/models.yaml is now the ONE file humans edit to assign a model to an agent. tools/sync_models.py regenerates every derived surface — the .opencode/models/*.txt files, opencode.json's `agent` block, and each non-tier agent's .md `model:` frontmatter — so an assignment cannot drift across surfaces. `--check` verifies with no writes and is enforced in CI by test_model_registry_in_sync_with_manifest. Hardened after adversarial review: - Deletes orphan <agent>.txt files left behind when an override is dropped from the manifest. The dispatcher's resolver reads <agent>.txt before default.txt, so a stale file would silently pin the old model. Tier .txt files are left to sync_tier_models.py. - Rejects a manifest key that does not name a real agent (no matching .opencode/agents/<name>.md) instead of silently appending a bogus opencode.json entry and leaving the real agent on the default model. - Validates the regenerated opencode.json BEFORE writing it, so a bad render aborts cleanly instead of corrupting the file on disk. Tier-ladder agents (task-implementor-tier-*) remain governed separately by tiers.yaml + sync_tier_models.py and are passed through untouched. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
40 lines
1.8 KiB
YAML
40 lines
1.8 KiB
YAML
# Model registry — single source of truth for per-agent LLM model
|
|
# assignment.
|
|
#
|
|
# Humans edit ONLY this file. `tools/sync_models.py` regenerates every
|
|
# derived artifact so a model assignment can never drift across surfaces:
|
|
#
|
|
# - .opencode/models/<agent>.txt — read by the dispatcher's resolver
|
|
# (_opencode_worker._resolve_role_model)
|
|
# - .opencode/models/default.txt — the fallback model
|
|
# - the `agent` block of opencode.json — read by OpenCode at startup
|
|
# (the channel that drives generation)
|
|
#
|
|
# opencode.json points each agent at `{file:.../<agent>.txt}`, so both
|
|
# consumers read the SAME generated file — the generation model and the
|
|
# dispatcher's recorded model cannot disagree.
|
|
#
|
|
# The pipeline launcher runs `tools/sync_models.py` at every startup
|
|
# (before OpenCode caches opencode.json), so editing this file + relaunch
|
|
# is all an operator needs. Run `python3 tools/sync_models.py` by hand to
|
|
# apply immediately, or `--check` to verify with no writes.
|
|
#
|
|
# Tier-ladder agents (task-implementor-tier-*) are governed separately by
|
|
# tiers.yaml + tools/sync_tier_models.py; sync_models.py leaves their
|
|
# opencode.json entries untouched.
|
|
#
|
|
# Every value is `providerID/modelID`; the provider MUST be declared in
|
|
# opencode.json's `provider` block.
|
|
|
|
# Fallback model for every agent not listed under `agents:` below.
|
|
default: local-claude/claude-haiku-4-5
|
|
|
|
# Per-agent overrides. Only agents that need a NON-default model appear
|
|
# here; every other agent resolves to `default` above.
|
|
agents:
|
|
conflict-resolver-worker: local-claude/claude-opus-4-6
|
|
estimator-implementation: local-claude/claude-sonnet-4-6
|
|
pr-review-worker: local-claude/claude-sonnet-4-6
|
|
pr-review-worker-dispute: local-claude/claude-opus-4-6
|
|
ca-test-infra-improver: google/gemini-2.5-pro
|