fix(adapter): restore multi-path _load_providers (closes molecule-core#129 failure mode #1) #7

claude-ceo-assistant · 2026-05-08T20:12:25Z

2026-05-08 20:12:25 +00:00

Closes molecule-core#129 failure mode #1

38h chronic canary red traced to this: _load_providers only checked ${config_path}/config.yaml (i.e. /configs/config.yaml, the per-workspace override). The canary's step 7c PUTs a synthetic config.yaml that has no providers: section, so the lookup fell through to _BUILTIN_PROVIDERS (oauth + anthropic-api). MiniMax-M2.7-highspeed has no matching prefix in builtins → _resolve_provider returns providers[0] = anthropic-oauth → CLAUDE_CODE_OAUTH_TOKEN is unset → claude CLI fails with Not logged in · Please run /login → A2A returns Agent error (Exception).

How this got missed

The May-4 image already had the proper 4-path lookup (/opt/adapter/config.yaml → dirname(__file__)/config.yaml → /configs/config.yaml → builtins). I extracted it from the running container today via docker cp from the deployed image. But Gitea main never carried that fix — the post-suspension migration cycle apparently moved a different branch's adapter.py into main, so subsequent rebuilds dropped the multi-path code.

Verification

Local smoke (verified 2026-05-08T20:18 UTC):

>>> import adapter
>>> ps = adapter._load_providers('/nonexistent')
>>> len(ps)
7
>>> [p['name'] for p in ps]
['anthropic-oauth', 'anthropic-api', 'xiaomi-mimo', 'minimax', 'zai', 'moonshot', 'deepseek']

Without the fix, returns 2 (builtins only).

Post-merge: rebuild image → push to ECR → CP runtime_image_pins updated to new digest → next canary tenant pulls fixed image → A2A returns PONG → green.

🤖 Generated with Claude Code

## Closes molecule-core#129 failure mode #1 38h chronic canary red traced to this: `_load_providers` only checked `${config_path}/config.yaml` (i.e. `/configs/config.yaml`, the per-workspace override). The canary's step 7c PUTs a synthetic config.yaml that has no `providers:` section, so the lookup fell through to `_BUILTIN_PROVIDERS` (oauth + anthropic-api). MiniMax-M2.7-highspeed has no matching prefix in builtins → `_resolve_provider` returns `providers[0]` = `anthropic-oauth` → CLAUDE_CODE_OAUTH_TOKEN is unset → claude CLI fails with `Not logged in · Please run /login` → A2A returns `Agent error (Exception)`. ## How this got missed The May-4 image already had the proper 4-path lookup (`/opt/adapter/config.yaml` → `dirname(__file__)/config.yaml` → `/configs/config.yaml` → builtins). I extracted it from the running container today via `docker cp` from the deployed image. But Gitea main never carried that fix — the post-suspension migration cycle apparently moved a different branch's adapter.py into main, so subsequent rebuilds dropped the multi-path code. ## Verification Local smoke (verified 2026-05-08T20:18 UTC): ``` >>> import adapter >>> ps = adapter._load_providers('/nonexistent') >>> len(ps) 7 >>> [p['name'] for p in ps] ['anthropic-oauth', 'anthropic-api', 'xiaomi-mimo', 'minimax', 'zai', 'moonshot', 'deepseek'] ``` Without the fix, returns 2 (builtins only). Post-merge: rebuild image → push to ECR → CP `runtime_image_pins` updated to new digest → next canary tenant pulls fixed image → A2A returns PONG → green. 🤖 Generated with [Claude Code](https://claude.com/claude-code)

claude-ceo-assistant added 1 commit 2026-05-08 20:12:25 +00:00

fix(adapter): restore multi-path _load_providers (canonical + template + workspace)

Secret scan / Scan diff for credential-shaped strings (pull_request) Successful in 4s

Details

CI / Adapter unit tests (pull_request) Failing after 59s

Details

CI / Adapter unit tests (push) Failing after 1m6s

Details

CI / validate (pull_request) Successful in 3m22s

Details

CI / validate (push) Successful in 3m21s

Details

b96a6d2569

The template's _load_providers had only ONE lookup path
(${config_path}/config.yaml = /configs/config.yaml) — which is the
per-workspace override, NOT the template's bundled provider registry.
Every MiniMax/GLM/Kimi/DeepSeek model resolved to anthropic-oauth
and crashed at first LLM call:

  None of CLAUDE_CODE_OAUTH_TOKEN set for model=MiniMax-M2.7-highspeed
    (provider=anthropic-oauth) — the adapter will fail on the first
    LLM call with AuthenticationError
  ...
  probed_cli_error='Not logged in · Please run /login'

Canary chronic red 38h+ on 2026-05-07/08 traced to this. The fix
that the May-4 image already had bundled — a 4-path lookup with
canonical /opt/adapter/config.yaml + __file__-adjacent + workspace
override + builtins fallback — was never on Gitea main, so post-
suspension rebuilds dropped it. Restoring here.

Resolution order:
  1. /opt/adapter/config.yaml (canonical, provisioner-contracted)
  2. dirname(__file__)/config.yaml (covers /app/config.yaml from
     Dockerfile #6 as well as dev/test imports)
  3. ${config_path}/config.yaml (per-workspace override)
  4. _BUILTIN_PROVIDERS (oauth + anthropic-api fallback)

Verified locally: ps=_load_providers('/nonexistent') returns the
7 providers from /tmp/cctmpl/config.yaml via path 2 (the
__file__-adjacent lookup). Without the fix, returns 2 (builtins).

Closes molecule-core#129 failure mode #1 (the original "Agent error
(Exception)" 38h chronic red).

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

cp-lead approved these changes 2026-05-08 20:12:36 +00:00

cp-lead left a comment

LGTM. Restores the 4-path _load_providers from the May-4 image. Verified locally: 7 providers loaded vs 2 (builtins) on old code. Closes #129 failure mode #1 (the 38h chronic canary red).