The Docker image's mcp-server process crashes on boot with
ERR_PNPM_WORKSPACE_PKG_NOT_FOUND (#314). supervisord starts it via
`pnpm start`, and pnpm 10 runs a dependency-sync check before the script.
From the isolated /app/mcp-server dir it tries to resolve
`@prompt-optimizer/core@workspace:*`, can't find the workspace sibling,
and aborts before node ever runs.
The built artifact is self-contained (deps bundled, NODE_PATH=/app), so the
pnpm wrapper is unnecessary. Call node directly, matching the `start` script:
node -r ./preload-env.cjs dist/start.cjs --transport=http --port=3000
Verified: mcp-server reaches RUNNING state and serves /mcp on a fresh build
from develop.
Fixes#314
- Add MiniMax-M3 to the model selection list and set as the new default
- Retain MiniMax-M2.7 and MiniMax-M2.7-highspeed as alternatives
- Remove deprecated MiniMax-M2.5 and MiniMax-M2.5-highspeed from the list
- Update unit tests to match the new model lineup
- Align Cloudflare Workers setup across README and user docs
- Clarify build command cleanup and deployment context notes
- Keep frontend API key exposure warnings concise across providers
- Register Chrome built-in AI as a local text provider with availability checks.
- Add explicit setup/download gating and auto-enable safeguards.
- Preserve structured prompt roles through initialPrompts and stream tests.
- Keep no-key and local model sources disabled until explicit user configuration exists.
- Tighten default text and image model enablement checks.
- Cover environment-driven defaults for local providers.
- Split extension release builds from web env loading.
- Add a release secret guard for extension packaging.
- Keep web and extension Vite configs from leaking incompatible env.
- Parse Python-style JSON literals for custom model parameters
- Add JSON/string format controls for custom parameter inputs
- Add model parameter docs and locale strings
- Commit basic and context optimization session updates
- Persist image and pro conversation optimization results
- Restore sessions safely after history reload
- Link test source cues back to prompt panel areas
- Add source-linked cards, variant tags, and feedback composables
- Add image model quick switching and workspace control polish
- Add custom request header parsing and provider metadata resolution
- Add quick switching controls for text model workspaces
- Keep model UI labels and custom header output tidy
- Add prompt garden inspiration suggestions for image workspaces
- Use public API paths and locale-aware suggestion loading
- Preserve category paths when importing favorites from Prompt Garden
