dsh-ccswitch-sync

Juggling CC Switch and DeepSeek Harness at the same time means repeating a tedious ritual: pick a provider and model in CC Switch, then open DSH settings to manually create a custom provider, wire up the base URL, paste the API key, choose the protocol, and list the models one by one — every single time you switch.
dsh-ccswitch-sync collapses that entire ritual into a single sync. It reads cc-switch.db (SQLite) directly without requiring CC Switch to be running, maps each non-official provider to a llm-pi-ai custom route (default prefix cc-), auto-detects the protocol (openai-responses or openai-completions for Codex, anthropic-messages for Claude), and assembles the model list from the provider catalog, the gateway /models endpoint, or a local catalog file. API keys are stored as credential references in .credentials.yaml rather than leaking into settings.yaml, and reasoning levels are carried over from the model catalog into the selector. Syncing is strictly opt-in — it will not silently overwrite your manual edits on restart or when CC Switch changes — and the excludeRoutes option lets you carve out specific routes you want to keep untouched.
It is built for users who juggle multiple providers and especially rely on gateways like Sub2API or JingHua Coding, where the configured model list shows only a handful of entries while the gateway actually exposes many more. The plugin calls GET /models on each base URL to discover the full set of chat models (filtering out image, realtime, audio, and embedding types), caches the result for 24 hours on disk, and falls back gracefully to the explicitly configured models when an endpoint does not support discovery.