README.md

Sync snapshot generator

These scripts produce the JSON snapshots that the openproxy sync command applies to the user's db.json. They are maintainer-only tooling — end users never need to run them. The committed snapshots live at src/core/model/sources/{9router,omniroute}.json and are embedded into the release binary via include_str!, so the CLI works fully offline.

Why a Node helper

Both upstreams ship their catalogs as JS / TS source modules with imports and helper functions. Parsing those reliably from Rust is brittle; running them as the actual modules from Node is trivial. The helper:

1. Shallow-clones (or refreshes) decolua/9router and diegosouzapw/OmniRoute into /tmp/openproxy-sync-cache/.
2. Dynamically imports each catalog module (open-sse/config/providerModels.js for 9router, open-sse/config/providerRegistry.ts for OmniRoute — the latter via tsx).
3. Normalises both into the same shape ({source, ref, providers: [...]}).
4. Writes src/core/model/sources/<source>.json.

Usage

# Refresh both snapshots (clones the upstreams to /tmp on first run):
node scripts/sync/normalize-sources.mjs

# Refresh only one:
node scripts/sync/normalize-sources.mjs --only=9router
node scripts/sync/normalize-sources.mjs --only=omniroute

# Pin a specific ref (defaults: 9router=master, omniroute=main):
node scripts/sync/normalize-sources.mjs --ref-9router=v0.4.55 --ref-omniroute=v3.8.0

# Use a local clone instead of cloning fresh:
node scripts/sync/normalize-sources.mjs \
  --src-9router=/path/to/9router \
  --src-omniroute=/path/to/OmniRoute

After running, commit the updated src/core/model/sources/*.json files alongside any Rust code changes. The runtime openproxy sync command then applies them to db.json against the user's machine.

Schema

{
  "source": "9router" | "omniroute",
  "ref": "v0.4.55",                 // git ref or "HEAD"
  "generatedAt": "2026-05-19T...",
  "providerIdToAlias": {            // map provider id -> openproxy alias
    "openai": "openai",
    "github-models": "ghm"
  },
  "providers": [
    {
      "id": "openai",
      "alias": "openai",
      "format": "openai",           // present iff upstream exposes it
      "authType": "apikey",
      "baseUrl": "...",
      "models": [
        { "id": "gpt-5.5", "name": "GPT-5.5", "kind": "llm", "contextLength": 1050000 }
      ]
    }
  ]
}

The kind field follows openproxy's existing classification: llm | embedding | image | tts | stt | search | fetch | video.