Files
OmniRoute/@omniroute/opencode-plugin/README.md

252 lines
18 KiB
Markdown

# @omniroute/opencode-plugin
First-class OpenCode plugin for the [OmniRoute AI Gateway](https://github.com/diegosouzapw/OmniRoute). Pulls a live model catalog from `/v1/models` (including `-low`/`-medium`/`-high`/`-thinking` variants as first-class IDs), aggregates combos via `/api/combos` using a least-common-denominator capability/limit join, sanitizes Gemini tool schemas in flight, and supports multiple side-by-side OmniRoute instances out of the box.
## Install
Once published to npm:
```sh
npm install @omniroute/opencode-plugin
```
Until then (or for local development), reference the built artifact directly. Either extract the package into your OpenCode plugins dir and point at the extracted `dist/index.js`:
```sh
# from inside the OmniRoute repo
cd @omniroute/opencode-plugin && npm run build && npm pack
# then extract into ~/.config/opencode/plugins/omniroute-opencode-plugin/
```
Peer dep: `@opencode-ai/plugin` (managed by your OpenCode install).
## Quick start (single instance)
```jsonc
// opencode.json
{
"$schema": "https://opencode.ai/config.json",
"plugin": [
[
"@omniroute/opencode-plugin",
{
"providerId": "omniroute",
"baseURL": "https://or.example.com",
},
],
],
}
```
```sh
opencode auth login --provider omniroute
# prompts for the OmniRoute API key, writes to ~/.local/share/opencode/auth.json
```
> ⚠ Use the `--provider` flag explicitly. `opencode auth login omniroute` is parsed as a positional `url` argument by current OC releases (≤1.15.5) and fails with `fetch() URL is invalid`. Tracked upstream.
Restart OpenCode. `/models` lists the full live catalog. Variants (`-low`, `-medium`, `-high`, `-thinking`) and combos appear as first-class IDs — OmniRoute is the source of truth, no client-side synthesis.
## Multi-instance (prod + preprod side-by-side)
> ⚠ OC ≤1.15.5 dedupes plugin loads by absolute module path. Two `plugin:` entries pointing at the same `dist/index.js` collapse into one (last-listed options win). Workaround: install the plugin twice into separate directories so each entry resolves to a distinct module file. v0.2.x will introduce an `instances: [...]` shape that registers N providers from a single load.
### Dual-install workaround (works today on OC ≤1.15.5)
Pack the plugin once, extract it twice into named directories, then point each `plugin:` entry at its own copy:
```sh
# 1. Build + pack the plugin (run from the plugin worktree)
cd /path/to/OmniRoute/@omniroute/opencode-plugin
npm run build
npm pack
# produces omniroute-opencode-plugin-0.1.0.tgz
# 2. Extract one copy per OmniRoute endpoint
mkdir -p ~/.config/opencode/plugins/omniroute-opencode-plugin-prod
mkdir -p ~/.config/opencode/plugins/omniroute-opencode-plugin-preprod
tar -xzf omniroute-opencode-plugin-0.1.0.tgz -C ~/.config/opencode/plugins/omniroute-opencode-plugin-prod --strip-components=1
tar -xzf omniroute-opencode-plugin-0.1.0.tgz -C ~/.config/opencode/plugins/omniroute-opencode-plugin-preprod --strip-components=1
```
Then in `~/.config/opencode/opencode.json` reference each directory by absolute path:
```jsonc
{
"$schema": "https://opencode.ai/config.json",
"plugin": [
[
"./plugins/omniroute-opencode-plugin-prod/dist/index.js",
{
"providerId": "omniroute",
"displayName": "OmniRoute",
"baseURL": "https://or.example.com",
},
],
[
"./plugins/omniroute-opencode-plugin-preprod/dist/index.js",
{
"providerId": "omniroute-preprod",
"displayName": "OmniRoute Preprod",
"baseURL": "https://or-preprod.example.com",
},
],
],
}
```
Paths are relative to `~/.config/opencode/`. Each entry now resolves to a distinct module file, so OC loads them as two separate plugin instances. Authenticate each:
```sh
opencode auth login --provider omniroute
opencode auth login --provider omniroute-preprod
```
Each entry gets its own provider id, its own model picker entry, its own slot in `auth.json`, and its own TTL cache. Closures are isolated per plugin instance — no cross-talk.
### After publish (`@omniroute/opencode-plugin` npm)
Once the package is published, the dual-install becomes two `npm install --prefix` commands instead of `tar -xzf`:
```sh
mkdir -p ~/.config/opencode/plugins/omniroute-opencode-plugin-prod
mkdir -p ~/.config/opencode/plugins/omniroute-opencode-plugin-preprod
npm install --prefix ~/.config/opencode/plugins/omniroute-opencode-plugin-prod @omniroute/opencode-plugin
npm install --prefix ~/.config/opencode/plugins/omniroute-opencode-plugin-preprod @omniroute/opencode-plugin
```
`opencode.json` paths become `./plugins/omniroute-opencode-plugin-prod/node_modules/@omniroute/opencode-plugin/dist/index.js` (and the preprod equivalent).
## Features
| Feature | What it does | Hook |
| ------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------ | ---------------------------- |
| Dynamic `/v1/models` | Pulls live catalog (455+ entries on prod) on each refresh, TTL-cached | `provider.models` |
| Variants pass-through | `-low`/`-medium`/`-high`/`-thinking` ship as first-class IDs from OmniRoute (no client synthesis) | `provider.models` |
| Combo LCD aggregation | Combos appear with intersected capabilities + min context/output across members | `provider.models` + `config` |
| `combo/<slug>` namespace + `Combo: ` prefix | Combos surface under `combo/claude-primary` (not the upstream UUID) and the picker shows `Combo: claude-primary` so they stand apart from raw provider/model pairs | both hooks |
| Nice names + cost | `/api/pricing/models` display names AND `/api/pricing` per-million-token cost overlaid onto the live catalog | both hooks |
| Compression pipeline tags | Combo names get tagged with their compression pipeline (e.g. `Combo: claude-primary [rtk:standard → caveman:full]`) when `features.compressionMetadata: true` | both hooks |
| Usable-only filter | Filter to providers with at least one healthy connection in `/api/providers` (opt-in via `features.usableOnly`) | both hooks |
| Disk-cache fallback | Last-known-good catalog persisted to disk; hydrates on a cold start when `/v1/models` is unreachable (default-on, opt-out via `features.diskCache: false`) | `config` |
| Bearer injection + suffix-spoof guard | Adds `Authorization` on baseURL-matched requests only | `auth.loader.fetch` |
| Gemini schema sanitization | Strips `$schema`/`$ref`/`additionalProperties` for `gemini-*`/`google-vertex-gemini/*` | `auth.loader.fetch` wrap |
| Multi-instance | Each plugin entry binds to its own `providerId`; closures isolated | factory |
| Config-hook shim | OC ≤1.15.5 fallback: writes static catalog into `config.provider[id]` (config hook is the only one that fires in `serve` mode on these versions) | `config` |
## Plugin options
| Option | Type | Default | Description |
| --------------- | -------- | ------------------------------------------ | ---------------------------------------------------------- |
| `providerId` | `string` | `"omniroute"` | OpenCode provider id; must be unique across plugin entries |
| `displayName` | `string` | `"OmniRoute"` or `OmniRoute (<id>)` | Label in the OC UI |
| `modelCacheTtl` | `number` | `300000` (5 min) | `/v1/models` TTL in ms |
| `baseURL` | `string` | resolved from `auth.json` after `/connect` | Override OmniRoute base URL |
| `features` | `object` | see below | Feature toggles (all opt-in/out, defaults preserve v0.1.0) |
### `features` block
Every field is optional. Defaults mirror v0.1.0 behaviour so existing `opencode.json` files do not need to change.
| Feature | Type | Default | What it does |
| --------------------- | --------- | ------- | ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| `combos` | `boolean` | `true` | Discover `/api/combos` and surface them as pseudo-models with LCD capabilities. Combos are keyed under the `combo/<slug>` namespace and labelled `Combo: <name>` in the model picker so they're distinguishable from raw provider/model pairs. |
| `enrichment` | `boolean` | `true` | Pull display names from `/api/pricing/models` AND per-million-token pricing (`input`, `output`, `cached``cacheRead`, `cache_creation``cacheWrite`) from `/api/pricing`, then overlay both onto the live catalog (so the UI shows `Claude 4.7 Opus` with `cost.input: 5`, `cost.output: 25` instead of raw IDs and zeroed cost). |
| `compressionMetadata` | `boolean` | `false` | Pull `/api/context/combos` so combo names get tagged with their compression pipeline, e.g. `Combo: claude-primary [rtk:standard → caveman:full]` |
| `usableOnly` | `boolean` | `false` | Read `/api/providers` and filter the catalog to providers that have at least one connection with `isActive: true` AND `testStatus: 'active'`. Subtract-filter semantics: providers unknown to BOTH the pricing-models catalog AND the connection table pass through (so synthetic prefixes like `agentrouter/*` survive). On fetch failure the filter is disabled for the refresh — never hides the whole catalog. |
| `diskCache` | `boolean` | `true` | Persist the last successful `/v1/models` + `/api/combos` + enrichment + connections + compression snapshot to `${OPENCODE_DATA_DIR ?? ~/.local/share/opencode}/plugins/omniroute-<providerId>.json`. On a subsequent cold start where `/v1/models` throws (network down / IP whitelist drop / 5xx) the static block hydrates from the snapshot so OC's model picker survives offline. Soft-fail on read/write — never blocks publishing. |
| `geminiSanitization` | `boolean` | `true` | Strip `$schema`/`$ref`/`additionalProperties` from tool params when the model id matches `gemini` |
| `mcpAutoEmit` | `boolean` | `false` | Auto-write an `mcp.<providerId>` remote entry into the OC config pointing at `<baseURL>/api/mcp/stream` with the resolved Bearer token |
| `mcpToken` | `string` | _unset_ | Optional separate Bearer for the auto-emitted MCP entry. Falls back to the provider's `apiKey` (from `auth.json`) when unset |
| `fetchInterceptor` | `boolean` | `true` | Inject `Authorization: Bearer` + default `Content-Type` on every outbound request targeting `baseURL` (suffix-spoof guarded) |
#### Example — enrichment + compression tags + MCP auto-emit
```jsonc
{
"plugin": [
[
"@omniroute/opencode-plugin",
{
"providerId": "omniroute",
"baseURL": "https://or.example.com",
"features": {
"combos": true,
"enrichment": true,
"compressionMetadata": true,
"mcpAutoEmit": true,
},
},
],
],
}
```
With `mcpAutoEmit: true`, the plugin synthesises an `mcp.omniroute` entry equivalent to a manual:
```jsonc
"mcp": {
"omniroute": {
"type": "remote",
"url": "https://or.example.com/api/mcp/stream",
"enabled": true,
"headers": { "Authorization": "Bearer <apiKey-from-auth.json>" }
}
}
```
If you want a narrower-scoped Bearer for MCP (different from the chat/inference key), set `features.mcpToken`. Operator overrides win: if you already set `mcp.omniroute` in `opencode.json`, the plugin will not overwrite it.
#### Example — production-leaning defaults (clean picker, offline resilience)
```jsonc
{
"plugin": [
[
"@omniroute/opencode-plugin",
{
"providerId": "omniroute",
"baseURL": "https://or.example.com",
"features": {
"combos": true,
"enrichment": true,
"compressionMetadata": true,
"usableOnly": true,
"diskCache": true,
},
},
],
],
}
```
- `usableOnly: true` drops models whose canonical provider has no healthy connection in your OmniRoute instance — your `/models` picker stays focused on what you can actually call.
- `diskCache: true` (default) writes a snapshot to `${OPENCODE_DATA_DIR}/plugins/omniroute-<providerId>.json` on every healthy refresh. On a cold start where `/v1/models` is unreachable (laptop offline, IP whitelist drop), the snapshot hydrates the static block so OC still shows the catalog instead of a stub.
- `compressionMetadata: true` annotates combo display names with their pipeline (e.g. `Combo: claude-primary [rtk:standard → caveman:full]`) so the picker advertises which compression each combo applies.
## Comparison vs `@omniroute/opencode-provider`
[`@omniroute/opencode-provider`](https://github.com/diegosouzapw/OmniRoute/tree/main/%40omniroute/opencode-provider) is the existing config-generator package — it writes a frozen `provider.<id>` block into `opencode.json` at build time. This plugin is the runtime integration.
| | `@omniroute/opencode-plugin` (this) | `@omniroute/opencode-provider` |
| ----------------- | ----------------------------------- | --------------------------------- |
| Type | OC plugin | Config generator (CLI/build-time) |
| Models | Live from `/v1/models` | Frozen at scaffold |
| Combos | LCD-aggregated live | None |
| Gemini sanitize | Yes | N/A |
| OC UI integration | `/connect`, `/models` | None |
| Multi-instance | Native | Manual |
Both can coexist; pick the one that fits your environment.
## Requirements
- Node `>=22.22.3` (per `engines.node`); tested on Node 22 and 24.
- OpenCode: verified end-to-end against `opencode@1.15.5` with `@opencode-ai/plugin@1.15.6`.
- OC plugin peer (`@opencode-ai/plugin`) `>=1.14.49` for the full feature set (provider hook surfaces models in `/models`). On `<=1.14.48`, the plugin falls back to its `config` hook, writing a static catalog snapshot into `config.provider[id]` so models still appear.
- The plugin uses the OC v1 plugin shape (`default: { id, server }`) — older OC releases that only walk named exports will reject it. Stay on OC ≥1.15.
## License
MIT. See [LICENSE](./LICENSE).