opencode-provider-litellm 0.5.1 → 0.5.2

package/README.md

@@ -21,34 +21,39 @@ All models and MCP tools from your LiteLLM proxy appear in OpenCode automaticall
 
 ### Environment variables
 
-| Variable | Description |
-|----------|-------------|
-| `LITELLM_URL` | Your LiteLLM proxy base URL |
-| `LITELLM_KEY` | API key for the proxy |
-| `LITELLM_PROVIDER_ID` | Provider ID in OpenCode (defaults to `LiteLLM`) |
-| `LITELLM_GCLOUD_TOKEN_AUTH` | Set to `1` to use Google ADC for auth (makes `LITELLM_KEY` optional) |
-| `GOOGLE_APPLICATION_CREDENTIALS` | Path to a Google ADC JSON file (used when `LITELLM_GCLOUD_TOKEN_AUTH=1`) |
+Environment variables take precedence over all other configuration. Use them to keep secrets out of checked-in files.
 
-### Inline config
+| Variable | Required | Description |
+|----------|----------|-------------|
+| `LITELLM_URL` | Yes | Your LiteLLM proxy base URL |
+| `LITELLM_KEY` | Yes* | API key for the proxy |
+| `LITELLM_PROVIDER_ID` | No | Provider name in OpenCode (default: `LiteLLM`) |
+| `LITELLM_GCLOUD_TOKEN_AUTH` | No | Set to `1` to use Google ADC for auth — makes `LITELLM_KEY` optional |
+| `GOOGLE_APPLICATION_CREDENTIALS` | No | Path to a Google ADC JSON file (used when `LITELLM_GCLOUD_TOKEN_AUTH=1`) |
 
-Alternatively, provide `url` and `apiKey` directly in your `opencode.json`:
+*`LITELLM_KEY` is optional when `LITELLM_GCLOUD_TOKEN_AUTH=1`.
+
+### Plugin options
+
+All env vars have an equivalent in the plugin options block in `opencode.json`. **Env vars take precedence** — use options for defaults that can be overridden per environment.
 
 ```jsonc
 {
   "plugin": [
     ["opencode-provider-litellm", {
-      "url": "https://your-litellm-proxy.example.com",
-      "apiKey": "sk-..."
+      "url": "https://your-litellm-proxy.example.com",  // LITELLM_URL
+      "apiKey": "sk-...",                                // LITELLM_KEY
+      "providerName": "MyLiteLLM",                       // LITELLM_PROVIDER_ID
+      "gcloudTokenAuth": true                            // LITELLM_GCLOUD_TOKEN_AUTH=1
+      // apiKey can be omitted when gcloudTokenAuth is true
     }]
   ]
 }
 ```
 
-> **Tip:** Environment variables take precedence over inline config. Use env vars to keep secrets out of checked-in files.
-
 ### Google Vertex AI (gcloud token auth)
 
-When your LiteLLM proxy is backed by Google Vertex AI, you can skip `LITELLM_KEY` and let the plugin automatically fetch a gcloud OAuth token:
+When your LiteLLM proxy is backed by Google Vertex AI, you can skip `LITELLM_KEY` and let the plugin automatically fetch a Google OAuth token:
 
 ```bash
 # 1. Authenticate with gcloud (creates an ADC JSON file)
@@ -62,9 +67,12 @@ export LITELLM_GCLOUD_TOKEN_AUTH=1
 opencode plugin opencode-provider-litellm
 ```
 
-The plugin reads your [Application Default Credentials](https://cloud.google.com/docs/authentication/application-default-credentials) JSON file and exchanges the refresh token for an access token before every LLM request. Tokens are cached for 50 minutes.
+The plugin reads your [Application Default Credentials](https://cloud.google.com/docs/authentication/application-default-credentials) JSON file and exchanges the refresh token for an access token. Tokens are cached for 50 minutes and used for both model discovery at startup and every LLM request.
 
-To use a custom credentials file, set `GOOGLE_APPLICATION_CREDENTIALS` to its path.
+**ADC file locations searched (in order):**
+1. `GOOGLE_APPLICATION_CREDENTIALS` env var (all platforms)
+2. `~/.config/gcloud/application_default_credentials.json` (Linux/macOS)
+3. `%APPDATA%/gcloud/application_default_credentials.json` (Windows)
 
 > **Note:** Only `authorized_user` credentials (from `gcloud auth application-default login`) are supported. Service account keys are not yet supported.
 
@@ -109,22 +117,23 @@ Skills appear in OpenCode's `/skills` menu and are loaded natively by the agent.
 
 ## How it works
 
-The plugin uses three OpenCode hooks:
+The plugin uses these OpenCode hooks:
 
 | Hook | Purpose |
 |------|---------|
 | `config` | Discovers models from LiteLLM and injects them into OpenCode |
 | `auth` | Provides a `/connect` entry point for pasting an API key |
 | `tool` | Exposes discovered MCP tools as native OpenCode tools |
-| `chat.headers` | Injects `Authorization: Bearer <token>` when `LITELLM_GCLOUD_TOKEN_AUTH=1` |
+| `chat.headers` | Injects `Authorization: Bearer <token>` when gcloud token auth is enabled |
 
 ## Troubleshooting
 
 | Problem | Solution |
 |---------|----------|
-| "Plugin config error" | Set `LITELLM_URL` and `LITELLM_KEY`, or add `url`/`apiKey` to plugin options |
+| "Plugin config error" | Set `LITELLM_URL` and `LITELLM_KEY` (or `LITELLM_GCLOUD_TOKEN_AUTH=1`) |
 | "Access denied" (403) | Verify the API key has access to the LiteLLM proxy |
-| "No models discovered" | Check that the proxy is reachable and the `/health` endpoint responds |
+| "No models discovered" | Check the proxy is reachable and `/health` responds |
+| No models with gcloud auth | Verify `gcloud auth application-default login` has been run and the ADC file exists |
 | Skills not showing | Verify the proxy-sidecar is running and the skills URL is in `opencode.json` |
 
 ## Development

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "opencode-provider-litellm",
-  "version": "0.5.1",
+  "version": "0.5.2",
   "description": "OpenCode plugin for any LiteLLM proxy — auto-discovers models, auth, and capabilities",
   "type": "module",
   "exports": {

package/src/discovery.test.ts

@@ -83,7 +83,7 @@ describe('discoverModels', () => {
         tool_call: true,
         reasoning: false,
         limit: { context: 8192, output: 8192 },
-        cost: { input: 0.0001, output: 0.0003 },
+        cost: { input: 100, output: 300 },
         modalities: { input: ['text'], output: ['text'] },
       },
       'qwen3-32b': {
@@ -91,7 +91,7 @@ describe('discoverModels', () => {
         tool_call: true,
         reasoning: true,
         limit: { context: 32768, output: 32768 },
-        cost: { input: 0.00005, output: 0.00015 },
+        cost: { input: 50, output: 150 },
         modalities: { input: ['text'], output: ['text'] },
       },
     })
@@ -107,6 +107,51 @@ describe('discoverModels', () => {
     expect(getToken).toHaveBeenCalled()
   })
 
+  it('converts per-token cost to per-1M tokens with cache costs', async () => {
+    const mockFetch = vi.fn()
+      .mockResolvedValueOnce({
+        ok: true,
+        status: 200,
+        json: async () => ({
+          healthy_endpoints: [
+            { model: 'anthropic/claude-sonnet', model_id: 'uuid-1' },
+          ],
+        }),
+      })
+      .mockResolvedValueOnce({
+        ok: true,
+        status: 200,
+        json: async () => ({
+          data: [{
+            model_name: 'anthropic/claude-sonnet',
+            model_info: {
+              max_input_tokens: 1_000_000,
+              max_output_tokens: 64_000,
+              supports_function_calling: true,
+              supports_reasoning: true,
+              supports_vision: true,
+              supports_pdf_input: true,
+              // Per-token costs (LiteLLM format)
+              input_cost_per_token: 0.000005,
+              output_cost_per_token: 0.000025,
+              cache_read_input_token_cost: 0.0000005,
+              cache_creation_input_token_cost: 0.00000375,
+            },
+          }],
+        }),
+      })
+    vi.stubGlobal('fetch', mockFetch)
+
+    const result = await discoverModels(config, getToken)
+
+    expect(result['anthropic/claude-sonnet']?.cost).toEqual({
+      input: 5,         // 0.000005 * 1M
+      output: 25,       // 0.000025 * 1M
+      cache_read: 0.5,  // 0.0000005 * 1M
+      cache_write: 3.75, // 0.00000375 * 1M
+    })
+  })
+
   it('returns empty object on timeout', async () => {
     vi.useFakeTimers()
 

package/src/discovery.ts

@@ -117,16 +117,17 @@ export async function discoverModels(
       }
 
       // Add cost info if available
+      // LiteLLM returns cost per single token; opencode expects cost per 1M tokens
       if (info.input_cost_per_token != null && info.output_cost_per_token != null) {
         modelConfig.cost = {
-          input: info.input_cost_per_token,
-          output: info.output_cost_per_token,
+          input: info.input_cost_per_token * 1_000_000,
+          output: info.output_cost_per_token * 1_000_000,
         }
         if (info.cache_read_input_token_cost != null) {
-          modelConfig.cost.cache_read = info.cache_read_input_token_cost
+          modelConfig.cost.cache_read = info.cache_read_input_token_cost * 1_000_000
         }
         if (info.cache_creation_input_token_cost != null) {
-          modelConfig.cost.cache_write = info.cache_creation_input_token_cost
+          modelConfig.cost.cache_write = info.cache_creation_input_token_cost * 1_000_000
         }
       }