@elizaos/plugin-openrouter 2.0.0-beta.1 → 2.0.3-beta.2

package/README.md

@@ -1,6 +1,6 @@
 # @elizaos/plugin-openrouter
 
-This plugin provides integration with various models available through the OpenRouter API via the ElizaOS platform.
+elizaOS plugin that routes text generation, image description, image generation, and text embedding through the [OpenRouter](https://openrouter.ai) API, giving Eliza agents access to hundreds of hosted models via a single API key.
 
 ## Usage
 
@@ -25,7 +25,7 @@ The plugin requires the OpenRouter API key and can be configured via environment
   "OPENROUTER_IMAGE_MODEL": "x-ai/grok-2-vision-1212", // Optional: Overrides default image model
   "OPENROUTER_IMAGE_GENERATION_MODEL": "google/gemini-2.5-flash-image-preview", // Optional: Overrides default image generation model
   "OPENROUTER_EMBEDDING_MODEL": "openai/text-embedding-3-small", // Optional: Overrides default embedding model
-  "OPENROUTER_EMBEDDING_DIMENSIONS": "1536", // Optional: Sets embedding vector dimensions (256, 384, 512, 768, 1024, 1536, 2048, 3072)
+  "OPENROUTER_EMBEDDING_DIMENSIONS": "1536", // Optional: Sets embedding vector dimensions (384, 512, 768, 1024, 1536, 2048, 3072)
   "OPENROUTER_BROWSER_BASE_URL": "https://your-proxy.example.com/openrouter"
   // Fallbacks if specific OPENROUTER models are not set
   "SMALL_MODEL": "google/gemini-flash",
@@ -103,23 +103,31 @@ app.listen(3000);
 - `OPENROUTER_IMAGE_MODEL`: Specific model to use for `IMAGE_DESCRIPTION`. Overrides `IMAGE_MODEL` if set.
 - `OPENROUTER_IMAGE_GENERATION_MODEL`: Specific model to use for `IMAGE` generation. Overrides `IMAGE_GENERATION_MODEL` if set.
 - `OPENROUTER_EMBEDDING_MODEL`: Specific model to use for `TEXT_EMBEDDING`. Overrides `EMBEDDING_MODEL` if set.
-- `OPENROUTER_EMBEDDING_DIMENSIONS`: Number of dimensions for embedding vectors. Supported values: 256, 384, 512, 768, 1024, 1536, 2048, 3072. Defaults to 1536.
-- `OPENROUTER_AUTO_CLEANUP_IMAGES`: Whether to automatically delete generated images after 30 seconds (default: "false"). Set to "true" to enable auto-cleanup.
-- `SMALL_MODEL`: Fallback model for small tasks (default: "google/gemini-2.0-flash-001"). Used if `OPENROUTER_SMALL_MODEL` is not set.
-- `LARGE_MODEL`: Fallback model for large tasks (default: "openai/gpt-5.1-nano"). Used if `OPENROUTER_LARGE_MODEL` is not set.
+- `OPENROUTER_TRANSCRIPTION_MODEL`: Specific model to use for `TRANSCRIPTION` (default: `openai/whisper-large-v3`). Overrides `TRANSCRIPTION_MODEL` if set.
+- `OPENROUTER_EMBEDDING_DIMENSIONS`: Number of dimensions for embedding vectors. Supported values: 384, 512, 768, 1024, 1536, 2048, 3072. Defaults to 1536.
+- `OPENROUTER_AUTO_CLEANUP_IMAGES`: Boolean flag for auto-cleanup of generated images, read by `shouldAutoCleanupImages()` in `utils/config.ts` (default: "false").
+- `SMALL_MODEL`: Fallback model for small tasks (default: "google/gemini-2.5-flash-lite"). Used if `OPENROUTER_SMALL_MODEL` is not set.
+- `LARGE_MODEL`: Fallback model for large tasks (default: "google/gemini-2.5-flash"). Used if `OPENROUTER_LARGE_MODEL` is not set.
 - `IMAGE_MODEL`: Fallback model for image analysis (default: "x-ai/grok-2-vision-1212"). Used if `OPENROUTER_IMAGE_MODEL` is not set.
 - `IMAGE_GENERATION_MODEL`: Fallback model for image generation (default: "google/gemini-2.5-flash-image-preview"). Used if `OPENROUTER_IMAGE_GENERATION_MODEL` is not set.
 - `EMBEDDING_MODEL`: Fallback model for text embeddings (default: "openai/text-embedding-3-small"). Used if `OPENROUTER_EMBEDDING_MODEL` is not set.
+- `TRANSCRIPTION_MODEL`: Fallback model for audio transcription (default: "openai/whisper-large-v3"). Used if `OPENROUTER_TRANSCRIPTION_MODEL` is not set.
 - `EMBEDDING_DIMENSIONS`: Fallback dimension setting for embeddings (default: "1536"). Used if `OPENROUTER_EMBEDDING_DIMENSIONS` is not set.
 
 ## Provided Models
 
-The plugin currently provides these model types:
-
-- `TEXT_SMALL`: Optimized for fast, cost-effective text generation using the configured small model. Supports `tools`, `toolChoice`, and `responseSchema` for structured output via native tool calling.
-- `TEXT_LARGE`: For more complex text generation tasks requiring larger models, using the configured large model. Supports `tools`, `toolChoice`, and `responseSchema` for structured output via native tool calling.
-- `IMAGE_DESCRIPTION`: Analyzes images and provides descriptive text and titles, using the configured image model.
-- `IMAGE`: Generates images from text prompts using the configured image generation model (e.g., Gemini 2.5 Flash Image Preview).
-- `TEXT_EMBEDDING`: Generates vector embeddings for text input, supporting configurable dimensions from 256 to 3072, using the configured embedding model.
-
-_Note: Audio Transcription is not currently implemented in this specific OpenRouter plugin._
+The plugin registers these model types:
+
+- `TEXT_NANO`: Fastest/cheapest text generation; falls back to the small model when no nano override is set.
+- `TEXT_SMALL`: Fast, cost-effective text generation (default: `google/gemini-2.5-flash-lite`). Supports `tools`, `toolChoice`, and `responseSchema`.
+- `TEXT_MEDIUM`: Mid-tier text generation; falls back to the small model when no medium override is set.
+- `TEXT_LARGE`: Complex text generation tasks (default: `google/gemini-2.5-flash`). Supports `tools`, `toolChoice`, and `responseSchema`.
+- `TEXT_MEGA`: Largest text tasks; falls back to the large model when no mega override is set.
+- `RESPONSE_HANDLER`: Should-respond decisions; falls back to the nano model.
+- `ACTION_PLANNER`: Action planning; falls back to the medium model.
+- `IMAGE_DESCRIPTION`: Analyzes images and provides descriptive text (default: `x-ai/grok-2-vision-1212`).
+- `IMAGE`: Generates images from text prompts (default: `google/gemini-2.5-flash-image-preview`).
+- `TEXT_EMBEDDING`: Vector embeddings with configurable dimensions (default: `openai/text-embedding-3-small`, 1536 dims).
+- `TRANSCRIPTION`: Transcribes audio through OpenRouter's `/audio/transcriptions` endpoint (default: `openai/whisper-large-v3`).
+
+Transcription inputs may be URL strings, `Buffer`, `Blob` / `File`, core `{ audioUrl, prompt? }`, or local `{ audio, model?, language?, temperature?, format?, mimeType? }` objects. The handler sends base64 `input_audio` JSON to OpenRouter and returns the transcript text.

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@elizaos/plugin-openrouter",
-  "version": "2.0.0-beta.1",
+  "version": "2.0.3-beta.2",
   "type": "module",
   "main": "dist/node/index.node.js",
   "module": "dist/node/index.node.js",
@@ -28,6 +28,12 @@
         "default": "./dist/node/index.node.js"
       },
       "default": "./dist/node/index.node.js"
+    },
+    "./*.css": "./dist/*.css",
+    "./*": {
+      "types": "./dist/*.d.ts",
+      "import": "./dist/*.js",
+      "default": "./dist/*.js"
     }
   },
   "files": [
@@ -49,14 +55,12 @@
   },
   "sideEffects": false,
   "dependencies": {
-    "@ai-sdk/openai": "^3.0.9",
-    "@ai-sdk/ui-utils": "^1.2.8",
     "@openrouter/ai-sdk-provider": "^2.0.0",
     "ai": "^6.0.30"
   },
   "devDependencies": {
     "@biomejs/biome": "^2.4.14",
-    "@elizaos/core": "2.0.0-beta.1",
+    "@elizaos/core": "2.0.3-beta.2",
     "@types/bun": "^1.3.5",
     "@types/json-schema": "^7.0.15",
     "@types/node": "^25.0.3",
@@ -65,7 +69,7 @@
     "vitest": "^4.0.0"
   },
   "peerDependencies": {
-    "@elizaos/core": "2.0.0-beta.1"
+    "@elizaos/core": "2.0.3-beta.2"
   },
   "scripts": {
     "dev": "bun --hot build.ts",
@@ -74,8 +78,8 @@
     "format": "bunx @biomejs/biome format --write .",
     "format:check": "bunx @biomejs/biome format .",
     "typecheck": "tsc --noEmit --noCheck -p tsconfig.json",
-    "test": "vitest run --config ./vitest.config.ts --passWithNoTests",
-    "test:unit": "vitest run __tests__/ --config ./vitest.config.ts --passWithNoTests",
+    "test": "vitest run --config ./vitest.config.ts",
+    "test:unit": "vitest run __tests__/ --config ./vitest.config.ts",
     "test:watch": "vitest",
     "lint:check": "bunx @biomejs/biome check .",
     "build": "bun run build.ts",
@@ -116,7 +120,7 @@
         "type": "string",
         "description": "Overrides the default small language model used for text/object generation.",
         "required": false,
-        "default": "google/gemini-2.0-flash-001",
+        "default": "google/gemini-2.5-flash-lite",
         "sensitive": false
       },
       "OPENROUTER_LARGE_MODEL": {
@@ -130,7 +134,7 @@
         "type": "string",
         "description": "General fallback environment variable for the small model name when OPENROUTER_SMALL_MODEL is not set.",
         "required": false,
-        "default": "google/gemini-2.0-flash-001",
+        "default": "google/gemini-2.5-flash-lite",
         "sensitive": false
       },
       "LARGE_MODEL": {
@@ -177,7 +181,7 @@
       },
       "OPENROUTER_EMBEDDING_DIMENSIONS": {
         "type": "number",
-        "description": "Number of dimensions for embedding vectors. Supported values: 256, 384, 512, 768, 1024, 1536, 2048, 3072.",
+        "description": "Number of dimensions for embedding vectors. Supported values: 384, 512, 768, 1024, 1536, 2048, 3072.",
         "required": false,
         "default": 1536,
         "sensitive": false
@@ -215,5 +219,6 @@
       "browser": "Browser-compatible build available via exports.browser",
       "node": "Node.js build available via exports.node"
     }
-  }
+  },
+  "gitHead": "82fe0f44215954c2417328203f5bd6510985c1fc"
 }