unified-ai-router 3.0.1 → 3.1.0

package/bruno/openai/models.bru

@@ -0,0 +1,38 @@
+meta {
+  name: models
+  type: http
+  seq: 5
+}
+
+get {
+  url: {{address}}/models
+  body: json
+  auth: none
+}
+
+headers {
+  Content-Type: application/json
+  Authorization: {{token}}
+}
+
+body:json {
+  {
+    "messages": [
+      {
+        "role": "system",
+        "content": "You are a helpful assistant."
+      },
+      {
+        "role": "user",
+        "content": "Hello, say two words only."
+      }
+    ],
+    "model": "x-ai/grok-4-fast:free",
+    "temperature": 0,
+    "stream": true
+  }
+}
+
+settings {
+  encodeUrl: false
+}

package/main.js

@@ -134,6 +134,44 @@ class AIRouter
 		}
 		throw new Error( `All providers failed. Last error: ${lastError.message}` );
 	}
+
+	async getModels ()
+	{
+		const models = [];
+		for ( const provider of this.providers )
+		{
+			if ( !provider.apiKey )
+			{
+				logger.warn( `Skipping provider ${provider.name} due to missing API key` );
+				continue;
+			}
+			try
+			{
+				logger.info( `Fetching models for provider: ${provider.name}` );
+				const client = new OpenAI({
+					apiKey: provider.apiKey,
+					baseURL: provider.apiUrl,
+					timeout: 60000,
+				});
+				const listResponse = await client.models.list();
+				const modelList = listResponse.data && listResponse.data.length > 0 ? listResponse.data : listResponse.body || [];
+				const model = modelList.find( m => { return m.id === provider.model || m.id === `models/${provider.model}` });
+				if ( model )
+				{
+					models.push( model );
+				}
+				else
+				{
+					logger.warn( `Model ${provider.model} not found in provider ${provider.name}` );
+				}
+			}
+			catch ( error )
+			{
+				logger.error( `Failed to list models for provider ${provider.name}: ${error.message}` );
+			}
+		}
+		return models;
+	}
 }
 
 module.exports = AIRouter;

package/openai-compatible-server/index.js

@@ -1,26 +1,19 @@
 const express = require( "express" );
 const cors = require( "cors" );
-const AIRouter = require( "../main" ); // your existing class
+const AIRouter = require( "../main" );
 const pino = require( "pino" );
 const pretty = require( "pino-pretty" );
-const stream = pretty({ colorize: true, ignore: "pid,hostname" });
-const logger = pino({ base: false }, stream );
+const pinoStream = pretty({ colorize: true, ignore: "pid,hostname" });
+const logger = pino({ base: false }, pinoStream );
 require( "dotenv" ).config({ quiet: true });
 
 const app = express();
 app.use( cors() );
 app.use( express.json() );
 
-/**
- * Initialize router with providers (could load from env/config)
- */
 const providers = require( "../provider" )
-
 const aiRouter = new AIRouter( providers );
 
-/**
- * OpenAI-compatible endpoint: POST /v1/chat/completions
- */
 app.post( "/v1/chat/completions", async ( req, res ) =>
 {
 	const { messages, model, stream, ...rest } = req.body;
@@ -38,7 +31,7 @@ app.post( "/v1/chat/completions", async ( req, res ) =>
 
 		try
 		{
-			const result = await aiRouter.chatCompletionWithResponse( messages, { model, ...rest });
+			const result = await aiRouter.chatCompletionWithResponse( messages, { model, stream, ...rest });
 
 			for await ( const chunk of result.data )
 			{
@@ -60,7 +53,7 @@ app.post( "/v1/chat/completions", async ( req, res ) =>
 	{
 		try
 		{
-			const result = await aiRouter.chatCompletionWithResponse( messages, { model, ...rest });
+			const result = await aiRouter.chatCompletionWithResponse( messages, { model, stream, ...rest });
 
 			res.json( result.data );
 		}
@@ -72,7 +65,34 @@ app.post( "/v1/chat/completions", async ( req, res ) =>
 	}
 });
 
-// Health check
+app.get( "/v1/models", async ( req, res ) =>
+{
+	try
+	{
+		const models = await aiRouter.getModels();
+		res.json({ data: models });
+	}
+	catch ( error )
+	{
+		logger.error( `Error in /v1/models: ${error.message}` );
+		res.status( 500 ).json({ error: { message: error.message } });
+	}
+});
+
+app.get( "/models", async ( req, res ) =>
+{
+	try
+	{
+		const models = await aiRouter.getModels();
+		res.json({ data: models });
+	}
+	catch ( error )
+	{
+		logger.error( `Error in /models: ${error.message}` );
+		res.status( 500 ).json({ error: { message: error.message } });
+	}
+});
+
 app.get( "/health", ( req, res ) => { return res.json({ status: "ok" }) });
 
 // Start server

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "unified-ai-router",
-  "version": "3.0.1",
+  "version": "3.1.0",
   "description": "A unified interface for multiple LLM providers with automatic fallback. This project includes an OpenAI-compatible server and a deployable Telegram bot with a Mini App interface. It supports major providers like OpenAI, Google, Grok, and more, ensuring reliability and flexibility for your AI applications.",
   "license": "ISC",
   "author": "mlibre",

package/provider.js

@@ -1,5 +1,4 @@
 module.exports = [
-
 	{
 		name: "openrouter",
 		apiKey: process.env.OPENROUTER_API_KEY,

package/readme.md

@@ -84,7 +84,20 @@ console.log(response);
 
 ### 🔌 OpenAI-Compatible Server
 
-The OpenAI-compatible server provides a drop-in replacement for the OpenAI API. It routes requests through the unified router with fallback logic, ensuring high availability.
+The OpenAI-compatible server provides a drop-in replacement for the OpenAI API. It routes requests through the unified router with fallback logic, ensuring high availability.  
+The server uses the provider configurations defined in [provider.js](provider.js) file, and requires API keys set in a `.env` file.
+
+#### Setup
+
+1. Copy the example environment file:
+
+   ```bash
+   cp .env.example .env
+   ```
+
+2. Configure your providers in `provider.js`. Add new provider or modify existing ones with the appropriate `name`, `apiKey` (referencing the corresponding env variable), `model`, and `apiUrl` for the providers you want to use.
+
+3. Edit `.env` and add your API keys for the desired providers (see [🔑 API Keys](#-api-keys) for sources).
 
 To start the server locally, run:
 

package/legacy/main.js

@@ -1,63 +0,0 @@
-const { ChatOpenAI } = require( "@langchain/openai" );
-const pino = require( "pino" );
-const pretty = require( "pino-pretty" );
-const stream = pretty({ colorize: true, ignore: "pid,hostname" });
-const logger = pino({ base: false }, stream );
-
-class AIRouter
-{
-	constructor ( providers )
-	{
-		this.providers = providers;
-	}
-
-	async chatCompletion ( messages, options = {}, stream = false )
-	{
-		const { stream: streamOption, tools, model, ...restOptions } = options;
-		const isStreaming = stream || streamOption;
-
-		logger.info( `Starting chatCompletion with ${this.providers.length} providers (streaming: ${isStreaming})` );
-		let lastError;
-
-		for ( const provider of this.providers )
-		{
-			try
-			{
-				logger.info( `Attempting with provider: ${provider.name}` );
-				let llm = new ChatOpenAI({
-					apiKey: provider.apiKey,
-					model: provider.model,
-					configuration: {
-						baseURL: provider.apiUrl,
-					},
-					...restOptions,
-				});
-
-				if ( tools && tools.length > 0 )
-				{
-					llm = llm.bindTools( tools );
-				}
-
-				if ( isStreaming )
-				{
-					const stream = await llm.stream( messages );
-					return stream;
-				}
-				else
-				{
-					const response = await llm.invoke( messages, { timeout: 60000 });
-					return response;
-				}
-			}
-			catch ( error )
-			{
-				lastError = error;
-				logger.error( `Failed with ${provider.name}:${error.message}` );
-				// Continue to next provider
-			}
-		}
-		throw new Error( `All providers failed. Last error: ${lastError.message}` );
-	}
-}
-
-module.exports = AIRouter;

package/legacy/openai-compatible-server.js

@@ -1,192 +0,0 @@
-const express = require( "express" );
-const cors = require( "cors" );
-const AIRouter = require( "../main" ); // your existing class
-const pino = require( "pino" );
-const pretty = require( "pino-pretty" );
-const stream = pretty({ colorize: true, ignore: "pid,hostname" });
-const logger = pino({ base: false }, stream );
-require( "dotenv" ).config({ quiet: true });
-
-const app = express();
-app.use( cors() );
-app.use( express.json() );
-
-/**
- * Initialize router with providers (could load from env/config)
- */
-const providers = require( "../provider" )
-
-const aiRouter = new AIRouter( providers );
-
-/**
- * OpenAI-compatible endpoint: POST /v1/chat/completions
- */
-app.post( "/v1/chat/completions", async ( req, res ) =>
-{
-	try
-	{
-		const { messages, model, stream, ...rest } = req.body;
-
-		if ( !messages || !Array.isArray( messages ) )
-		{
-			return res.status( 400 ).json({ error: { message: "messages must be an array" } });
-		}
-
-		if ( stream )
-		{
-			// Streaming mode → use Server-Sent Events (SSE)
-			res.setHeader( "Content-Type", "text/event-stream" );
-			res.setHeader( "Cache-Control", "no-cache" );
-			res.setHeader( "Connection", "keep-alive" );
-
-			try
-			{
-				const response = await aiRouter.chatCompletion( messages, { model, ...rest }, true );
-				const id = `chatcmpl-${Date.now()}`;
-				const created = Math.floor( Date.now() / 1000 );
-				let fullResponse = null;
-				for await ( const chunk of response )
-				{
-					const modelName = chunk?.response_metadata?.model_name || model || "unknown";
-					const systemFingerprint = chunk?.response_metadata?.system_fingerprint || null;
-					let delta = { ... chunk.delta || { content: chunk.content || "" } };
-					if ( !delta.role ) delta.role = "assistant";
-					delta.reasoning = delta.reasoning || null;
-					delta.reasoning_details = delta.reasoning_details || [];
-					let toolCallsDelta = null;
-					if ( chunk.tool_calls && chunk.tool_calls.length > 0 )
-					{
-						toolCallsDelta = chunk.tool_calls.map( ( tc, index ) =>
-						{
-							return {
-								id: tc.id || `call_${Date.now()}_${Math.random().toString( 36 ).substr( 2, 9 )}`,
-								type: "function",
-								index,
-								function: {
-									name: tc.name,
-									arguments: JSON.stringify( tc.args || {})
-								}
-							};
-						});
-						delta.tool_calls = toolCallsDelta;
-						delta.content = "";
-					}
-					const chunkFinishReason = delta.finish_reason || chunk?.response_metadata?.finish_reason || null;
-					const chunkNativeFinishReason = delta.native_finish_reason || chunk?.response_metadata?.native_finish_reason || chunkFinishReason || null;
-					if ( chunk.content && !fullResponse ) fullResponse = chunk; // Capture full for reasoning if available
-					const payload = {
-						id,
-						provider: "OpenAI",
-						object: "chat.completion.chunk",
-						created,
-						model: modelName,
-						system_fingerprint: systemFingerprint,
-						choices: [
-							{
-								logprobs: null,
-								delta,
-								index: 0,
-								finish_reason: chunkFinishReason,
-								native_finish_reason: chunkNativeFinishReason,
-							},
-						],
-					};
-					const usage = chunk?.response_metadata?.usage;
-					if ( usage && typeof usage === "object" && !Array.isArray( usage ) && Object.keys( usage ).length > 0 )
-					{
-						payload.usage = chunk?.response_metadata?.usage || null;
-					}
-					res.write( `data: ${JSON.stringify( payload )}\n\n` );
-				}
-				// Send done signal
-				res.write( "data: [DONE]\n\n" );
-				res.end();
-			}
-			catch ( err )
-			{
-				logger.error( err );
-				res.write( `data: ${JSON.stringify({ error: err.message })}\n\n` );
-				res.write( "data: [DONE]\n\n" );
-				res.end();
-			}
-		}
-		else
-		{
-			// Non-streaming → return one-shot completion
-			const response = await aiRouter.chatCompletion( messages, { model, ...rest }, false );
-			let reasoning = null;
-			let refusal = null;
-			let toolCalls = null;
-			if ( response.contentBlocks )
-			{
-				const reasoningBlocks = response.contentBlocks.filter( b => { return b.type === "reasoning" || b.type === "thinking" });
-				reasoning = reasoningBlocks.length > 0 ? reasoningBlocks.map( b => { return b.text }).join( "\n" ) : null;
-				const refusalBlocks = response.contentBlocks.filter( b => { return b.type === "refusal" });
-				refusal = refusalBlocks.length > 0 ? refusalBlocks.map( b => { return b.text }).join( "\n" ) : null;
-			}
-			if ( response.tool_calls && response.tool_calls.length > 0 )
-			{
-				toolCalls = response.tool_calls.map( ( tc, index ) =>
-				{
-				  return {
-					 id: tc.id || `call_${Date.now()}_${Math.random().toString( 36 ).substr( 2, 9 )}`,
-					 type: "function",
-					 index,
-					 function: {
-							name: tc.name,
-							arguments: JSON.stringify( tc.args || {})
-					 }
-				  };
-				});
-			}
-			const systemFingerprint = response.response_metadata?.system_fingerprint || null;
-			const finishReason = response.response_metadata?.finish_reason || null;
-			const nativeFinishReason = response.response_metadata?.native_finish_reason || finishReason || null;
-			const messageContent = toolCalls && toolCalls.length > 0 ? "" : response.content;
-			let finalResult = {
-				id: `chatcmpl_${Date.now()}`,
-				provider: "OpenAI",
-				object: "chat.completion",
-				created: Math.floor( Date.now() / 1000 ),
-				model: response.response_metadata?.model_name || model || "unknown",
-				system_fingerprint: systemFingerprint,
-				choices: [
-					{
-						logprobs: null,
-						finish_reason: finishReason,
-						native_finish_reason: nativeFinishReason,
-						index: 0,
-						message: {
-							role: "assistant",
-							content: messageContent,
-							refusal,
-							reasoning,
-							tool_calls: toolCalls
-						},
-					},
-				],
-			}
-			const usage = response?.response_metadata?.usage;
-			if ( usage && typeof usage === "object" && !Array.isArray( usage ) && Object.keys( usage ).length > 0 )
-			{
-				finalResult.usage = response?.response_metadata?.usage || null;
-			}
-			res.json( finalResult );
-		}
-	}
-	catch ( err )
-	{
-		logger.error( err );
-		res.status( 500 ).json({ error: { message: err.message } });
-	}
-});
-
-// Health check
-app.get( "/health", ( req, res ) => { return res.json({ status: "ok" }) });
-
-// Start server
-const PORT = process.env.PORT || 3000;
-app.listen( PORT, () =>
-{
-	logger.info( `🚀 OpenAI-compatible API listening at http://localhost:${PORT}/v1/chat/completions` );
-});