feat: Add Tetrate Agent Router Service as a provider (#7136)

feat: Add TARS as a provider

Update gui with TARS provider

Change default model to gpt-5 mini

Author: johnlanda

core/llm/llms/TARS.ts

@@ -0,0 +1,18 @@
+import { LLMOptions } from "../../index.js";
+import { osModelsEditPrompt } from "../templates/edit.js";
+
+import OpenAI from "./OpenAI.js";
+
+class TARS extends OpenAI {
+  static providerName = "tars";
+  static defaultOptions: Partial<LLMOptions> = {
+    apiBase: "https://api.router.tetrate.ai/v1",
+    model: "gpt-5-mini",
+    promptTemplates: {
+      edit: osModelsEditPrompt,
+    },
+    useLegacyCompletionsEndpoint: false,
+  };
+}
+
+export default TARS;

core/llm/llms/index.ts

@@ -53,6 +53,7 @@ import SambaNova from "./SambaNova";
 import Scaleway from "./Scaleway";
 import SiliconFlow from "./SiliconFlow";
 import ContinueProxy from "./stubs/ContinueProxy";
+import TARS from "./TARS";
 import TestLLM from "./Test";
 import TextGenWebUI from "./TextGenWebUI";
 import Together from "./Together";
@@ -116,6 +117,7 @@ export const LLMClasses = [
   Inception,
   Voyage,
   LlamaStack,
+  TARS,
 ];
 
 export async function llmFromDescription(

docs/customize/model-providers/more/tetrate_agent_router_service.mdx

@@ -0,0 +1,44 @@
+---
+title: "Tetrate Agent Router Service"
+---
+
+The Tetrate Agent Router Service (TARS) provides a unified API gateway for accessing various AI models with fast inference capabilities. TARS acts as an intelligent router that can distribute requests across multiple model providers, offering enterprise-grade reliability and performance optimization.
+
+## Getting Started
+
+1. Obtain an API key from the [Tetrate Agent Router Service portal](https://api.router.tetrate.ai/)
+2. Configure Continue to use TARS as your model provider
+
+## Configuration
+
+Update your Continue configuration file (`~/.continue/config.json` or `~/.continue/config.yaml`) with your TARS API key:
+
+<Tabs>
+  <Tab title="YAML">
+  ```yaml title="config.yaml"
+  models:
+    - name: TARS GPT-5 Mini
+      provider: tars
+      model: gpt-5-mini
+      apiKey: <YOUR_TARS_API_KEY>
+  ```
+  </Tab>
+  <Tab title="JSON">
+  ```json title="config.json"
+  {
+    "models": [
+      {
+        "title": "TARS GPT-5 Mini",
+        "provider": "tars",
+        "model": "gpt-5-mini",
+        "apiKey": "<YOUR_TARS_API_KEY>"
+      }
+    ]
+  }
+  ```
+  </Tab>
+</Tabs>
+
+## Available Models
+
+TARS supports routing to various models. Go to the [Tetrate Agent Router Service model catalog](https://router.tetrate.ai/models) for a full list of supported models. 

extensions/vscode/config_schema.json

@@ -233,7 +233,8 @@
             "morph",
             "ovhcloud",
             "venice",
-            "inception"
+            "inception",
+            "tars"
           ],
           "markdownEnumDescriptions": [
             "### OpenAI\nUse gpt-4, gpt-3.5-turbo, or any other OpenAI model. See [here](https://openai.com/product#made-for-developers) to obtain an API key.\n\n> [Reference](https://docs.continue.dev/reference/Model%20Providers/openai)",
@@ -283,7 +284,8 @@
             "### Morph\nMorph provides a fast apply model. To get started, obtain an API key from [here](https://morphllm.com/dashboard).",
             "### OVHcloud AI Endpoints is a serverless inference API that provides access to a curated selection of models (e.g., Llama, Mistral, Qwen, Deepseek). It is designed with security and data privacy in mind and is compliant with GDPR. To get started, create an API key on the OVHcloud [AI Endpoints website](https://endpoints.ai.cloud.ovh.net/). For more information, including pricing, visit the OVHcloud [AI Endpoints product page](https://www.ovhcloud.com/en/public-cloud/ai-endpoints/).",
             "### Venice\n Venice.AI is a privacy-focused generative AI platform, allowing users to interact with open-source LLMs without storing any private user data.\nHosted models support the OpenAI API standard, providing seamless integration for users seeking privacy and flexibility.\nTo get started with the Venice API, either purchase a pro account, stake $VVV for daily inference allotments, or fund your account with USD.\nVisit the [API settings page](https://venice.ai/settings/api) or learn more at the [Venice API documentation](https://venice.ai/api).",
-            "### Inception\n Inception Labs offer a new generation of diffusion-based LLMs.\nVisit the [API settings page](https://platform.inceptionlabs.ai/) or learn more at the [Inception docs](https://platform.inceptionlabs.ai/docs)."
+            "### Inception\n Inception Labs offer a new generation of diffusion-based LLMs.\nVisit the [API settings page](https://platform.inceptionlabs.ai/) or learn more at the [Inception docs](https://platform.inceptionlabs.ai/docs).",
+            "### TARS\nTARS is an OpenAI-compatible proxy router. To get started, obtain an API key and configure the provider in your config.json."
           ],
           "type": "string"
         },

gui/public/logos/TetrateAgentRouterService.png

@@ -998,4 +998,30 @@ To get started, [register](https://dataplatform.cloud.ibm.com/registration/stepo
     packages: [{ ...models.AUTODETECT }],
     apiKeyUrl: "https://venice.ai/chat",
   },
+  tars: {
+    title: "Tetrate Agent Router Service",
+    provider: "tars",
+    icon: "TetrateAgentRouterService.png",
+    description: "TARS API for fast inference with various models",
+    tags: [ModelProviderTags.RequiresApiKey],
+    collectInputFor: [
+      {
+        inputType: "text",
+        key: "apiKey",
+        label: "API Key",
+        placeholder: "Enter your Tetrate Agent Router Service API key",
+        required: true,
+      },
+    ],
+    packages: [
+      {
+        ...models.AUTODETECT,
+        params: {
+          ...models.AUTODETECT.params,
+          title: "TARS",
+        },
+      },
+    ],
+    apiKeyUrl: "https://api.router.tetrate.ai/",
+  },
 };