feat: add vLLM model provider (#968)

Users can now use vLLM as a model provider. Tested with llama3.2 3b and with Rubra models.

docs/docs/05-configuration/02-model-providers.md

@@ -36,6 +36,7 @@ If this model provider hasn't been configured, then the API would return somethi
 ```
 
 To configure a model provider using the API, a `POST` request can be made to `/api/model-providers/azure-openai-model-provider/configure` with each required configuration parameter set in the body:
+
 ```json
 {
 	"OBOT_AZURE_OPENAI_MODEL_PROVIDER_ENDPOINT": "...",
@@ -107,3 +108,10 @@ Voyage is Anthropic's recommended text-embedding provider. The Voyage model prov
 The Ollama model provider requires the configuration parameter `OBOT_OLLAMA_MODEL_PROVIDER_HOST`. This host must point to a running instance of Ollama. For your reference, the default host and port for Ollama is `127.0.0.1:11434`. Obot doesn't set this by default.
 
 To set up and run an instance of Ollama, refer to the [Ollama GitHub readme](https://github.com/ollama/ollama/blob/main/README.md).
+
+## vLLM
+
+The vLLM model provider requires the following configuration parameters:
+
+- `OTTO8_VLLM_MODEL_PROVIDER_ENDPOINT`: The endpoint to use for your vLLM server (e.g., `your-vllm-server.com` or `https://your-vllm-server.com`)
+- `OTTO8_VLLM_MODEL_PROVIDER_API_KEY`: The API key for authentication with your vLLM server

ui/admin/app/components/model-providers/ModelProviderForm.tsx

@@ -61,6 +61,7 @@ const translateUserFriendlyLabel = (label: string) => {
         "OBOT_OLLAMA_MODEL_PROVIDER",
         "OBOT_VOYAGE_MODEL_PROVIDER",
         "OBOT_GROQ_MODEL_PROVIDER",
+        "OBOT_VLLM_MODEL_PROVIDER",
         "OBOT_ANTHROPIC_BEDROCK_MODEL_PROVIDER",
     ];
 

ui/admin/app/components/model-providers/constants.ts

@@ -1,6 +1,7 @@
 export const CommonModelProviderIds = {
     OLLAMA: "ollama-model-provider",
     GROQ: "groq-model-provider",
+    VLLM: "vllm-model-provider",
     VOYAGE: "voyage-model-provider",
     ANTHROPIC: "anthropic-model-provider",
     OPENAI: "openai-model-provider",
@@ -12,6 +13,7 @@ export const ModelProviderLinks = {
     [CommonModelProviderIds.VOYAGE]: "https://www.voyageai.com/",
     [CommonModelProviderIds.OLLAMA]: "https://ollama.com/",
     [CommonModelProviderIds.GROQ]: "https://groq.com/",
+    [CommonModelProviderIds.VLLM]: "https://docs.vllm.ai/",
     [CommonModelProviderIds.AZURE_OPENAI]:
         "https://azure.microsoft.com/en-us/explore/",
     [CommonModelProviderIds.ANTHROPIC]: "https://www.anthropic.com",
@@ -42,6 +44,11 @@ export const ModelProviderRequiredTooltips: {
         "Api Key":
             "Groq API Key. Can be created and fetched from https://console.groq.com/keys",
     },
+    [CommonModelProviderIds.VLLM]: {
+        Endpoint:
+            "Endpoint for the vLLM OpenAI service (eg. http://localhost:8000)",
+        "Api Key": "VLLM API Key set when starting the vLLM server",
+    },
     [CommonModelProviderIds.AZURE_OPENAI]: {
         Endpoint:
             "Endpoint for the Azure OpenAI service (e.g. https://<resource-name>.<region>.api.cognitive.microsoft.com/)",
@@ -89,6 +96,10 @@ export const ModelProviderSensitiveFields: Record<string, boolean | undefined> =
         // Groq
         OBOT_GROQ_MODEL_PROVIDER_API_KEY: true,
 
+        // VLLM
+        OBOT_VLLM_MODEL_PROVIDER_ENDPOINT: false,
+        OBOT_VLLM_MODEL_PROVIDER_API_KEY: true,
+
         // Anthropic Bedrock
         OBOT_ANTHROPIC_BEDROCK_MODEL_PROVIDER_ACCESS_KEY_ID: true,
         OBOT_ANTHROPIC_BEDROCK_MODEL_PROVIDER_SECRET_ACCESS_KEY: true,

ui/admin/app/routes/_auth.model-providers.tsx

@@ -33,6 +33,7 @@ const sortModelProviders = (modelProviders: ModelProvider[]) => {
             CommonModelProviderIds.OLLAMA,
             CommonModelProviderIds.VOYAGE,
             CommonModelProviderIds.GROQ,
+            CommonModelProviderIds.VLLM,
         ];
         const aIndex = preferredOrder.indexOf(a.id);
         const bIndex = preferredOrder.indexOf(b.id);