docs: add llama.cpp custom endpoint example, closes #783

content/docs/configuration/librechat_yaml/ai_endpoints/llama-cpp.mdx

@@ -0,0 +1,47 @@
+---
+title: llama.cpp
+icon: BookOpen
+description: Configure llama.cpp as a custom endpoint in LibreChat.
+---
+
+[llama.cpp](https://github.com/ggml-org/llama.cpp) includes `llama-server`, which provides an OpenAI-compatible API for locally hosted models.
+
+## Configuration
+
+Set the same API key for LibreChat and `llama-server`. For example, add this to the `.env` file used by LibreChat:
+
+```bash filename=".env"
+LLAMA_CPP_API_KEY=replace-with-the-same-key
+```
+
+Start `llama-server` with a model alias and that key:
+
+```bash
+llama-server \
+  --model /path/to/model.gguf \
+  --alias local-model \
+  --port 8080 \
+  --api-key "replace-with-the-same-key"
+```
+
+Add this endpoint under `endpoints.custom` in your `librechat.yaml`:
+
+```yaml filename="librechat.yaml"
+    - name: "llama.cpp"
+      apiKey: "${LLAMA_CPP_API_KEY}"
+      baseURL: "http://localhost:8080/v1"
+      models:
+        default: ["local-model"]
+        fetch: true
+      titleConvo: true
+      titleModel: "current_model"
+      summarize: false
+      summaryModel: "current_model"
+      modelDisplayLabel: "llama.cpp"
+```
+
+## Notes
+
+- `--alias local-model` sets the model ID returned by the server, which must match the entry in `models.default`.
+- `fetch: true` lets LibreChat discover models from the server's `/v1/models` endpoint.
+- The server listens on `127.0.0.1:8080` by default. If LibreChat runs in Docker, use an address reachable from the API container instead of `localhost`. If you change the server's bind address, keep it on a trusted network and retain API-key protection.

content/docs/configuration/librechat_yaml/ai_endpoints/meta.json

@@ -14,6 +14,7 @@
     "helicone",
     "huggingface",
     "lemonade",
+    "llama-cpp",
     "litellm",
     "mistral",
     "mlx",