> ## Documentation Index
> Fetch the complete documentation index at: https://docs.bor-os.io/llms.txt
> Use this file to discover all available pages before exploring further.

# Openai compatible

# OpenAI-compatible

The **OpenAI-compatible** provider (`openai-compatible`) connects BOR to any endpoint that speaks the OpenAI Chat Completions API: a local model server, a proxy, or another vendor's OpenAI-shaped API. Adapter: `server/providers/openai-compatible.js`.

## When to use it

* **Local models** — Ollama, LM Studio, vLLM, llama.cpp servers, etc.
* **Proxies / gateways** — LiteLLM, OpenRouter-style routers, your company's gateway.
* **Other vendors** that expose an OpenAI-compatible endpoint.

## Setup

This provider **requires a base URL** (`needsBaseUrl: true`). In [onboarding](../getting-started/onboarding.md):

1. Pick **OpenAI-compatible**.
2. Enter the **base URL** of your endpoint (e.g. `http://localhost:11434/v1` for Ollama, or your proxy URL).
3. Enter an **API key** if the endpoint needs one (local servers often accept any value).
4. Enter the **model name** exactly as your endpoint expects it.

## Notes

* The model name is whatever your endpoint serves — there's no fixed catalog, so type it exactly.
* Vision/image support depends on the underlying model; if it's text-only, screenshot-driven features won't work.
* Streaming and image blocks use the shared multimodal utilities, assuming the endpoint follows the OpenAI streaming format.

## Examples

| Backend         | Base URL                    | Notes                                              |
| --------------- | --------------------------- | -------------------------------------------------- |
| Ollama          | `http://localhost:11434/v1` | Run a local model; key can be anything.            |
| LM Studio       | `http://localhost:1234/v1`  | Local server with an OpenAI-compatible API.        |
| A proxy/gateway | your gateway URL            | Point at the gateway; it routes to the real model. |

## Troubleshooting

* **Connection refused** — the local server isn't running or the base URL/port is wrong.
* **404 / unknown model** — the model name doesn't match what the endpoint serves.
* **No streaming** — the endpoint must support OpenAI-style streamed responses.
* See [Troubleshooting](../troubleshooting.md).