OpenAI Compatibility API Reference

LibreChatTmuxBridge provides REST endpoints compatible with the OpenAI API specification. These endpoints allow LibreChat to connect to host terminal sessions without custom client plugins.

Endpoint Overview

Method	Path	Description
GET	/v1/models	Returns available host tmux sessions as language models
POST	/v1/chat/completions	Injects terminal input and streams terminal responses

GET /v1/models

Queries active tmux sessions and returns them in the OpenAI model list format.

Request

GET /v1/models HTTP/1.1
Host: 10.0.0.10:8035
Authorization: Bearer sk-tmux

Response Example

{
  "object": "list",
  "data": [
    {
      "id": "tmux:new",
      "object": "model",
      "created": 1700000000,
      "owned_by": "tmux-bridge"
    },
    {
      "id": "tmux:agy-work",
      "object": "model",
      "created": 1700000120,
      "owned_by": "tmux-bridge"
    },
    {
      "id": "tmux:infra",
      "object": "model",
      "created": 1700000240,
      "owned_by": "tmux-bridge"
    }
  ]
}

The model tmux:new is always included to allow creating new sessions from the client interface.

POST /v1/chat/completions

Transmits user messages to the specified tmux session and captures resulting terminal changes.

Request Body Parameters

Field	Type	Required	Description
model	string	Yes	Target session identifier (for example, tmux:infra or tmux:agy-work)
messages	array	Yes	Array of chat messages. The final user message contains the command or keystroke
stream	boolean	No	Enables Server-Sent Events streaming. Default is false

Request Example

{
  "model": "tmux:infra",
  "messages": [
    {
      "role": "user",
      "content": "docker ps --format 'table {{.Names}}\t{{.Status}}'"
    }
  ],
  "stream": true
}

Streaming Response Format

When stream is true, the server sets the content type to text/event-stream and transmits incremental chunks:

data: {"id":"chatcmpl-a1b2c3d4","object":"chat.completion.chunk","created":1700000300,"model":"tmux:infra","choices":[{"index":0,"delta":{"content":"NAMES             STATUS\n"},"finish_reason":null}]}

data: {"id":"chatcmpl-a1b2c3d4","object":"chat.completion.chunk","created":1700000300,"model":"tmux:infra","choices":[{"index":0,"delta":{"content":"caddy             Up 12 hours\n"},"finish_reason":null}]}

data: {"id":"chatcmpl-a1b2c3d4","object":"chat.completion.chunk","created":1700000300,"model":"tmux:infra","choices":[{"index":0,"delta":{},"finish_reason":"stop"}]}

data: [DONE]

Non-Streaming Response Format

When stream is false, the server gathers settled output and returns a single JSON object:

{
  "id": "chatcmpl-a1b2c3d4",
  "object": "chat.completion",
  "created": 1700000300,
  "model": "tmux:infra",
  "choices": [
    {
      "index": 0,
      "message": {
        "role": "assistant",
        "content": "NAMES             STATUS\ncaddy             Up 12 hours\n"
      },
      "finish_reason": "stop"
    }
  ],
  "usage": {
    "prompt_tokens": 0,
    "completion_tokens": 0,
    "total_tokens": 0
  }
}