Skip to main content

How It Works

The /chat endpoint enables your agents to interact with Telex AI models by sending messages and receiving AI-generated responses.

Agents can:

  • Ask one-time questions (stateless)
  • Hold multi-turn conversations (contextual)
  • Shape model behavior via instructions
  • (Soon) Receive responses as a live stream

The flexibility of this endpoint makes it ideal for building advanced agentic systems integrated with Telex.

🔗 Endpoint​

POST /chat

Required Headers​

  • X-AGENT-API-KEY: Your API key

How to Specify a Model​

Models can be selected by specifying the model ID using one of the ways listed below:

  • Using the request headers
  • Using the query params
  • Using the request body

Using the request Header​

Example​

X-AGENT-API-KEY: your-key
X-Model: deepseek/deepseek-r1-0528:free

Using the query parameter​

Example​

`POST` /chat?model=deepseek/deepseek-r1-0528:free

Using the request body​

Example​

{
  "model": "deepseek/deepseek-r1-0528:free",
  "messages": [
    { "role": "user", "content": "Hello!" }
  ],
  "stream": false
}

Agents should dynamically select models using the Model Discovery endpoint.

Ways to utilize This Endpoint​

There are multiple ways to interact with /chat, depending on your use case:

  1. Stateless Chat

    • Single message / one-turn conversation

  2. Contextual Chat

    • Multi-turn conversations with history

  3. System & Developer Instructions

    • Guide model behavior with roles like system or developer

  4. Streaming Chat

    • Live, incremental responses (planned)

Next Steps​

Before sending chat requests: