Flux RouterDocumentation

Tools and function calling

Give the model functions to call using the standard OpenAI tools and tool_calls flow through FluxRouter.

Tool calling (also called function calling) lets the model ask your code to run a function and feed the result back into the conversation. FluxRouter speaks the OpenAI wire format, so you use the standard tools and tool_calls flow against https://api.fluxrouter.ai/v1/chat/completions. Nothing about the request is Flux-specific except the base URL and the flux-auto model.

How the flow works

  1. You send the conversation plus a tools array describing the functions the model is allowed to call.
  2. If the model decides to call a function, it responds with tool_calls instead of a final answer. Each tool call has an id, a function name, and JSON arguments.
  3. Your code runs the function and appends a message with role: "tool", the matching tool_call_id, and the function's result.
  4. You send the conversation back. The model uses the tool result to produce its final answer.

Minimal example

This example gives the model one tool, get_weather, and handles the round trip.

python
import json
from openai import OpenAI

client = OpenAI(
    api_key="sk-...",                          # your Flux key
    base_url="https://api.fluxrouter.ai/v1",   # the one line you change
)

tools = [
    {
        "type": "function",
        "function": {
            "name": "get_weather",
            "description": "Get the current weather for a city.",
            "parameters": {
                "type": "object",
                "properties": {
                    "city": {"type": "string", "description": "City name"}
                },
                "required": ["city"],
            },
        },
    }
]

messages = [{"role": "user", "content": "What's the weather in Lisbon?"}]

# 1. First call: the model decides whether to use the tool.
first = client.chat.completions.create(
    model="flux-auto",
    messages=messages,
    tools=tools,
)

choice = first.choices[0].message
messages.append(choice)

# 2. Run any tool calls the model requested.
if choice.tool_calls:
    for call in choice.tool_calls:
        args = json.loads(call.function.arguments)
        result = {"city": args["city"], "temp_c": 21, "sky": "clear"}  # your real function here
        messages.append(
            {
                "role": "tool",
                "tool_call_id": call.id,
                "content": json.dumps(result),
            }
        )

    # 3. Second call: the model answers using the tool result.
    final = client.chat.completions.create(
        model="flux-auto",
        messages=messages,
        tools=tools,
    )
    print(final.choices[0].message.content)
else:
    print(choice.content)

curl (single turn)

bash
curl https://api.fluxrouter.ai/v1/chat/completions \
  -H "Authorization: Bearer $FLUX_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "flux-auto",
    "messages": [{ "role": "user", "content": "What is the weather in Lisbon?" }],
    "tools": [
      {
        "type": "function",
        "function": {
          "name": "get_weather",
          "description": "Get the current weather for a city.",
          "parameters": {
            "type": "object",
            "properties": { "city": { "type": "string" } },
            "required": ["city"]
          }
        }
      }
    ]
  }'

Forcing or disabling tool use

The standard OpenAI tool_choice field works as usual:

  • "tool_choice": "auto" (default) lets the model decide.
  • "tool_choice": "none" forbids tool use for this turn.
  • "tool_choice": {"type": "function", "function": {"name": "get_weather"}} forces a specific tool.

Make sure your model supports tools

Tool calling depends on the model that serves the request. flux-auto routes to a sensible model, and most general-purpose models in the Flux catalog support tools, but support is a property of the underlying model, not of FluxRouter.

If your application always needs tool calling, pin a tool-capable model instead of relying on the router. Pass a flux-pinned-* id (for example flux-pinned-claude-sonnet or flux-pinned-gpt-5) so every request lands on a model you have verified supports tools. See Models for the full list of pinned aliases.

Anthropic shape

If you call the Anthropic-compatible base at https://api.fluxrouter.ai/anthropic, use Anthropic's native tools and tool_use / tool_result blocks exactly as you would calling Anthropic directly. The mechanics differ by wire format, but the round trip is the same: the model requests a tool, you run it, you send the result back.