Tool calling

Tool calling (function calling) lets a model ask your code to run a function and then use the result in its answer. api14 uses the OpenAI tools format for every model that supports tools: OpenAI, Anthropic, Google, Mistral, Qwen, DeepSeek and others. To find tool-capable models, filter by supported parameters on the Models list (supported_parameters includes tools).

Request parameters

import OpenAI from "openai";

const client = new OpenAI({
  baseURL: "https://api14.com/api/v1",
  apiKey: process.env.API14_API_KEY,
});

const result = await client.chat.completions.create({
  "model": "openai/gpt-5-mini",
  "messages": [
    {
      "role": "user",
      "content": "What is the weather in Istanbul?"
    }
  ],
  "tools": [
    {
      "type": "function",
      "function": {
        "name": "get_weather",
        "description": "Get the current weather",
        "parameters": {
          "type": "object",
          "properties": {
            "city": {
              "type": "string"
            }
          },
          "required": [
            "city"
          ]
        }
      }
    }
  ]
});
console.log(result.choices[0].message.tool_calls);

The loop

1. Send messages and tools.
2. If the response has finish_reason: "tool_calls", run each entry in message.tool_calls.
3. Add the assistant message, then one { "role": "tool", "tool_call_id", "content" } message per call, and send the request again.
4. Repeat until the model answers with plain content.

Response with a tool call:

{
  "id": "gen-1760000000-tool1",
  "model": "anthropic/claude-sonnet-4.5",
  "choices": [
    {
      "index": 0,
      "finish_reason": "tool_calls",
      "message": {
        "role": "assistant",
        "content": null,
        "tool_calls": [
          {
            "id": "call_01",
            "type": "function",
            "function": { "name": "get_weather", "arguments": "{\"city\":\"Istanbul\"}" }
          }
        ]
      }
    }
  ],
  "usage": { "prompt_tokens": 182, "completion_tokens": 21, "total_tokens": 203, "cost": 0.000861 }
}

The follow-up request includes the tool result:

{
  "model": "anthropic/claude-sonnet-4.5",
  "tools": [ ... same tools ... ],
  "messages": [
    { "role": "user", "content": "What's the weather in Istanbul?" },
    { "role": "assistant", "content": null, "tool_calls": [ { "id": "call_01", "type": "function", "function": { "name": "get_weather", "arguments": "{\"city\":\"Istanbul\"}" } } ] },
    { "role": "tool", "tool_call_id": "call_01", "content": "{\"temp_c\": 21, \"sky\": \"clear\"}" }
  ]
}

Keep sending tools with every request in the loop. Many providers reject a tool message if the request has no tools definitions.

Tips

• arguments is a string of JSON. Parse it and validate it, because models can produce invalid JSON.
• When streaming, tool calls arrive in delta.tool_calls fragments keyed by index. Join the arguments strings before parsing them.
• To route only to providers that support tools, set provider: { "require_parameters": true }. See Provider routing.
• Each loop iteration is a separate billed request.

Errors

A model that doesn't support tools may return an upstream 400 or 404 ("No endpoints found that support tool use"). Other errors are the same as for Chat completions. See Errors.