Tool calling
Let a model request a function that your application executes deliberately.
Describe a tool
Add a tools array with a function name, description and JSON Schema parameters. Choose a model with tools support. provider.require_parameters: true avoids silently dropping required tool settings on an incompatible route.
Offer one function
The example describes a weather lookup. tool_choice: auto lets the model decide whether to request it; declaring a function does not execute any code in your application.
curl
curl --request POST \
--url 'https://api.tokenlane.de/v1/chat/completions' \
--header 'Authorization: Bearer YOUR_API_KEY' \
--header 'Content-Type: application/json' \
--data '{
"model": "deepseek/deepseek-v4-flash-0731",
"messages": [
{
"role": "user",
"content": "Hello"
}
],
"max_tokens": 400,
"tools": [
{
"type": "function",
"function": {
"name": "get_weather",
"description": "Read the weather for a city.",
"parameters": {
"type": "object",
"properties": {
"city": {
"type": "string"
}
},
"required": [
"city"
],
"additionalProperties": false
}
}
}
],
"tool_choice": "auto",
"provider": {
"require_parameters": true
}
}'
node
import OpenAI from 'openai';
const client = new OpenAI({ baseURL: "https://api.tokenlane.de/v1", apiKey: process.env.TOKENLANE_API_KEY });
const response = await client.chat.completions.create({
"model": "deepseek/deepseek-v4-flash-0731",
"messages": [
{
"role": "user",
"content": "Hello"
}
],
"max_tokens": 400,
"tools": [
{
"type": "function",
"function": {
"name": "get_weather",
"description": "Read the weather for a city.",
"parameters": {
"type": "object",
"properties": {
"city": {
"type": "string"
}
},
"required": [
"city"
],
"additionalProperties": false
}
}
}
],
"tool_choice": "auto",
"provider": {
"require_parameters": true
}
}, {});
console.log(response);
python
import os
import json
from openai import OpenAI
client = OpenAI(base_url="https://api.tokenlane.de/v1", api_key=os.environ["TOKENLANE_API_KEY"])
body = json.loads(r'''{
"model": "deepseek/deepseek-v4-flash-0731",
"messages": [
{
"role": "user",
"content": "Hello"
}
],
"max_tokens": 400,
"tools": [
{
"type": "function",
"function": {
"name": "get_weather",
"description": "Read the weather for a city.",
"parameters": {
"type": "object",
"properties": {
"city": {
"type": "string"
}
},
"required": [
"city"
],
"additionalProperties": false
}
}
}
],
"tool_choice": "auto",
"provider": {
"require_parameters": true
}
}''')
response = client.chat.completions.create(
model=body["model"],
messages=body["messages"],
max_tokens=body["max_tokens"],
tool_choice=body["tool_choice"],
tools=body["tools"],
extra_body=json.loads(r'''{
"provider": {
"require_parameters": true
}
}''')
)
print(response)
vercel
// This operation uses HTTP directly.
const response = await fetch("https://api.tokenlane.de/v1/chat/completions", {
method: "POST",
headers: {
"Authorization": "Bearer " + process.env.TOKENLANE_API_KEY,
"Content-Type": "application/json",
},
body: `{
"model": "deepseek/deepseek-v4-flash-0731",
"messages": [
{
"role": "user",
"content": "Hello"
}
],
"max_tokens": 400,
"tools": [
{
"type": "function",
"function": {
"name": "get_weather",
"description": "Read the weather for a city.",
"parameters": {
"type": "object",
"properties": {
"city": {
"type": "string"
}
},
"required": [
"city"
],
"additionalProperties": false
}
}
}
],
"tool_choice": "auto",
"provider": {
"require_parameters": true
}
}`,
});
console.log(await response.text());
Complete the conversation
Inspect message.tool_calls, validate the function name and parsed arguments, apply your permissions and execute the allowed action. Append the assistant tool-call message and a tool message with the matching tool_call_id, then request the final answer. Bound the number of tool rounds and handle parallel calls only if your implementation supports them.
Keep execution safe
Treat arguments as untrusted model output. Never let a model invent permissions, arbitrary code or destinations. If no route supports the requested tool shape, choose an eligible model; do not remove validation to make the request succeed. Use timeouts and an explicit error result when a tool fails.