docs/chat-template/tool_demo.py

"""
## 来源


## 参考

- [Llama-3.1的工具调用 | 2024/08](https://huggingface.co/meta-llama/Llama-3.1-405B-Instruct)
- [huggingface统一工具调用 | 2024/08/12](https://huggingface.co/blog/unified-tool-use)


"""

import os
import json
from transformers import AutoTokenizer
from transformers.utils import get_json_schema


# MODEL_PATH = "meta-llama/Llama-3.1-405B-Instruct"
# MODEL_PATH = "NousResearch/Hermes-3-Llama-3.1-405B"  # messages里的tool_calls必须要有content字段
# MODEL_PATH = "../../test/Llama-4-Maverick-17B-128E-Instruct/"
# MODEL_PATH = "meta-llama/Llama-4-Maverick-17B-128E-Instruct"
# MODEL_PATH = "Qwen/Qwen3-235B-A22B-Instruct-2507"
# MODEL_PATH = "mistralai/Mistral-7B-Instruct-v0.1"  # messages里不支持tool_calls，不支持 role=tool，不支持 tools
# MODEL_PATH = "mistralai/Ministral-8B-Instruct-2410" # 支持 tools, 支持tool_calls(必须要有id), 格式非主流
# MODEL_PATH = "deepseek-ai/DeepSeek-R1"  # 不支持tools，tool_calls也有问题
# MODEL_PATH = "deepseek-ai/DeepSeek-V3.1"
# MODEL_PATH = "google/gemma-3-27b-it"  # 不支持任何tool
# MODEL_PATH = "moonshotai/Kimi-K2-Instruct"
MODEL_PATH = "xai-org/grok-2"

tokenizer = AutoTokenizer.from_pretrained(MODEL_PATH)

# First, define a tool
def get_current_temperature(location: str) -> float:
    """
    Get the current temperature at a location.
    
    Args:
        location: The location to get the temperature for, in the format "City, Country"
    Returns:
        The current temperature at the specified location in the specified units, as a float.
    """
    return 22.  # A real function should probably actually get the temperature!

# Next, create a chat and apply the chat template
messages = [
  {"role": "system", "content": "You are a bot that responds to weather queries."},
  {"role": "user", "content": "Hey, what's the temperature in Paris right now?"},
#   {"role": "assitant", "content": "test1"},
#   {"role": "user", "content": "test2"},
]

# step1:
print("###" * 10, "llm with tools", MODEL_PATH)
print(json.dumps(messages, ensure_ascii=False, indent=2))
inputs = tokenizer.apply_chat_template(messages, tools=[get_current_temperature], add_generation_prompt=True, tokenize=False)
print("###" * 5, "prompt")
print(inputs)
json_schema = get_json_schema(get_current_temperature)
print("###" * 5, "json_schema")
print(json.dumps(json_schema, ensure_ascii=False, indent=2))


# step2: 调用 LLM，以下是LLM的返回
tool_call = {"name": "get_current_temperature", "arguments": {"location": "Paris, France"}}
# messages.append({"role": "assistant", "tool_calls": [{"type": "function", "function": tool_call}]})
# messages.append({"role": "assistant", "tool_calls": [{"type": "function", "function": tool_call}], "content": ""}) # Hermes-3-Llama-3.1-405B 的content 不能为空
# messages.append({"role": "assistant", "tool_calls": [{"id": "123456789", "type": "function", "function": tool_call}]})  # Ministral-8B-Instruct-2410 仍然报错

# step3: 调用tool，以下是 tool 的返回
messages.append({"role": "tool", "name": "get_current_temperature", "content": "22.0"})

# step4: 再调用 LLM
print("###" * 10, "tool_calls & tool_response")
print(json.dumps(messages, ensure_ascii=False, indent=2))
inputs = tokenizer.apply_chat_template(messages, add_generation_prompt=True, tokenize=False)
print("###" * 5, 'prompt')
print(inputs)