56 lines
1.9 KiB
Python
56 lines
1.9 KiB
Python
import json # Import the json module
|
|
from qwen_agent.agents import Assistant
|
|
|
|
# Define LLM
|
|
llm_cfg = {
|
|
'model': 'qwen3:0.6B',
|
|
|
|
# Use the endpoint provided by Alibaba Model Studio:
|
|
# 'model_type': 'qwen_dashscope',
|
|
# 'api_key': os.getenv('DASHSCOPE_API_KEY'),
|
|
|
|
# Use a custom endpoint compatible with OpenAI API:
|
|
'model_server': 'http://localhost:11434/v1', # api_base
|
|
'api_key': 'EMPTY',
|
|
|
|
# Other parameters:
|
|
# 'generate_cfg': {
|
|
# # Add: When the response content is `<think>this is the thought</think>this is the answer;
|
|
# # Do not add: When the response has been separated by reasoning_content and content.
|
|
# 'thought_in_content': True,
|
|
# },
|
|
}
|
|
|
|
# Define Tools
|
|
tools = [
|
|
{'mcpServers': { # You can specify the MCP configuration file
|
|
'time': {
|
|
'command': 'uvx',
|
|
'args': ['mcp-server-time', '--local-timezone=Asia/Shanghai']
|
|
},
|
|
"fetch": {
|
|
"command": "uvx",
|
|
"args": ["mcp-server-fetch"]
|
|
}
|
|
}
|
|
},
|
|
'code_interpreter', # Built-in tools
|
|
]
|
|
|
|
# Define Agent
|
|
bot = Assistant(llm=llm_cfg, function_list=tools)
|
|
|
|
# Streaming generation
|
|
messages = [{'role': 'user', 'content': 'fetch https://qwenlm.github.io/blog/ and give me a summery of the energy efficiency of training and inference of Qwen'}]
|
|
# Initialize responses variable before the loop in case the loop doesn't run
|
|
final_responses = None
|
|
for responses in bot.run(messages=messages, enable_thinking=True,max_tokens=8192):
|
|
# The loop assigns the latest response to final_responses
|
|
final_responses = responses
|
|
|
|
# Pretty-print the final response object
|
|
if final_responses:
|
|
print(json.dumps(final_responses, indent=2)) # Use indent=2 (or 4) for pretty printing
|
|
print(final_responses['content'])
|
|
else:
|
|
print("No response received from the agent.") |