LLM Plugin

Configuration

1. Rename PLUGINS/LLM/CONFIG.example.py to CONFIG.py
2. Add one or more LLM configurations in the LLM_CONFIGS list

Each configuration item contains the following fields:

Field	Description
api_key	API key; use "ollama" for local models
base_url	OpenAI-compatible API endpoint
model	Model name
proxy	Proxy address; use None if not needed
tags	Tag list for selecting configurations by scenario

Selecting Models by Tag

from PLUGINS.LLM.CONFIG import get_model

# Use the first configuration with the "fast" tag
client = get_model(tag="fast")

# Use the first configuration with both "powerful" and "function_calling" tags
client = get_model(tag=["powerful", "function_calling"])

# When no tag is provided, the first configuration in the list is used by default
client = get_model()

Built-in Tags

Tags used in ASP (freely customizable):

• cheap — Low-cost model
• fast — Fast response model
• powerful — High-performance model
• function_calling — Supports function calling
• structured_output — Supports structured output

Proxy Format

The proxy field supports the following formats:

"proxy": "http://192.168.1.100:3128"                          # HTTP
"proxy": "http://user:pass@192.168.1.100:3128"                # HTTP with authentication
"proxy": "socks5://192.168.1.100:1080"                        # SOCKS5
"proxy": "socks5://user:pass@192.168.1.100:1080"              # SOCKS5 with authentication

Configuration Example

LLM_CONFIGS = [
    {
        "api_key": "ollama",
        "base_url": "http://localhost:11434/v1",
        "model": "qwen3.5:9b",
        "proxy": None,
        "tags": ["fast", "cheap", "function_calling", "structured_output"]
    },
    {
        "api_key": "sk-xxx",
        "base_url": "https://dashscope.aliyuncs.com/compatible-mode/v1",
        "model": "qwen3-max",
        "proxy": None,
        "tags": ["powerful", "function_calling", "structured_output"]
    },
]