Skip to main content
The Agent Factory module provides functions for creating and managing AI agent teams with different specialized capabilities.

​
create_model

Create a model instance for a specific user. 
def create_model(user_id: str)
​
user_id
str
required
User ID for model configuration
​
return
OpenAILike
Configured model instance (Groq or Custom provider)

​
Example

from agent.agent_factory import create_model

model = create_model("user_123")

​
create_team_for_user

Create a full AI Team for a specific user with multiple specialized agents. 
def create_team_for_user(user_id: str, client=None)
​
user_id
str
required
User ID to create team for
​
client
discord.Client
default:"None"
Discord client instance for bio tools
​
return
tuple
Tuple of (model, team) instances

​
Team Members

The created team includes:
  • Code Agent: Sandbox execution, code generation (GPT-5)
  • Perplexity Agent: Real-time web research (Sonar Pro)
  • Compound Agent: Fast code execution (Groq Compound)
  • Context QnA Agent: Chat history analysis
  • MCP Agent (optional): MCP tool interactions

​
Example

from agent.agent_factory import create_team_for_user

model, team = create_team_for_user("user_123")

​
get_or_create_team

Get existing team for a user or create a new one. Implements LRU cache with automatic resource cleanup. 
async def get_or_create_team(user_id: str, client=None)
​
user_id
str
required
User ID to get or create team for
​
client
discord.Client
default:"None"
Discord client instance
​
return
Team
Team instance (existing or newly created)

​
Caching Behavior

  • Maintains up to MAX_AGENTS teams in memory
  • Uses LRU eviction when cache is full
  • Automatically cleans up evicted team resources
  • Thread-safe with per-user locks

​
Example

from agent.agent_factory import get_or_create_team

team = await get_or_create_team("user_123")
result = await team.arun("Generate a report")

​
get_prompt

Retrieve system prompt from Phoenix or fallback to local prompt. 
def get_prompt() -> str
​
return
str
System prompt content

​
Example

from agent.agent_factory import get_prompt

prompt = get_prompt()

​
Configuration

The module uses these environment variables:
  • PROVIDER: Model provider (groq or custom)
  • MODEL_NAME: Model identifier
  • CUSTOM_PROVIDER_API_KEY: Custom provider API key
  • GROQ_API_KEY: Groq API key
  • MAX_AGENTS: Maximum teams in cache (default: 100)
  • POSTGRES_URL: Database URL for sessions
  • FIRECRAWL_API_KEY: Optional Firecrawl integration

Build docs developers (and LLMs) love

Get started for freeTalk to us