peargent 0.1.0

A modern Python framework for building intelligent AI agents with advanced tracing, history management, and seamless LLM integration

Peargent

A modern Python framework for building intelligent AI agents with advanced tracing, history management, and seamless LLM integration.

Features

• Advanced Tracing - Track every action, decision, and API call with detailed telemetry and database persistence
• Smart History Management - Built-in conversation history with intelligent context windowing and buffer management
• Multi-LLM Support - Seamlessly switch between OpenAI, Anthropic, Groq, and more
• Type-Safe Tools - Pydantic-powered tool system with automatic validation
• Agent Pools - Run multiple agents concurrently with shared context
• Streaming Support - Real-time streaming responses with event handling
• Cost Tracking - Monitor token usage and costs across all LLM providers

Installation

pip install peargent

Optional Dependencies

For PostgreSQL database tracing:

pip install peargent[postgresql]

Quick Start

Basic Agent

from peargent import create_agent
from peargent.models import groq

agent = create_agent(
    name="assistant",
    persona="You are a helpful AI assistant.",
    model=groq("llama-3.3-70b-versatile")
)

result = agent.run("What is the capital of France?")
print(result)

Agent with Tools

from peargent import create_agent, tool
from peargent.models import openai

@tool
def get_weather(city: str) -> str:
    """Get the current weather for a city."""
    return f"The weather in {city} is sunny and 72 degrees F"

agent = create_agent(
    name="weather_bot",
    persona="You are a weather assistant.",
    model=openai("gpt-4"),
    tools=[get_weather]
)

result = agent.run("What's the weather in San Francisco?")
print(result)

Database Tracing

from peargent import create_agent
from peargent.models import openai
from peargent.telemetry import DatabaseTracer

# SQLite
tracer = DatabaseTracer(
    db_type="sqlite",
    db_path="./traces.db"
)

agent = create_agent(
    name="traced_agent",
    persona="You are a helpful assistant.",
    model=openai("gpt-4o"),
    tracer=tracer
)

result = agent.run("Explain quantum computing")
# All traces automatically saved to database

Streaming Responses

from peargent import create_agent
from peargent.models import openai

agent = create_agent(
    name="streaming_agent",
    persona="You are a helpful assistant.",
    model=openai("gpt-4o")
)

for chunk in agent.stream("Tell me a story"):
    print(chunk, end="", flush=True)

Agent Pools

from peargent import AgentPool
from peargent.models import openai, groq

pool = AgentPool(
    agents=[
        {"name": "researcher", "persona": "Research expert", "model": openai("gpt-4o")},
        {"name": "writer", "persona": "Content writer", "model": groq("llama-3.3-70b-versatile")}
    ]
)

results = pool.run_all("Explain artificial intelligence")
for agent_name, result in results.items():
    print(f"{agent_name}: {result}")

Core Concepts

Agents

Agents are autonomous entities that can process requests, use tools, maintain conversation history, and interact with LLMs.

Tools

Tools are Python functions decorated with @tool that agents can call. They're automatically validated using Pydantic schemas.

Tracing

The telemetry system tracks all agent activities including LLM calls, tool usage, token consumption, and costs. Traces can be saved to SQLite or PostgreSQL databases.

History Management

Built-in conversation history with configurable buffer sizes, context windowing, and automatic truncation strategies.

Models

Unified interface for multiple LLM providers:

• openai() - OpenAI models
• anthropic() - Anthropic Claude models
• groq() - Groq models
• More providers coming soon

Documentation

Full documentation is available at [your-docs-url-here]

Examples

Check out the /examples directory for more comprehensive examples:

• Multi-agent pools
• Database tracing
• Streaming responses
• Custom tools
• Cost tracking
• And more!

Requirements

• Python >= 3.9
• requests >= 2.31
• python-dotenv >= 1.0
• pydantic >= 2.0

License

MIT License - see LICENSE file for details

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

Author

Created by Quanta-Naut

Links

• GitHub Repository
• Documentation
• PyPI Package
• Issue Tracker