gradient-agent 0.1.0

Python SDK for building and instrumenting AI agents with FastAPI integration

Gradient Agent

A Python SDK for building and instrumenting AI agents with FastAPI integration and comprehensive runtime tracking.

Features

• @entrypoint decorator - Easy FastAPI server creation from agent functions
• Runtime instrumentation - Complete observability for LangGraph-based agents
• Request tracking - Monitor agent performance and behavior
• Context management - Maintain state across agent executions

Quick Start

from gradient_sdk import entrypoint

@entrypoint
def my_agent(prompt: str, max_tokens: int = 100) -> str:
    """Simple echo agent"""
    return f"Echo: {prompt} [max_tokens={max_tokens}]"

# The decorator automatically creates a FastAPI server
# Access at http://localhost:8080 when running

Installation

pip install gradient-agent

Usage

Basic Agent

from gradient_agent import entrypoint

@entrypoint
def simple_agent(prompt: str) -> str:
    return f"Response to: {prompt}"

LangGraph Integration

from gradient_agent import entrypoint
from langgraph.graph import StateGraph
from gradient_agent.runtime import get_runtime_manager

@entrypoint
def langgraph_agent(prompt: str) -> str:
    # Your LangGraph agent code here
    # Runtime tracking is automatically enabled
    pass

Manual Server Control

from gradient_agent import get_app
import uvicorn

# Import your agent modules first
import my_agent

# Get the FastAPI app
app = get_app()

# Run with custom configuration
uvicorn.run(app, host="0.0.0.0", port=8080)

API Endpoints

When you use the @entrypoint decorator, your agent automatically gets:

• POST /completions - Main agent endpoint
• GET /health - Health check
• GET / - Basic info

Runtime Tracking

The SDK includes comprehensive runtime tracking for LangGraph-based agents:

• Node execution timing
• Request context management
• Performance metrics
• Error tracking

Development

This SDK is part of the larger Gradient ecosystem for AI agent development and deployment.

License

MIT