astra-runtime 0.1.0

Astra Runtime - Build AI agents, teams, and RAG pipelines in Python with embedded runtime and FastAPI server

Astra Runtime

🚀 Build AI agents, teams, and RAG pipelines in pure Python.

Astra Runtime provides everything you need to build intelligent AI applications:

• Embedded Mode: Use agents directly in your Python code
• Server Mode: Deploy as REST APIs with FastAPI

Installation

pip install astra-runtime

Optional Dependencies

# With MongoDB support
pip install astra-runtime[mongodb]

# With AWS Bedrock support
pip install astra-runtime[aws]

# Everything
pip install astra-runtime[all]

Quick Start

Embedded Mode

import asyncio
from astra import Agent, Gemini

agent = Agent(
    model=Gemini(model="gemini-2.0-flash"),
    instructions="You are a helpful assistant"
)

async def main():
    response = await agent.invoke("Hello!")
    print(response.content)

asyncio.run(main())

Server Mode

from astra import Agent, Gemini
from astra.server import create_app

agent = Agent(
    name="assistant",
    model=Gemini(model="gemini-2.0-flash"),
    instructions="You are a helpful assistant"
)

app = create_app(agents={"assistant": agent})

# Run with: uvicorn main:app --reload

With Tools

from astra import Agent, Tool, tool, Gemini

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

agent = Agent(
    model=Gemini(model="gemini-2.0-flash"),
    instructions="You are a weather assistant",
    tools=[get_weather]
)

RAG (Retrieval-Augmented Generation)

from astra import Agent, Rag, LanceDB, HuggingFaceEmbedder, Gemini

# Create RAG pipeline
rag = Rag(
    vector_db=LanceDB(path="./my_db"),
    embedder=HuggingFaceEmbedder()
)

# Ingest documents
await rag.ingest("path/to/documents/")

# Create agent with RAG
agent = Agent(
    model=Gemini(model="gemini-2.0-flash"),
    instructions="Answer questions using the provided context",
    rag=rag
)

Features

Feature	Description
🤖 Agents	Build intelligent agents with tools, memory, and context
📚 RAG	Retrieval-Augmented Generation with custom pipelines
🗄️ Storage	LibSQL, MongoDB, and LanceDB backends
🛡️ Guardrails	PII filtering, content moderation, prompt injection detection
🔧 Tools	Easy function calling with @tool decorator
👥 Teams	Multi-agent collaboration and delegation
🌐 Server	FastAPI-based REST API with streaming support
💾 Memory	Short-term and long-term agent memory
🔌 Middleware	Input/output processing pipelines

Model Support

• Google Gemini: Gemini(model="gemini-2.0-flash")
• OpenAI: OpenAI(model="gpt-4")
• AWS Bedrock: Bedrock(model="anthropic.claude-3")
• HuggingFace Local: HuggingFaceLocal("Qwen/Qwen2.5-0.5B-Instruct")

Documentation

• 📖 Examples
• 🐛 Issues
• 💬 Discussions

License

MIT License - see LICENSE for details.