I'm excited to open source Fury. A lightweight python package making it easy to build with LLMs, even if it's your first time.
https://github.com/huwprosser/fury
With Fury the following things work out-of-the-box:
- Parallel Tool Calls
- Tool Call announcements eg "I am using the bash tool" (great for UI/UX)
- Multimodal inputs
- Audio input transcription.
from fury import Agent
agent = Agent(
model="your-model-name", # e.g., "gpt-4o" or a local model
system_prompt="You are a helpful assistant.",
base_url="http://127.0.0.1:8080/v1", # or https://openrouter.ai/api/v1, https://api.openai.com/v1
api_key="your-api-key",
)
response = agent.ask("Hello!", history=[])
print(response)
Examples:
- By far the most powerful example I have built so far is the coding agent. It is based on Pi.dev (powering OpenClaw) and offers compaction, extensible skills and terminal control out the box.
- The CRON example shows you how to create an OpenClaw-style heartbeat system with Fury (room for improvement here).
Pip package coming soon! For now, install directly from Github.
Build neat stuff 🚀