AgentHub
the only SDK you need to connect to state-of-the-art LLMs
4 followers
the only SDK you need to connect to state-of-the-art LLMs
4 followers
AgentHub changes that: 1. Async, stateful, streaming interface for multi-turn agentic execution—switch models with zero code changes. 2. Model capabilities (e.g. interleaved thinking, caching) are validated and aligned—no performance loss when switching models. It also ships with a lightweight, fine-grained tracing board for auditing LLM executions—just add a single parameter and every run is permanently traced.
Hi Product Hunt! 👋
What makes it different:
🚀 Zero Code Changes: A unified, asynchronous, and stateful API that handles provider quirks automatically—significantly flattening the learning curve.
🧠 No Performance Loss: We rigorously align model-specific features like interleaved thinking and caching, ensuring 100% reasoning fidelity with no manual handling.
🔍 One-Parameter Tracing: Permanently audit every LLM execution with zero complex setup.
AgentHub is fully open-source and built for developer ergonomics.
GitHub：https://github.com/Prism-Shadow/AgentHub
I’d love to know: what’s the most annoying API quirk you’ve had to deal with? I'm here to answer any questions! 🛠️