🛠 Agents-Sandbox: An Interactive AI Agent Laboratory
The Agents-Sandbox project has been released to study AI agent architectures based on LangChain and Ollama. The laboratory allows for comparing memory performance (buffer, summary), tool calling (ReAct, native bind_tools), and various RAG systems (Chroma, LlamaIndex, Haystack, NumPy) through a single CLI.
🌍 The project demonstrates a modular approach to development, allowing for a visual comparison of the effectiveness of retrieval methods (RAG) and routing strategies on local models.
👤 It is ideal for transitioning from simple chats to complex agentic systems, allowing users to see "under the hood" how a model makes decisions regarding function calling or document searching.