OLLAMA_LLM
Built 5 AI agent challenges using AWS Strands SDK and Ollama llama3.2:3b — covering conversational agents, tool calling, persistent memory with Mem0 and FAISS, and an MCP protocol chatbot. Fully local, no API keys required.
Builders Skill Sprint — AI Agents with Strands SDK & Ollama
This project is a series of 5 progressive AI agent challenges built using the AWS Strands Agents SDK and Ollama (llama3.2:3b), running fully locally with no cloud API keys required.
Challenge 1 — Simple Agent: Built a minimal conversational agent using OllamaModel and Strands Agent class with an interactive chat loop and automatic conversation history.
Challenge 2 — Tools Agent: Extended the agent with 3 custom tools — a calculator, a weather checker, and an age calculator. The agent autonomously selects the right tool based on the user's question using the @tool decorator pattern.
Challenge 3 — Memory Agent: Added persistent long-term memory using Mem0, FAISS vector storage, and Ollama's nomic-embed-text embedding model. The agent stores and recalls personal facts across sessions entirely on local disk.
Challenge 4 — Full Agent: Combined all tools and memory into one unified agent with 6 tools and smart system prompt routing, enabling cross-capability reasoning like recalling a stored city and then checking its weather in a single response.
Challenge 5 — MCP Chatbot: Implemented the Model Context Protocol (MCP) using a local FastMCP server and Strands MCPClient. Tools are served over stdio JSON-RPC, auto-discovered at runtime, and used by the agent through the open MCP standard.
Enjoyed reading this content? Let the author know!
Your likes, comments, shares, and saves help creators reach more builders.
Loading recommendations
Loading article