AI_Devs 4 is a Polish course on building applications with large language models. Its five weeks cover programming with LLMs, context management, nondeterminism and evaluation, working with AI, and generative applications. Every lesson ends with a practical task in capture-the-flag format.
This repository holds one solution script per lesson, some rewrites and side quests, and its biggest part: Dave.
Dave#
Dave is an autonomous agent for solving the course tasks:
- a coordinator and planner that break a task into steps;
- a model router that picks a model per step (OpenAI models or local ones through Ollama or LM Studio);
- a RestrictedPython sandbox for running generated code;
- knowledge and memory modules;
- a tool registry: HTTP, LLM, vision, audio, PDF, scraping, e-mail, graph, shell, and spawning or delegating to sub-agents;
- an MCP server and a web interface.
Every LLM call is traced with LangFuse, and course data and model outputs are cached on disk, so runs are reproducible. A rule for the coding agents working in the repo: validate the answer before reporting success.
Tech: Python, uv, OpenAI API, Ollama, LangFuse, PromptFoo, Qdrant, MCP, FireCrawl, Docker.