Skip to content

Home

Jump to bottom
kwarodom edited this page May 19, 2026 · 2 revisions

Agentic Coding Fitness — Wiki

Welcome to the wiki for Agentic Coding Fitness @ Rust Tech Bar — a weekly hands-on "Vibe Coding" series in Bangkok where we build AI muscle memory by shipping real agentic systems instead of sitting through lectures.

What this repo is

A living, week-by-week codebase that walks from plain LLM API calls all the way to multi-agent systems, RAG, MCP tool servers, and physical IoT (smart lights + a Tello drone). Each weekN/ directory is one Tuesday session — usually a small runnable script plus the PDF/slides that go with it.

The philosophy: practice first, ship workable solutions, touch the real world. No deep frameworks until they earn their place; no abstractions until concrete pain demands them.

Curriculum at a glance

Week Theme Highlights
2 Claude API foundations Single-turn, streaming, multi-turn conversation
3 Tool use & smart assistants Function calling; Tapo L530 smart-light control via local HTTP wrapper
4 Autonomous pipelines & hardware Research pipeline → NotebookLM; Tello drone flight choreography
5 Auto-agents Self-directed agent loops
6 TypeScript agents + memory AGENTS.md/CLAUDE.md patterns, Jest, memory notes
7 MCP & agent tooling MCP filesystem server, Poe API, tool-search agents
8 RAG & knowledge agents RAG lab
9 Multi-agent systems (MAS) CrewAI, LangGraph, AG2, parallel swarms — sequential, supervisor, and parallel topologies
11 Agent Forge: Mastery Lab Self-contained index.html web app to revise, simulate, and play with every topology from weeks 1–10 (Sim + Live API modes)
12 (placeholder)
14 MAS design-thinking workshop Pre-test, KG coding, design-thinking workshop, answers

Detail pages: see the sidebar / week-specific wiki pages as they're added.

Repo layout

weekN/                  one Tuesday session each
claudemanagedagent/     experimental CMMS / FDD agent (dede_afddv2.py)
AgentMemoryDemo/        memory pattern demo
check_tapo.py, scan.py  Tapo smart-light glue
ollama_stream.py        local Ollama streaming client
SKILL.md                PR-description authoring skill
GSD_vs_PlanMD_Analysis.md   methodology comparison notes
requirements.txt        Python deps (Anthropic SDK, djitellopy, duckduckgo-search, ...)
pyproject.toml + uv.lock    uv-managed Python project

Getting started

  1. Python 3.10+ — clone, create venv, pip install -r requirements.txt (or uv sync).
  2. Drop an ANTHROPIC_API_KEY into .env at the repo root.
  3. Hardware (optional):
    • Tapo: edit tapo_config.json with your TP-Link creds + bulb LAN IP.
    • Tello: connect your laptop to the drone's Wi-Fi SSID before running week4/dronecontrol.py.
  4. Some weeks (6, 11) use Node — npm install inside that week folder.

How to use this wiki

  • Home (this page): orientation and curriculum map.
  • Per-week pages: deeper notes, gotchas, and follow-up exercises — added as sessions happen.
  • Cross-cutting topics: MCP, RAG, MAS topologies, agent memory, hardware integration.

If you attended a session and want to add notes, edit the relevant week page directly — the wiki is a Git repo (agenticaicodingfitness.wiki.git), so PR-style etiquette doesn't apply; just push.

Goals

  • Build real stuff — not toy demos.
  • Solve real problems — IoT, knowledge work, ops.
  • Generate real impact — code you'd actually deploy on Monday.

See you Tuesday at Rust Bar.

Clone this wiki locally