> ai-factory-planner
A task planner for AI Factory: breaks a goal into steps, selects the right skills, keeps an execution log and highlights points that need human control.
> run ./lab --watch
Experiments and research
The lab is a working bench for LLM, RAG, AI planning, code intelligence and admin workflows. I document testable hypotheses, not shiny promises: what can be automated, where a human must stay in control and how to bring it to production.
llm-playground
A shared bench for quick checks: prompts, system instructions, tool calling, JSON responses, streaming and model behavior on the same inputs.
related note RAG pipeline without magicai-factory-planner
A task planner for AI Factory: breaks a goal into steps, selects the right skills, keeps an execution log and highlights points that need human control.
related note AI Factory task planning loopcode-miner
A tool for mining repository context: file graph, dependencies, important entrypoints, risky zones and a compact context pack for an LLM.
related note Context mining for large codebasesrag-lab
A RAG bench for the retrieval layer: chunking, pgvector, rerank, eval sets, citations, guardrails and answer-quality observability.
related note RAG pipeline without magicprompt-studio
Versioning prompts and quality criteria: response comparison, regression cases, tone checks and a careful path from idea to repeatable scenario.
related note Prompt evals before prompt vibesmoonshine-ai-tools
Admin AI workflows for MoonShine: draft generation, translations, filling content blocks and controlling what actually reaches the frontend.
related note MoonShine AI workflows for content ops