Analysis updated 2026-05-18
Copy a ready-made pi agent configuration to route tasks between multiple self-hosted models.
Reuse a research pipeline skill that retrieves, checks, and summarizes sources with citations.
Automatically generate a report on a newly released HuggingFace model.
Set up a dedicated writing subagent for drafts, outlines, and edits.
| muellerzr/learning-pi-through-force | 100/awesome-machine-learning | adam-s/car-diagnosis | |
|---|---|---|---|
| Stars | 8 | 8 | 8 |
| Language | Python | Python | Python |
| Last pushed | — | 2024-08-07 | — |
| Maintenance | — | Stale | — |
| Setup difficulty | — | easy | moderate |
| Complexity | — | 1/5 | 3/5 |
| Audience | developer | developer | researcher |
Figures from each repo's GitHub metadata at analysis time.
This repository is a personal setup shared by the author of a blog about learning to use an AI coding agent called pi. It packages up the configuration files, custom skills, and custom subagents that the author built on top of a standard pi install, so anyone else who uses pi can copy the same setup into their own machine and get the same behavior. At the center of it is a model routing configuration that decides which AI model handles which kind of task. The author runs three self-hosted models: one used for strong reasoning and planning, one used as a fast coding assistant, and one used as a cheap model for quick lookups and retrieval. A settings file assigns each type of subagent role, such as scout, planner, or worker, to one of these models along with how much the model is allowed to think before answering. On top of the models, the repository includes two custom subagents. One is a fast retriever built for pulling together research sources in parallel, deliberately using the cheaper model so the stronger model is saved for harder synthesis work. The other is an editorial agent meant for writing tasks like outlines, drafts, and rewrites. It also includes four custom skills. One is a multi-step research pipeline that scopes a question, plans research, retrieves sources, checks them against each other, and produces a final written report with citations. Another safely fetches academic papers from arXiv. A third pulls model information from HuggingFace. The last one researches a newly released HuggingFace model and produces a full report, which can optionally be emailed or sent to a Kindle device. Every credential in the repository has been replaced with a placeholder, so before using any of this, a person needs to fill in their own API keys and endpoint addresses. The install instructions are simple copy commands that drop the files into the standard pi configuration folder. This is aimed squarely at people who already use the pi agent and want to see or reuse someone else's working configuration.
A shared personal configuration for the pi AI coding agent, including custom model routing, research skills, and writing subagents.
Mainly Python. The stack also includes Python, pi agent, GLM-5.2.
Mainly developer.
This repo across BitVibe Labs
Verify against the repo before relying on details.