[p]rog[mo]

by whitmo

117 downloads
Not rated
GitHub

About

Experimental project to create an MCP agent to help with coding

Details

Author
whitmo
Downloads
117
Categories
Other

- Knowledge management with a local vector store (Qdrant) for CRUD on tokenized text.
- Documentation-driven development manager with projects, resources, and archives.
- Automated code review with branch creation, test addition, and iterative adjustment.
- Test manager focused on running, fixing, and improving test coverage.
- Configurable via .codereview or .clinefile for code review guides.
- Constant mode for periodic code review and change logging.

The server is distributed as a Rust binary or container. It works with MCP clients such as Cline. Detailed installation and configuration steps are not provided in the README.

[p]rog[mo]

Coverage Status

_program more_

An agent for handling out of band common coding tasks

- knowledge management
- documentation driven development
- code review
- test writing and running and gardening

Knowledge Management

p-mo works with a vector datastore to provide specific context to Cline and other mcp clients. It also also for the basic crud operations of tokenizing text sources, uploading them to the vector store and deleting them when no longer needed.

Supported
- Qdurant: containerized locally run vector store.

Research

Graph RAG - https://www.semanticpartners.com/post/a-triple-store-rag-retriever - https://github.com/indradb/indradb

DDD manager

Starting with the high level "why", manages our critical path and the narrative of our build. Records features and their user stories, technical specification, decisions and everything needed for operating our system. Uses the "external brain" format for knowledge management: projects for planning active efforts (w/ completion horizons), resources for active reference and policy material, archive for archived resources and projects.

Code Review

Using a .codereview file or the .clinefile as a guide (plus any initial prompting), cuts a review branch and iterates over the code, adds tests, runs tests, makes adjustments. Can be run in a constant mode which periodically pulls in latest commits and keeps a change log of review comments and change commits

Test manager

In some ways, more or less the same as "Code Review", but focussed only on running and fixing tests, improving coverage, and changes to the code to improve testing isolation, speed, layering, etc. Can provide "last run" data for a repo's tests.

No reviews yet — be the first

Sign in to leave a review

Use Google, GitHub, or an email account so ratings stay tied to real people.

Email sign in

No reviews posted yet.