| ▲ | Show HN: Replen – maps your repos to a knowledge graph to match open-source(github.com) | |
| 1 points by nsokin 4 hours ago | ||
Hi all. I built Replen as a way to improve my AI dev workflows. In essence, it automates what was a manual process for me and turns it from a reactive stance to a proactive one. That being reading/ scrolling lists and creators that showcased cool and trending repos. If i saw something I liked i’d then manually checked with Claude or Codex to see if theres anything here worth taking because I had a hunch there might be. Replen now does this for me. And cuts out a number of back and fourths. It knows what every project of mine does, what its trying to do, how its built, what it depends on, and finds genuine matches from the open source frontier (all the new projects that are just not in the training data of any of these models). Replen then builds an ontology and knowledge graph across all your repos, to find the relationships and gaps and then runs a self tuning matching loop against the wider oss frontier. It goes looking for libraries, algos, techniques and ideas that you can get your agent to adopt, port, cherry pick or rebuild. Because of the embeddings and ontology, it understands your whole portfolio and can suggest better ways to do things. i.e. I personally do a lot of scraping and Claude always suggests the same way to do it, but /davinchi/scrapling just does it better and gets past CF, now with Replen, it knows this is the better option. By no means is it perfect and the learning loop needs users and their interactions to get better for everyone so with that in mind, id love feedback on a couple of things. 1. Do you find yourself doing what I do or is this just a me thing? (Validates whether this is useful to anyone but me) 2. Is the ‘quite by design’ about right? (I worked on my own exp of 1-3 suggestions that I actually act on per month being the right cadence, does that feel right? A couple of things to preempt comments. No code leaves your machine. No API keys needed. Feel free to fully self host. Engine is Apache 2.0. Npx Replen (MCP and Skill). Happy to answer any qs on capability graph, self-tuning loop or modality gate in comments. Thx | ||