Send broken Python snippets to Inkeep for an automated fix.
Route coding requests through a local proxy to Claude 3.5 Sonnet.
Automatically fail over to another model when hitting rate limits.
Read the integration guide for details on the proxy's architecture.
| obrh1/freesont3.5 | 0marildo/imago | 0xdfi/glm-5.2-1m-4x-dgx-spark | |
|---|---|---|---|
| Stars | 3 | 3 | 3 |
| Language | Python | Python | Python |
| Setup difficulty | hard | easy | hard |
| Complexity | 3/5 | 2/5 | 5/5 |
| Audience | developer | general | researcher |
Figures from each repo's GitHub metadata at analysis time.
README does not document installation steps, refers readers to a separate integration guide file.
This repository contains two Python scripts, called Inkeep and Inkeep2, that act as a bridge between a developer's own tools and the Claude 3.5 Sonnet AI model, routed through a local proxy the author built. The first script, Inkeep, takes broken Python code snippets and returns a corrected version, using a model the author describes as a QA expert for fixing structural problems. The second script, Inkeep2, gives more direct access to Claude 3.5 Sonnet using what the author calls a multi turn wrapper strategy, intended to get around some of the standard limitations built into typical chat interfaces. The README describes several technical features. The proxy includes built in handling for proof of work challenges meant to avoid bot detection, and it can automatically fail over to a different model if the primary one hits a rate limit. It also describes a step that strips out certain keywords before sending requests, which the author says is meant to avoid automated content classifiers flagging the traffic. A separate integration guide file is referenced for a fuller technical explanation of how the proxy and wrapper logic work. The author states they wrote none of the code themselves and instead acted as the architect, directing AI tools to implement the entire project. This is presented as an example of what they call an AI native development approach, where a person designs the system and the AI writes all of the actual code. The project asks for donations to a cryptocurrency wallet address to support continued development. There is no license file mentioned in the README, and no information is given about installation requirements or how to run the proxy server itself beyond the referenced integration guide.
Two Python scripts that route requests through a local proxy to Claude 3.5 Sonnet, built entirely by directing AI tools rather than hand-written code.
Mainly Python. The stack also includes Python.
Setup difficulty is rated hard, with roughly 1h+ to a first successful run.
Mainly developer.
This repo across BitVibe Labs
double-check against the repo, no cap.