The core goal of Quack AI is to turn a team's implicit engineering standards into a living index that can be queried, reasoned about, and measured, and to present it directly in the editor where code is written. It addresses the problem that every team has its own way of doing things, but these expectations are mostly scattered in members' heads, in review comments, or in a CONTRIBUTING.md that no one reads.
Main capabilities:
- Structured guidelines: maintainers only need to write expectations once, covering naming, error handling, testing, review etiquette, and more; Quack parses existing documentation and past PR comments to help generate an initial list of guidelines.
- In-editor contextual assistance: the extension pulls only the guidelines relevant to the currently open file, checks code diffs, and generates code that conforms to the guidelines before reviewers see it.
- Feedback loop: maintainers can see which rules are violated most and which rules no one understands, thereby discovering leaks in the contribution process and rewriting guidelines that were not implemented.
How to use it:
- Runs on open-source models and does not depend on anyone else's cloud services; Quack can connect to Ollama and any open-source model you specify.
- Supports self-hosting the entire stack, keeping every line of source code on your own infrastructure.
- Open source under the Apache 2.0 license, so you can fork it, run it, and use the parts you need.
- The project consists of three parts: the API (guideline storage, retrieval, and Ollama-based LLM orchestration), the editor extension (inline presentation of guidelines and checks), and the web app (for maintainers to curate and review guidelines).
Use cases:
- R&D teams that want to codify and automatically check their team's engineering standards.
- Scenarios that need to reduce review round trips and align with standards during the coding phase.
- Teams with data privacy requirements that want to run LLMs fully locally.
Limitations and current status:
- Quack AI is no longer maintained (the product has been sunsetted).
- The code remains open source on GitHub under Apache 2.0, and you can maintain your own fork.
- After working at SourcePulse, the founders have left the developer tools space and are currently building Rimward (safety monitoring for solar sites).
