Storia AI is developing an open-source copilot designed to understand a company's codebase and its context, assisting software engineers as they focus more on reading, judging, and architecting code. The initial offering, Sage, functions like a Perplexity-style agent, aiding developers in understanding, evaluating, and generating software. Sage allows developers to query existing codebases with questions such as selecting the appropriate vector database based on project constraints, comparing technologies like Redis and Milvus, and assessing the functionality and integration needs of a codebase within an organization.
Sage provides answers supported by documentation and external references, including GitHub, Stack Overflow, technical design documents, and project management tools, minimizing the risk of hallucinations. Currently, Sage maintains up-to-date knowledge of open-source repositories, with plans to expand its understanding to every line of code available online. For team environments, Sage can also be tailored to understand private codebases.
The challenge of creating an AI system capable of comprehending codebases and empowering developers to architect superior code efficiently requires innovative research, as standard RAG and generic LLMs are insufficient.