Pomniter

The project was initially named Breadcrumb but has been renamed to Pomniter, inspired by the Russian word помнить (pomnit) meaning "to remember." The core concept behind Pomniter is to make screenshots more understandable and searchable, so users can find them using natural language queries rather than relying on filenames or exact keywords.

The problem Pomniter aims to solve is the difficulty in finding screenshots when needed due to the lack of contextual understanding. Simply storing images does not provide enough information about their content. For example, a screenshot containing "mid = left + (right - left) / 2" should be more than just the text; it should also convey the context, such as being related to programming, DSA, binary search, interview preparation, and potentially useful for future reference.

To achieve this, Pomniter will need to go beyond traditional OCR (Optical Character Recognition) and understand the context surrounding the text in the screenshot. The system should be able to handle queries like "BFS," "shortest path," or "operating systems" and return relevant screenshots based on the user's intention rather than just exact keyword matches.

The initial version of Pomniter will focus on the core loop: saving a screenshot, understanding its content, storing the learned information, and allowing users to search using natural language. AI technologies like image understanding, OCR, embeddings, and retrieval-augmented generation (RAG) will be employed to make this process more effective.

The system will use image understanding to extract text and context, structured metadata to organize the information, and semantic search to find relevant results based on the user's natural language queries.