Recall: A Private AI Memory Companion That Runs Locally
This is a submission for the Hacktoberfest Weekend Challenge: Build for a Friend What I Built I built Recall , a local-first AI memory companion for a friend who constantly has the same problem: they remember that they saw, wrote, saved, or discussed something somewhere, but can't remember where. Recall turns their personal data — notes, documents, conversations, bookmarks, and other files — into…
This submission for the Hacktoberfest Weekend Challenge introduces Recall, a local-first AI memory companion designed for a friend who struggles to locate specific memories scattered across notes, documents, conversations, bookmarks, and other files. Recall transforms personal data into a private, searchable memory, allowing users to ask questions such as where a particular document was saved, what they said about a Goa trip, or a list of learning objectives for the year. Crucially, all personal memories remain on the user's own machine.
The core technology powering Recall is built around open-source AI and local inference. Essential components include Ollama for running open-weight language models locally, Qwen and other open-weight models for reasoning and response generation, BGE-M3 for generating local embeddings, Qdrant for vector storage and retrieval, Whisper for processing voice notes, and FastAPI for the backend. The user interface is built using Next.js.
Recall's pipeline begins with personal data ingestion, which is then broken down into chunks with accompanying metadata. Local embeddings are generated using BGE-M3, followed by vector storage and retrieval managed by Qdrant. Voice notes are processed via Whisper. The retrieved information is then analyzed by a local open-weight language model before providing an answer.
This approach treats personal data as a persistent personal memory that can be retrieved through semantic search, metadata, entities, dates, and relationships, all happening locally.
Open innovation is particularly significant in this context. By utilizing open-weight models and local inference, Recall ensures that sensitive personal data remains on the user's device, preventing it from being sent to third-party AI providers. This approach also enhances the system's flexibility, enabling users to swap models, modify retrieval pipelines, adjust memory representation, fine-tune components, or run the entire system on different hardware without the need for a complete rebuild around a closed API.
Therefore, open innovation is essential in creating a secure, customizable, and hackable AI memory companion.
Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.