The first working foundation
The earliest Chronicler build joined document ingestion, authentication, retrieval, and a desktop interface into one testable system. Rough, but end to end.
Chronicler / historyOur history
Built from one stubborn question
What if asking your own documents felt as natural as asking a colleague, without giving up the source, the privacy, or the control? Chronicler grew from that question into a document intelligence workspace that ships in three deployment shapes.
Retrieval that could be checked
Citations and document attribution became part of the answer itself rather than a debugging view, and OCR brought scanned pages into the same pipeline as native text.
From search to a knowledge workspace
Collections, conversational context, and multi-format parsing turned isolated answers into a repeatable research workflow.
A swappable pipeline
Embeddings, vector store, reranker, and language model each became replaceable. Run against OpenAI, or keep the whole thing local with Ollama, without changing how the product behaves.
Ready for teams
Workspaces, roles, an administration console, self-hosting, updates, and licensing brought Chronicler beyond a personal prototype. GDPR erasure and an audit log were built in rather than promised.
Indexing moved off member devices
The v0.2.0-beta release moved embedding to a hosted service, so the desktop app stopped downloading a 390MB model and uploads finish indexing whether or not the app stays open. Self-hosted admins choose where embedding happens: the server's own model, member devices, Cohere, or OpenAI.
Toward the first public release
Three deployment options, one product. We are refining reliability, privacy controls, and the experience around one goal: knowledge people can actually verify.
“The product has changed quickly. The principle has not: the evidence should always travel with the answer.”