Repository navigation
Use Cases
Ask your coding agent to implement account recovery from the project's specification and ADR. It searches with Ragmir, opens the cited passages, searches again for exceptions, then updates code and tests. The retrieved passages enter the agent's usual model conversation.
Ask the agent to investigate a checkout timeout using the production runbook, an incident report, and the retry implementation. It searches the symptoms and constraints, expands the evidence that explains the failure, then proposes the narrow code change and a regression test. The cited passages enter the agent's usual model conversation.
Ask the agent to plan an API v2 migration from the migration plan, versioning ADR, and affected callers. It identifies compatibility rules, rollback conditions, and test coverage before it edits code. The retrieved passages enter the agent's usual model conversation.
Ask whether internal architecture rules require attachments in PostgreSQL or object storage. The chat retrieves the supporting passages with Ragmir, then generates a cited answer using a model on your machine or on infrastructure you operate. A downloaded model served by Ollama is one local option; your own inference server is another deployment choice. Local mode can keep the exchange on the machine; self-hosted mode sends excerpts to your server. Control the client, model, tools, access, and logs within the chosen environment.
All four workflows can use spreadsheets with cited headers, scanned PDFs with local OCR, and separate monorepo bases. The consuming application owns synthesis, code changes, and validation. Ragmir supplies indexed evidence, not a guarantee that every requested fact is documented. See Agent Integration and TypeScript API.