Release Notes - v1.0.1.dev3

Release Date: 2026-04-16
Changes: v1.0.1.dev3 → dev

Summary

v1.0.1.dev3 is a small but focused development release for Cognee that improves memory ingestion, search accuracy, and overall stability. It introduces a bulk import flow, UI refinements, and several reliability fixes to make everyday workflows smoother for power users and teams.

Highlights

• Bulk CSV/JSON memory import to add many memories at once
• Improved search relevance and faster query responses
• UI polish for the Memories list (better metadata and actions)
• Fixes for sync, token refresh, and memory deletion reliability

Breaking Changes

• Memory import/export JSON schema change: The export format has a small schema change — the memory id and source fields are now required and are expected to be strings. If you rely on automated import scripts, please update them to include 'id' as a string and include a 'source' field per memory. Existing imports from earlier v1.x should be compatible after converting ids to strings.
• API: The /v1/memories create endpoint now validates and requires the 'source' field for imported records. Clients that create memories without 'source' will receive a validation error until they include it.

New Features

• Bulk import: A new bulk import option (CSV/JSON) lets you create or update large sets of memories in one operation. This speeds onboarding and migrations and reduces manual work.
• Import preview: Before importing, a preview step shows how rows will map to memory fields (title, text, tags, source) so you can adjust column mappings.
• Memory card enhancements: Memory items in the web UI now show source, tag badges, and last-updated time for faster context at a glance.

Improvements

• Search relevance: Ranking improvements produce more relevant results for short and long queries, improving hit quality when you look up past items.
• Performance: Faster memory ingestion paths and query caching reduce latency on common operations, especially for accounts with many memories.
• Deduplication: Improved duplicate detection during imports and manual creation reduces clutter and accidental repeats.
• UX polish: Cleaner list controls, clearer import progress indicators, and improved error messages to help users recover from problems faster.
• CLI/Dev experience: The developer CLI gives clearer feedback and progress reporting for long-running import or sync tasks.

Bug Fixes

• Fixed a crash that could occur when saving very large embedding vectors during ingestion.
• Resolved an issue where deleting a memory from the UI would sometimes not remove it from search results until a full reindex.
• Fixed intermittent token-refresh failures that caused background sync to stop and required manual sign-in.
• Addressed several small UI bugs: tag overflow in the memory list, incorrect timestamps in certain timezones, and misaligned buttons on small screens.

Technical Changes

• Refactored ingestion pipeline for better observability and retries; improved logging for failed imports to speed troubleshooting.
• Search subsystem tweaks to support the new ranking rules and caching layer.
• Internal dependency updates and small refactors to reduce memory usage during large imports.

Generated by Cognee Release Notes Generator on 2026-04-16