v1.4.1 — Reliability & Search Improvements

Release Date: 2026-07-31
Changes: v1.4.1 → main

Summary

This patch release focuses on reliability, search relevance, and stability. It fixes several user-visible bugs, improves search results and ingestion robustness, and reduces memory and latency in common workloads.

Highlights

• Improved search relevance so queries return more useful results.
• More robust ingestion: fewer failed uploads and clearer error messages.
• Stability fixes that prevent rare crashes and reduce memory usage.

Breaking Changes

• None. This is a patch release with bug fixes and small improvements; no breaking changes are introduced.

New Features

• Search result previews: When you open a dataset (a collection of documents you've added), search results now include short preview snippets so you can quickly judge relevance without opening the full document. This helps you find the right information faster.
• Clearer ingestion error messages: If adding files or text to a dataset fails, the system now returns clearer, actionable messages explaining what went wrong and how to fix it (e.g., unsupported file type or quota limit). This reduces time spent diagnosing upload issues.

Improvements

• Search relevance tuning: Ranking adjustments improve which documents appear first for typical queries, so answers and search hits are more likely to match intent.
• More robust document ingestion: Retries and better handling of transient failures reduce the chances of partial uploads, leading to fewer missing documents after bulk imports.
• Cleaner metadata handling: Deleting a dataset now reliably removes associated metadata so deleted datasets no longer leave behind orphaned records.
• Better error surfaces: User-facing messages across the UI and API are clearer and include suggested next steps (for example, retrying an upload or checking file size).

Performance

• Lower memory usage during large-file ingestion: Ingestion of big files now uses streaming where possible, so importing large documents consumes less memory.
• Faster query latency for common searches: Optimizations to the search path reduce average response times for typical queries, making interactive workflows feel snappier.

Security

• Safer handling of uploaded content: Improved validation of uploaded files to reduce the risk of malformed inputs causing instability or unexpected behavior.

Bug Fixes

• Fixed a rare crash that could occur when processing certain malformed documents during ingestion.
• Fixed an issue where deleting a dataset left behind some metadata entries.
• Fixed incorrect or confusing error messages returned by the ingestion API when uploads failed.
• Fixed a case where empty files could cause uploads to fail without a useful error message.

Technical Changes

• Internal refactors to the ingestion pipeline to enable streaming and retries, improving maintainability and reliability.
• Improved logging and diagnostics to make it easier for support and engineers to trace uploads and search queries.
• Additional tests covering previously flaky scenarios around file uploads and dataset deletion.

Compatibility

— The Cognee Team · 2026-07-31