Search Optimization & Agent Documents 🔍

This release focuses on one practical goal: making knowledge easier to find and manage. Search now returns results with better speed and relevance, and agent-level documents sit on a clearer storage foundation.

Search and storage foundation

At the search layer, we added BM25 indexing with the ICU tokenizer and enabled PostgreSQL pg_search. Together, these upgrades give full-text retrieval a stronger base for both recall and ranking quality.

For agent knowledge organization, this release introduces the new agent_documents table and adds a description field to topic metadata. This makes content structure and retrieval context more explicit.

On security hardening, we added API key hash storage to improve key handling safety.

Improvements and fixes

• Fixed changelog auto-generation in the release workflow.
• Corrected stable renderer tar source path.
• Resolved market M2M token registration for trust-client scenarios.