Changelog

Track new features, improvements, bug fixes, and breaking changes across all ChaozCode products. Subscribe to our RSS feed to stay updated.

v8.0.0 Latest

March 2026

New Features

  • Memory Spine expanded to 140,000+ vectors with FTS5 full-text search
  • ML Router upgraded to 254 specialized agents with cascade routing v14.2 — 100% accuracy on 410 test scenarios
  • Total platform tools expanded to 1,050+ across 14 MCP servers
  • Zearch Engine: 623 API endpoints now live across 26 route modules
  • ZIcon AI icon generation added — available at /auth/apps/zicon/
  • ChaozPilot v2: improved context management and multi-agent streaming — at /auth/tools/chaozpilot/
  • HelixHyper knowledge graph v7.0.0: 811+ nodes, 1,063 edges, 124 REST endpoints, CORS middleware, SQLite hydration on startup, OpenAPI/Swagger at /docs, self-healing, and full enterprise analytics (PageRank, eigenvector centrality, community detection, influence propagation)
  • GitChaozOxide: Rust-native git operations with integrated security scanning
  • 14 MCP servers active — added ionos-mcp, gitchaoz-mcp, zemail-mcp

Improvements

  • ML Router training: 4-layer cascade hybrid with 100% accuracy on 410 test scenarios (from 68 agents to 254)
  • Memory Spine v3.0.0-enterprise-fts5: vector count grown from 55K to 140,000+
  • Zearch throughput: 1,000+ req/s with <50ms p95 latency
  • Admin dashboard: 8-tab UI with live cascade/training/config views

v0.4.0

January 28, 2026

New Features

  • Added memory_consolidate tool for merging old memories based on decay threshold
  • New knowledge_graph_build and knowledge_graph_query tools for relationship mapping
  • Introduced codebase_generate tool for AI-powered code generation from context
  • Added batch operations: memory_batch_store, memory_batch_tag, memory_batch_delete
  • New agent_handoff tool for seamless context transfer between agents
  • Added memory_priority_score for intelligent memory ranking
  • New conversation management: conversation_start, conversation_add_turn, conversation_save

Improvements

  • Memory search now uses hybrid vector + keyword matching for 40% better relevance
  • Reduced average response time from 12ms to 8ms across all operations
  • Improved llm_context_window token estimation accuracy to 99%+
  • Vector capacity increased from 25K to 50K vectors per instance
  • Added automatic deduplication during memory storage

Bug Fixes

  • Fixed race condition in concurrent memory writes #234
  • Resolved memory leak in long-running webhook connections #241
  • Fixed incorrect timestamp parsing for ISO 8601 dates with timezone offsets
  • Corrected tag filtering when using special characters

v0.3.2

January 15, 2026

Security

  • Patched potential XSS vulnerability in webhook payload rendering
  • Added rate limiting for authentication endpoints (100 req/min)
  • Improved API key hashing with Argon2id

Bug Fixes

  • Fixed memory_search returning stale results after recent stores
  • Resolved WebSocket disconnection issues on high-latency networks
  • Corrected pagination cursor handling for large result sets

v0.3.1

January 8, 2026

Improvements

  • Reduced memory footprint by 30% through optimized vector storage
  • Added support for custom embedding models via configuration
  • Improved error messages with actionable suggestions

Bug Fixes

  • Fixed memory_timeline not respecting hour parameter
  • Resolved duplicate entries in memory_cluster output

v0.3.0 Breaking

December 20, 2025

Breaking Changes

  • Renamed memory_add to memory_store for consistency
  • Changed memory_search response format: results now include score field
  • Removed deprecated memory_fetch (use memory_retrieve instead)
  • API version bumped to v2 - v1 endpoints deprecated

New Features

  • Introduced codebase_analyze and codebase_context tools
  • Added memory_insights for pattern analysis across memories
  • New memory_link tool for creating memory relationships
  • Added MCP protocol support for Copilot integration

Improvements

  • Rewrote search engine with HNSW indexing for 10x faster queries
  • Added automatic retry logic for transient failures
  • Improved documentation with interactive examples

v0.2.5

December 1, 2025

New Features

  • Added memory_pin and memory_unpin for persistent context
  • New memory_export tool for bulk data extraction
  • Introduced webhook support for real-time event notifications

Improvements

  • Added support for filtering by multiple tags (AND/OR logic)
  • Improved memory decay algorithm for more natural forgetting
  • Added created_at and updated_at timestamps to all memories

v0.2.0

November 15, 2025

New Features

  • Initial release of Memory Spine persistent memory system
  • Core operations: memory_store, memory_search, memory_retrieve, memory_delete
  • Vector-based semantic search with cosine similarity
  • Tag-based organization and filtering
  • REST API with JSON request/response format
  • Authentication via API keys