Changelog
New updates and improvements to Hippocortex. Follow along as we ship.
Plugin now auto-updates itself when a newer version is detected and prompts the user to restart the gateway.
New admin API for injecting memories with embeddings, plus system_update memories that surface until the user updates.
Fixed retrieval queries using only last user message. Now uses last 4 turns for better semantic context.
Increased default working memory and retrieval depth to take advantage of newer models with 1M token context windows.
Fixed a retrieval blindspot where short user replies like "Ja", "Ok", or "Yes" produced empty or irrelevant search results. The retrieval query now uses the last 4 conversation turns for semantic context.
Major release: the Context Engine eliminates session overflow by storing older messages in Hippocortex and retrieving them semantically. Plus a 1-click setup wizard and automatic Vault integration.
A global shared knowledge layer across all agents with a 6-stage promotion pipeline and SVG visualization in the admin dashboard.
Full-featured support hub with 21 API endpoints, 11 public pages, AI chat with RAG, and comprehensive knowledge base.
Six previously standalone packages wired into the production cloud-api with an adapter layer, 8 feature flags, and 21 new integration tests.
Major retrieval quality improvements with hybrid vector+keyword search, type boosting, and smarter budget enforcement.
Fixed execute signature alignment, agent_end hook lifecycle, and baseUrl configuration propagation.
Added vault query and reveal tools so agents can search and retrieve secrets from the Hippocortex Vault directly. Main Vault is auto-provisioned on first use.
Major upgrade to memory retrieval quality. Memories are now embedded with OpenAI text-embedding-3-small and searched via cosine similarity. Multiple new quality filters reduce noise and improve relevance.
First stable npm publish of the plugin with content security policy fixes for secure deployment.
First public release of the Hippocortex OpenClaw plugin. Provides capture, synthesize, and learn tools for any OpenClaw-compatible agent.
Added vault methods for querying and revealing secrets stored in the Hippocortex Vault.
First public release of the Hippocortex JavaScript/TypeScript SDK. Covers all core API endpoints with full type safety.
Added vault query and reveal methods to both async and sync clients, plus new dataclasses for vault responses.
Added first-party adapters for popular AI agent frameworks, making integration a one-liner.
First public release of the Hippocortex Python SDK on PyPI. Async-first with a sync wrapper.
Comprehensive backend overhaul across 13 phases: vector indexing, hardened learn endpoint, memory distillation, intelligent lifecycle, predictive caching, and more.
Introduced the Vault subsystem for encrypted secret storage, with API endpoints and MCP tool integrations.
First production deployment of the Hippocortex Cloud API with full enterprise feature set.