v3.41.4 v3.42.0
ysharma3501/NovaSRSep 14, 2026 → Sep 15, 20260 days between releases
What's New
Swarm consensus now uses trust weights, vector search is faster, and security is hardened with MCP and Hive-mind.
Breaking Changes
- New `hiveToken` required for `hive-mind_join`, `hive-mind_leave`, and voting in Hive-mind MCP tools
New Features
- Trust-weighted swarm consensus
- HNSW product-quantization for vector search
- MCP governance policy enforcement via `RUFLO_MCP_ENFORCE_POLICY`
- Hive-mind capability tokens
- Real per-outcome confidence scores in learning loop
Improvements
- Vector search performance and accuracy via HNSW product-quantization
- MMR re-ranking speedup
- Swarm consensus logic now utilizes trust weights
- Learning loop uses real confidence scores
- Security hardening via MCP governance and Hive-mind Sybil protection
Bug Fixes
- Swarm consensus now correctly uses trust weights for voting
- Vector search dispatch fixed to use valid embeddings data
- Hive-mind voting hardened against Sybil attacks
- Learning loop now uses real per-outcome confidence scores
Upgrade Notes
Configure the new `RUFLO_MCP_ENFORCE_POLICY` environment variable and ensure your `hiveToken` is available for `hive-mind_join`, `hive-mind_leave`, and voting operations.
The Two Releases
v3.41.4 — agentdb_pattern-search/store degraded-state flag
Sep 14, 2026· 1 month ago
Fixes the `agentdb_pattern-store` and `agentdb_pattern-search` handlers to surface a `degraded: true` flag when the ReasoningBank controller is unhealthy or unavailable.
Full release notesv3.42.0 — Swarm, Memory & Security Reliability Batch
Sep 15, 2026· 1 month ago
A major reliability batch release featuring Swarm consensus improvements, vector search performance enhancements, and significant security hardening including MCP governance enforcement and Hive-mind protection against Sybil attacks.
Full release notes