RelayCore v1.2.0

RelayCore v1.2.0

RelayCore v1.2.0 turns the shared intelligence line into a more usable runtime workflow: startup context loading is automatic, durable memory capture can be assisted instead of hand-wired every time, and Mission Control now exposes confidence-aware memory review.

Highlights

• auto-memory MCP helpers: memory_auto_prepare, memory_auto_extract, memory_auto_capture, memory_auto_commit
• AI confidence scoring for extracted memory drafts, with manual override support in Mission Control
• versioned RULES.md plus relaycore sync-rules projection into RelayCore rule memory
• refined Memory Viewer that emphasizes readable memory, tag-first filtering, and legacy memory visibility
• clearer single-database and multi-runtime guidance for real Codex / Claude style deployments

What Changed In This Release

• expanded the MCP surface so runtimes can bootstrap task context and capture durable observations without replaying full chat history
• added auto-memory extraction heuristics, duplicate-aware proposal routing, and split note capture for batched observations
• introduced optional OpenAI-backed confidence assessments for auto-memory candidates and manual confidence review controls in the web UI
• added repository rule sync tooling so human-reviewed RULES.md documents can be projected into active RelayCore rule memories
• refreshed runtime docs, startup guidance, and deployment examples around the canonical ~/.relaycore/relaycore.db topology
• improved Mission Control memory browsing with better grouping, provenance drill-down, and imported memory visibility

Included CLI

• relaycore init-db
• relaycore serve
• relaycore export
• relaycore mcp-http
• relaycore consolidate-db
• relaycore sync-rules --rules-file RULES.md
• python -m relaycore init-db
• python -m relaycore serve
• python -m relaycore export
• python -m relaycore mcp-http

Runtime Notes

• Core service: Python 3.9+
• MCP bridge: Python 3.10+ with pip install -e .[mcp]
• AI confidence scoring is optional and activates when OPENAI_API_KEY is configured

Validation

• local automated tests passed
• pytest status on July 29, 2026: 81 passed

Packaging

• source distribution
• universal wheel for the current Python package