V0.7 Governed Memory & Knowledge Graph
Lucid begins connecting intelligence across entities, evidence and time, introducing Atlas alongside more durable and inspectable research context.
- Type
- Changelog
- Published
- 2026-07-27
- Last updated
- 2026-08-23
- Reading time
- 1 min
Features ✅
- Atlas – Explore connections between people, organisations, places, publishers, events and topics through an interactive intelligence graph.
- Durable Signals conversations – Conversations used to create and refine Signals now persist alongside their research and evidence.
- Governed memory – Lucid can retain useful research context with provenance and controls around how that information is reused.
- Human review for agent workflows – Longer-running workflows can pause for review before continuing with actions that require approval.
Improvements 🧱
- Evidence-first relationships – Intelligence graphs increasingly use articles as evidence for relationships rather than overwhelming the graph with individual stories.
- Inspectable memory – Added clearer provenance and review controls around retained context.
- More resilient graph exploration – Improved caching, freshness and performance for larger intelligence views.
- Better research continuity – Strengthened the relationship between a user request, the research it produces and the resulting Signal or intelligence output.
Fixes 🔧
- Fixed missing entities in public and authenticated intelligence graph views.
- Fixed inconsistent authentication behaviour across graph experiences.
- Improved recovery for Signal runs interrupted before reaching a final state.
- Fixed several graph completeness, loading and navigation issues.