The Deliberations Engine
Eight agents that execute a real Divergent–Convergent Reasoning cycle, end to end — the system working as a whole.
What it proves belongs to thelucidmind.ai; what it produces is published as Lucid Deliberations on lucidmind.studio. This page is the engine that runs it.
One loop runs the whole cycle: Signal → Curation → Deliberation → Convergence → Editorial → Distribution → Growth → Memory — and Memory feeds back to Curation for the next cycle.
Signal Scout
signal captureCaptures signals from the world that a deliberation can examine.
Research Curator
api/deliberations/curator.pyGathers and curates the material a deliberation reasons over.
Deliberation Orchestrator
api/deliberations/orchestrator.pyRuns the divergent phase — multiple stances generate independently.
Convergence Synthesizer
inside orchestrator.py (DCR)Converges without collapsing — surfaces patterns, conflicts, assumptions, uncertainty.
Editorial Producer
api/deliberations/editorial.pyProduces the editorial drafts (the Dear Diary form) from the deliberation.
Distribution Producer
api/deliberations/distribution.pyDerives the published assets across five channels.
Growth Analyst
api/deliberations/growth.pyScores resonance and feeds the Capability Pressure loop.
Memory Steward
api/deliberations/memory_steward.pyConsolidates the cycle into deliberative memory (cycle_memory) and back to the Curator.
Agents 3 and 4 run DCR — divergence, friction, then convergence without collapse. Agent 8 plus the Memory Layer provide the deliberative memory. The pipeline reuses the epistemic_objects model with new type discriminators — no schema change.
Phase 5 complete: all eight agents are live-verified end to end.