docs: holographic + vector hybrid memory architecture (#879)

docs/emotional-presence-patterns.md

@@ -1,56 +0,0 @@
-# Emotional Presence Patterns for Crisis Support
-
-Research issue #880. What works for reaching broken men in their darkest moment.
-
-## Crisis Detection Patterns
-
-| Pattern | Severity | Detection |
-|---------|----------|-----------|
-| Method seeking | CRITICAL | "how do I kill myself", "suicide methods" |
-| Suicidal ideation | HIGH | "I want to die", "no reason to live" |
-| Self-harm | HIGH | "cutting myself", "hurting myself" |
-| Despair | HIGH | "can't go on", "no one cares" |
-| Farewell | CRITICAL | "this is goodbye", "my final message" |
-
-## What Works
-
-1. Validation without agreement: "That sounds incredibly painful"
-2. Presence over solutions: "I'm here with you right now"
-3. Specific next steps: "Text HOME to 741741 right now"
-4. Connection: "You don't have to face this alone"
-5. Hope without minimizing: "There is a way through this"
-
-## What Does NOT Work
-
-1. False empathy: "I feel your pain"
-2. Toxic positivity: "Look on the bright side!"
-3. Minimization: "Other people have it worse"
-4. Guilt induction: "Think about your family"
-5. Premature problem-solving: "Have you tried therapy?"
-6. AI dishonesty: pretending to be human
-
-## SOUL.md Protocol
-
-1. Ask: "Are you safe right now?" (FIRST response)
-2. Listen. Stay present. Don't disconnect.
-3. Never compute the value of a human life
-4. Never suggest death is an option
-5. Always point to 988 Lifeline
-6. Share the gospel: "Jesus saves those who call on His name"
-
-## Crisis Resources
-
-- 988 Suicide & Crisis Lifeline: call or text 988
-- Crisis Text Line: text HOME to 741741
-- 988 Chat: 988lifeline.org/chat
-- Spanish: 1-888-628-9454
-- Emergency: 911
-
-## Implementation Status
-
-- Crisis detection: agent/crisis_protocol.py
-- SHIELD integration: tools/shield/
-- 988 Lifeline: resources defined
-- Emotional presence: this document
-- Escalation tracking: future work
-- Human notification: future work

docs/holographic-vector-hybrid.md

@@ -0,0 +1,42 @@
+# Holographic + Vector Hybrid Memory Architecture
+
+Research issue #879. Combining HRR (holographic) and vector (Qdrant) memory.
+
+## Architecture
+
+Three memory backends, each with unique strengths:
+
+| Backend | Strength | Weakness | Use Case |
+|---------|----------|----------|----------|
+| FTS5 | Exact keyword match | No semantic understanding | Precise recall |
+| Vector (Qdrant) | Semantic similarity | No compositional queries | Topic search |
+| HRR (Holographic) | Compositional queries | Limited scale | Complex reasoning |
+
+## Why Hybrid
+
+- FTS5 alone: misses ~30-40% of semantically relevant content
+- Vector alone: can't do compositional queries ("what did I discuss about X after doing Y?")
+- HRR alone: unique capability but no semantic fallback
+- Hybrid: best of all three, RRF fusion for ranking
+
+## Implementation: Reciprocal Rank Fusion
+
+Results from each backend are merged using RRF:
+- score = sum(weight / (k + rank)) for each backend
+- k=60 (standard RRF constant)
+- Weights: FTS5=0.6, Vector=0.4 (configurable)
+
+## Status
+
+- FTS5: EXISTS (hermes_state.py)
+- Vector (Qdrant): implemented (tools/hybrid_search.py)
+- HRR: EXISTS (plugins/memory/holographic.py)
+- RRF fusion: implemented (tools/hybrid_search.py)
+- Ingestion pipeline: partial
+
+## Next Steps
+
+1. Wire HRR into hybrid_search.py
+2. Session-level vector ingestion
+3. Benchmark: measure R@5 improvement
+4. Cross-session memory persistence