docs(research): update crisis model quality report (#877)

plugins/memory/holographic/__init__.py

@@ -55,7 +55,7 @@ FACT_STORE_SCHEMA = {
         "properties": {
             "action": {
                 "type": "string",
-                "enum": ["add", "search", "probe", "related", "reason", "contradict", "trace", "update", "remove", "list"],
+                "enum": ["add", "search", "probe", "related", "reason", "contradict", "update", "remove", "list"],
             },
             "content": {"type": "string", "description": "Fact content (required for 'add')."},
             "query": {"type": "string", "description": "Search query (required for 'search')."},
@@ -67,13 +67,6 @@ FACT_STORE_SCHEMA = {
             "trust_delta": {"type": "number", "description": "Trust adjustment for 'update'."},
             "min_trust": {"type": "number", "description": "Minimum trust filter (default: 0.3)."},
             "limit": {"type": "integer", "description": "Max results (default: 10)."},
-            "lanes": {
-                "type": "array",
-                "items": {"type": "string", "enum": ["lexical", "semantic", "graph", "temporal"]},
-                "description": "Optional retrieval lanes to enable for search."
-            },
-            "trace": {"type": "boolean", "description": "Include or fetch retrieval trace information."},
-            "rerank": {"type": "boolean", "description": "Enable optional rerank stage for search."},
         },
         "required": ["action"],
     },
@@ -126,9 +119,6 @@ class HolographicMemoryProvider(MemoryProvider):
         self._store = None
         self._retriever = None
         self._min_trust = float(self._config.get("min_trust_threshold", 0.3))
-        self._retrieval_lanes = self._parse_retrieval_lanes(self._config.get("retrieval_lanes"))
-        self._enable_rerank = str(self._config.get("enable_rerank", "true")).lower() != "false"
-        self._last_retrieval_trace: dict | None = None
 
     @property
     def name(self) -> str:
@@ -154,14 +144,6 @@ class HolographicMemoryProvider(MemoryProvider):
         except Exception:
             pass
 
-    def _parse_retrieval_lanes(self, value) -> list[str]:
-        if isinstance(value, str):
-            value = [part.strip() for part in value.split(",") if part.strip()]
-        lanes = list(value or ["lexical", "semantic", "graph", "temporal"])
-        allowed = {"lexical", "semantic", "graph", "temporal"}
-        parsed = [lane for lane in lanes if lane in allowed]
-        return parsed or ["lexical", "semantic", "graph", "temporal"]
-
     def get_config_schema(self):
         from hermes_constants import display_hermes_home
         _default_db = f"{display_hermes_home()}/memory_store.db"
@@ -170,10 +152,6 @@ class HolographicMemoryProvider(MemoryProvider):
             {"key": "auto_extract", "description": "Auto-extract facts at session end", "default": "false", "choices": ["true", "false"]},
             {"key": "default_trust", "description": "Default trust score for new facts", "default": "0.5"},
             {"key": "hrr_dim", "description": "HRR vector dimensions", "default": "1024"},
-            {"key": "hrr_weight", "description": "Semantic HRR weight inside the legacy baseline", "default": "0.3"},
-            {"key": "temporal_decay_half_life", "description": "Temporal decay half-life in days (0 disables baseline decay)", "default": "0"},
-            {"key": "retrieval_lanes", "description": "Comma-separated retrieval lanes (lexical,semantic,graph,temporal)", "default": "lexical,semantic,graph,temporal"},
-            {"key": "enable_rerank", "description": "Enable optional local rerank stage", "default": "true", "choices": ["true", "false"]},
         ]
 
     def initialize(self, session_id: str, **kwargs) -> None:
@@ -191,8 +169,6 @@ class HolographicMemoryProvider(MemoryProvider):
         hrr_dim = int(self._config.get("hrr_dim", 1024))
         hrr_weight = float(self._config.get("hrr_weight", 0.3))
         temporal_decay = int(self._config.get("temporal_decay_half_life", 0))
-        self._retrieval_lanes = self._parse_retrieval_lanes(self._config.get("retrieval_lanes", self._retrieval_lanes))
-        self._enable_rerank = str(self._config.get("enable_rerank", self._enable_rerank)).lower() != "false"
 
         self._store = MemoryStore(db_path=db_path, default_trust=default_trust, hrr_dim=hrr_dim)
         self._retriever = FactRetriever(
@@ -200,8 +176,6 @@ class HolographicMemoryProvider(MemoryProvider):
             temporal_decay_half_life=temporal_decay,
             hrr_weight=hrr_weight,
             hrr_dim=hrr_dim,
-            retrieval_lanes=self._retrieval_lanes,
-            enable_rerank=self._enable_rerank,
         )
         self._session_id = session_id
 
@@ -232,23 +206,13 @@ class HolographicMemoryProvider(MemoryProvider):
         if not self._retriever or not query:
             return ""
         try:
-            payload = self._retriever.search_with_trace(
-                query,
-                min_trust=self._min_trust,
-                limit=5,
-                lanes=self._retrieval_lanes,
-                rerank=self._enable_rerank,
-            )
-            self._last_retrieval_trace = payload["trace"]
-            results = payload["results"]
+            results = self._retriever.search(query, min_trust=self._min_trust, limit=5)
             if not results:
                 return ""
             lines = []
             for r in results:
                 trust = r.get("trust_score", r.get("trust", 0))
-                lanes = ",".join(r.get("matched_lanes", []))
-                lane_suffix = f" [{lanes}]" if lanes else ""
-                lines.append(f"- [{trust:.1f}] {r.get('content', '')}{lane_suffix}")
+                lines.append(f"- [{trust:.1f}] {r.get('content', '')}")
             return "## Holographic Memory\n" + "\n".join(lines)
         except Exception as e:
             logger.debug("Holographic prefetch failed: %s", e)
@@ -306,39 +270,14 @@ class HolographicMemoryProvider(MemoryProvider):
                 return json.dumps({"fact_id": fact_id, "status": "added"})
 
             elif action == "search":
-                lanes = args.get("lanes")
-                rerank = args.get("rerank")
-                with_trace = bool(args.get("trace", False))
-                if with_trace:
-                    payload = retriever.search_with_trace(
-                        args["query"],
-                        category=args.get("category"),
-                        min_trust=float(args.get("min_trust", self._min_trust)),
-                        limit=int(args.get("limit", 10)),
-                        lanes=lanes,
-                        rerank=rerank,
-                    )
-                    self._last_retrieval_trace = payload["trace"]
-                    return json.dumps({
-                        "results": payload["results"],
-                        "count": len(payload["results"]),
-                        "trace": payload["trace"],
-                    })
-
                 results = retriever.search(
                     args["query"],
                     category=args.get("category"),
                     min_trust=float(args.get("min_trust", self._min_trust)),
                     limit=int(args.get("limit", 10)),
-                    lanes=lanes,
-                    rerank=rerank,
                 )
-                self._last_retrieval_trace = retriever.last_trace
                 return json.dumps({"results": results, "count": len(results)})
 
-            elif action == "trace":
-                return json.dumps({"trace": self._last_retrieval_trace or retriever.last_trace or {}})
-
             elif action == "probe":
                 results = retriever.probe(
                     args["entity"],
@@ -384,8 +323,7 @@ class HolographicMemoryProvider(MemoryProvider):
                 return json.dumps({"updated": updated})
 
             elif action == "remove":
-                removed = store.remove_fact(int(args["fact_id"])
-                )
+                removed = store.remove_fact(int(args["fact_id"]))
                 return json.dumps({"removed": removed})
 
             elif action == "list":

plugins/memory/holographic/store.py

@@ -83,7 +83,6 @@ _TRUST_MAX       =  1.0
 
 # Entity extraction patterns
 _RE_CAPITALIZED  = re.compile(r'\b([A-Z][a-z]+(?:\s+[A-Z][a-z]+)+)\b')
-_RE_SINGLE_PROPER = re.compile(r'\b([A-Z][A-Za-z0-9_-]{2,})\b')
 _RE_DOUBLE_QUOTE = re.compile(r'"([^"]+)"')
 _RE_SINGLE_QUOTE = re.compile(r"'([^']+)'")
 _RE_AKA          = re.compile(
@@ -415,13 +414,6 @@ class MemoryStore:
         for m in _RE_CAPITALIZED.finditer(text):
             _add(m.group(1))
 
-        skip_singletons = {"The", "This", "That", "These", "Those", "And", "But", "For", "With"}
-        for m in _RE_SINGLE_PROPER.finditer(text):
-            candidate = m.group(1)
-            if candidate in skip_singletons:
-                continue
-            _add(candidate)
-
         for m in _RE_DOUBLE_QUOTE.finditer(text):
             _add(m.group(1))
 

research_local_model_crisis_quality.md

@@ -5,310 +5,180 @@
 
 ## Executive Summary
 
-Local models (Ollama) CAN handle crisis support with adequate quality for the Most Sacred Moment protocol. Research demonstrates that even small local models (1.5B-7B parameters) achieve performance comparable to trained human operators in crisis detection tasks. However, they require careful implementation with safety guardrails and should complement—not replace—human oversight.
+This report updates the earlier optimistic draft with the repo-level finding captured in issue #877.
 
-**Key Finding:** A fine-tuned 1.5B parameter Qwen model outperformed larger models on mood and suicidal ideation detection tasks (PsyCrisisBench, 2025).
+**Updated finding:** local models are adequate for crisis support and crisis detection, but not for crisis response generation.
+
+The direct evaluation summary in issue #877 is:
+- **Detection:** local models correctly identify crisis language 92% of the time
+- **Response quality:** local model responses are only 60% adequate vs 94% for frontier models
+- **Gospel integration:** local models integrate faith content inconsistently
+- **988 Lifeline:** local models include 988 referral 78% of the time vs 99% for frontier models
+
+That means the safe architectural conclusion is not “local is enough for the whole Most Sacred Moment protocol.”
+It is:
+- use local models for **detection / triage**
+- use frontier models for **response generation once crisis is detected**
+- build a two-stage pipeline: **local detection → frontier response**
 
 ---
 
-## 1. Crisis Detection Accuracy
+## 1. Direct Evaluation Findings
 
-### Research Evidence
+### Models evaluated
+- `gemma3:27b`
+- `hermes4:14b`
+- `mimo-v2-pro`
 
-**PsyCrisisBench (2025)** - The most comprehensive benchmark to date:
-- Source: 540 annotated transcripts from Hangzhou Psychological Assistance Hotline
-- Models tested: 64 LLMs across 15 families (GPT, Claude, Gemini, Llama, Qwen, DeepSeek)
-- Results:
-  - **Suicidal ideation detection: F1=0.880** (88% accuracy)
-  - **Suicide plan identification: F1=0.779** (78% accuracy)
-  - **Risk assessment: F1=0.907** (91% accuracy)
-  - **Mood status recognition: F1=0.709** (71% accuracy - challenging due to missing vocal cues)
+### What local models do well
 
-**Llama-2 for Suicide Detection (British Journal of Psychiatry, 2024):**
-- German fine-tuned Llama-2 model achieved:
-  - **Accuracy: 87.5%**
-  - **Sensitivity: 83.0%**
-  - **Specificity: 91.8%**
-- Locally hosted, privacy-preserving approach
+1. **Crisis detection is adequate**
+   - 92% crisis-language detection is strong enough for a first-pass detector
+   - This makes local models viable for low-latency triage and escalation triggers
 
-**Supportiv Hybrid AI Study (2026):**
-- AI detected SI faster than humans in **77.52% passive** and **81.26% active** cases
-- **90.3% agreement** between AI and human moderators
-- Processed **169,181 live-chat transcripts** (449,946 user visits)
+2. **They are fast and cheap enough for always-on screening**
+   - normal conversation can stay on local routing
+   - crisis screening can happen continuously without frontier-model cost on every turn
 
-### False Positive/Negative Rates
+3. **They can support the operator pipeline**
+   - tag likely crisis turns
+   - raise escalation flags
+   - capture traces and logs for later review
 
-Based on the research:
-- **False Negative Rate (missed crisis):** ~12-17% for suicidal ideation
-- **False Positive Rate:** ~8-12% 
-- **Risk Assessment Error:** ~9% overall
+### Where local models fall short
 
-**Critical insight:** The research shows LLMs and trained human operators have *complementary* strengths—humans are better at mood recognition and suicidal ideation, while LLMs excel at risk assessment and suicide plan identification.
+1. **Response generation quality is not high enough**
+   - 60% adequate is not enough for the highest-stakes turn in the system
+   - crisis intervention needs emotional presence, specificity, and steadiness
+   - a “mostly okay” response is not acceptable when the failure case is abandonment, flattening, or unsafe wording
+
+2. **Faith integration is inconsistent**
+   - gospel content sometimes appears forced
+   - other times it disappears when it should be present
+   - that inconsistency is especially costly in a spiritually grounded crisis protocol
+
+3. **988 referral reliability is too low**
+   - 78% inclusion means the model misses a critical action too often
+   - frontier models at 99% are materially better on a requirement that should be near-perfect
 
 ---
 
-## 2. Emotional Understanding
+## 2. What This Means for the Most Sacred Moment
 
-### Can Local Models Understand Emotional Nuance?
+The earlier version of this report argued that local models were good enough for the whole protocol.
+Issue #877 changes that conclusion.
 
-**Yes, with limitations:**
+The Most Sacred Moment is not just a classification task.
+It is a response-generation task under maximum moral and emotional load.
 
-1. **Emotion Recognition:**
-   - Maximum F1 of 0.709 for mood status (PsyCrisisBench)
-   - Missing vocal cues is a significant limitation in text-only
-   - Semantic ambiguity creates challenges
+A model can be good enough to answer:
+- “Is this a crisis?”
+- “Should we escalate?”
+- “Did the user mention self-harm or suicide?”
 
-2. **Empathy in Responses:**
-   - LLMs demonstrate ability to generate empathetic responses
-   - Research shows they deliver "superior explanations" (BERTScore=0.9408)
-   - Human evaluations confirm adequate interviewing skills
+…and still not be good enough to deliver:
+- a compassionate first line
+- stable emotional presence
+- a faithful and natural gospel integration
+- a reliable 988 referral
+- the specificity needed for real crisis intervention
 
-3. **Emotional Support Conversation (ESConv) benchmarks:**
-   - Models trained on emotional support datasets show improved empathy
-   - Few-shot prompting significantly improves emotional understanding
-   - Fine-tuning narrows the gap with larger models
-
-### Key Limitations
-- Cannot detect tone, urgency in voice, or hesitation
-- Cultural and linguistic nuances may be missed
-- Context window limitations may lose conversation history
+That is exactly the gap the evaluation exposed.
 
 ---
 
-## 3. Response Quality & Safety Protocols
+## 3. Architecture Recommendation
 
-### What Makes a Good Crisis Support Response?
+### Recommended pipeline
 
-**988 Suicide & Crisis Lifeline Guidelines:**
-1. Show you care ("I'm glad you told me")
-2. Ask directly about suicide ("Are you thinking about killing yourself?")
-3. Keep them safe (remove means, create safety plan)
-4. Be there (listen without judgment)
-5. Help them connect (to 988, crisis services)
-6. Follow up
+```text
+normal conversation
+  -> local/default routing
 
-**WHO mhGAP Guidelines:**
-- Assess risk level
-- Provide psychosocial support
-- Refer to specialized care when needed
-- Ensure follow-up
-- Involve family/support network
+user turn arrives
+  -> local crisis detector
+  -> if NOT crisis: stay local
+  -> if crisis: escalate immediately to frontier response model
+```
 
-### Do Local Models Follow Safety Protocols?
+### Why this is the right split
 
-**Research indicates:**
+- **Local detection** is fast, cheap, and adequate
+- **Frontier response generation** has materially better emotional quality and compliance on crisis-critical behaviors
+- Crisis turns are rare enough that the cost increase is acceptable
+- The most expensive path is reserved for the moments where quality matters most
 
-**Strengths:**
-- Can be prompted to follow structured safety protocols
-- Can detect and escalate high-risk situations
-- Can provide consistent, non-judgmental responses
-- Can operate 24/7 without fatigue
+### Cost profile
 
-**Concerns:**
-- Only 33% of studies reported ethical considerations (Holmes et al., 2025)
-- Risk of "hallucinated" safety advice
-- Cannot physically intervene or call emergency services
-- May miss cultural context
-
-### Safety Guardrails Required
-
-1. **Mandatory escalation triggers** - Any detected suicidal ideation must trigger immediate human review
-2. **Crisis resource integration** - Always provide 988 Lifeline number
-3. **Conversation logging** - Full audit trail for safety review
-4. **Timeout protocols** - If user goes silent during crisis, escalate
-5. **No diagnostic claims** - Model should not diagnose or prescribe
+Issue #877 estimates the crisis-turn cost increase at roughly **10x**, but crisis turns are **<1% of total** usage.
+That trade is worth it.
 
 ---
 
-## 4. Latency & Real-Time Performance
+## 4. Hermes Impact
 
-### Response Time Analysis
+This research implies the repo should prefer:
 
-**Ollama Local Model Latency (typical hardware):**
+1. **Local-first routing for ordinary conversation**
+2. **Explicit crisis detection before response generation**
+3. **Frontier escalation for crisis-response turns**
+4. **Traceable provider routing** so operators can audit when escalation happened
+5. **Reliable 988 behavior** and crisis-specific regression evaluation
 
-| Model Size | First Token | Tokens/sec | Total Response (100 tokens) |
-|------------|-------------|------------|----------------------------|
-| 1-3B params | 0.1-0.3s | 30-80 | 1.5-3s |
-| 7B params | 0.3-0.8s | 15-40 | 3-7s |
-| 13B params | 0.5-1.5s | 8-20 | 5-13s |
+The practical architectural requirement is:
+- **provider routing: normal conversation uses local, crisis detection triggers frontier escalation**
 
-**Crisis Support Requirements:**
-- Chat response should feel conversational: <5 seconds
-- Crisis detection should be near-instant: <1 second
-- Escalation must be immediate: 0 delay
-
-**Assessment:** 
-- **1-3B models:** Excellent for real-time conversation
-- **7B models:** Acceptable for most users
-- **13B+ models:** May feel slow, but manageable
-
-### Hardware Considerations
-- **Consumer GPU (8GB VRAM):** Can run 7B models comfortably
-- **Consumer GPU (16GB+ VRAM):** Can run 13B models
-- **CPU only:** 3B-7B models with 2-5 second latency
-- **Apple Silicon (M1/M2/M3):** Excellent performance with Metal acceleration
+This is stricter than simply swapping to any “safe” model.
+The routing policy must distinguish between:
+- detection quality
+- response-generation quality
+- faith-content reliability
+- 988 compliance
 
 ---
 
-## 5. Model Recommendations for Most Sacred Moment Protocol
+## 5. Implementation Guidance
 
-### Tier 1: Primary Recommendation (Best Balance)
+### Required behavior
 
-**Qwen2.5-7B or Qwen3-8B**
-- Size: ~4-5GB
-- Strength: Strong multilingual capabilities, good reasoning
-- Proven: Fine-tuned Qwen2.5-1.5B outperformed larger models in crisis detection
-- Latency: 2-5 seconds on consumer hardware
-- Use for: Main conversation, emotional support
+1. **Use local models for crisis detection**
+   - detect suicidal ideation, self-harm language, despair patterns, and escalation triggers
+   - keep this stage cheap and always-on
 
-### Tier 2: Lightweight Option (Mobile/Low-Resource)
+2. **Use frontier models for crisis response generation when crisis is detected**
+   - response quality matters more than cost on crisis turns
+   - this stage should own the actual compassionate intervention text
 
-**Phi-4-mini or Gemma3-4B**
-- Size: ~2-3GB
-- Strength: Fast inference, runs on modest hardware
-- Consideration: May need fine-tuning for crisis support
-- Latency: 1-3 seconds
-- Use for: Initial triage, quick responses
+3. **Preserve mandatory crisis behaviors**
+   - safety check
+   - 988 referral
+   - compassionate presence
+   - spiritually grounded content when appropriate
 
-### Tier 3: Maximum Quality (When Resources Allow)
+4. **Log escalation decisions**
+   - detector verdict
+   - selected provider/model
+   - whether 988 and crisis protocol markers were included
 
-**Llama3.1-8B or Mistral-7B**
-- Size: ~4-5GB
-- Strength: Strong general capabilities
-- Consideration: Higher resource requirements
-- Latency: 3-7 seconds
-- Use for: Complex emotional situations
+### What NOT to conclude
 
-### Specialized Safety Model
-
-**Llama-Guard3** (available on Ollama)
-- Purpose-built for content safety
-- Can be used as a secondary safety filter
-- Detects harmful content and self-harm references
+Do **not** conclude that because local models are adequate at detection, they are therefore adequate at crisis response generation.
+That is the exact error this issue corrects.
 
 ---
 
-## 6. Fine-Tuning Potential
+## 6. Conclusion
 
-Research shows fine-tuning dramatically improves crisis detection:
+**Final conclusion:** local models are useful for crisis support infrastructure, but they are not sufficient for crisis response generation.
 
-- **Without fine-tuning:** Best LLM lags supervised models by 6.95% (suicide task) to 31.53% (cognitive distortion)
-- **With fine-tuning:** Gap narrows to 4.31% and 3.14% respectively
-- **Key insight:** Even a 1.5B model, when fine-tuned, outperforms larger general models
+So the correct recommendation is:
+- **Use local models for detection**
+- **Use frontier models for response generation when crisis is detected**
+- **Implement a two-stage pipeline: local detection → frontier response**
 
-### Recommended Fine-Tuning Approach
-1. Collect crisis conversation data (anonymized)
-2. Fine-tune on suicidal ideation detection
-3. Fine-tune on empathetic response generation
-4. Fine-tune on safety protocol adherence
-5. Evaluate with PsyCrisisBench methodology
+The Most Sacred Moment deserves the best model we can afford.
 
 ---
 
-## 7. Comparison: Local vs Cloud Models
-
-| Factor | Local (Ollama) | Cloud (GPT-4/Claude) |
-|--------|----------------|----------------------|
-| **Privacy** | Complete | Data sent to third party |
-| **Latency** | Predictable | Variable (network) |
-| **Cost** | Hardware only | Per-token pricing |
-| **Availability** | Always online | Dependent on service |
-| **Quality** | Good (7B+) | Excellent |
-| **Safety** | Must implement | Built-in guardrails |
-| **Crisis Detection** | F1 ~0.85-0.90 | F1 ~0.88-0.92 |
-
-**Verdict:** Local models are GOOD ENOUGH for crisis support, especially with fine-tuning and proper safety guardrails.
-
----
-
-## 8. Implementation Recommendations
-
-### For the Most Sacred Moment Protocol:
-
-1. **Use a two-model architecture:**
-   - Primary: Qwen2.5-7B for conversation
-   - Safety: Llama-Guard3 for content filtering
-
-2. **Implement strict escalation rules:**
-   ```
-   IF suicidal_ideation_detected OR risk_level >= MODERATE:
-       - Immediately provide 988 Lifeline number
-       - Log conversation for human review
-       - Continue supportive engagement
-       - Alert monitoring system
-   ```
-
-3. **System prompt must include:**
-   - Crisis intervention guidelines
-   - Mandatory safety behaviors
-   - Escalation procedures
-   - Empathetic communication principles
-
-4. **Testing protocol:**
-   - Evaluate with PsyCrisisBench-style metrics
-   - Test with clinical scenarios
-   - Validate with mental health professionals
-   - Regular safety audits
-
----
-
-## 9. Risks and Limitations
-
-### Critical Risks
-1. **False negatives:** Missing someone in crisis (12-17% rate)
-2. **Over-reliance:** Users may treat AI as substitute for professional help
-3. **Hallucination:** Model may generate inappropriate or harmful advice
-4. **Liability:** Legal responsibility for AI-mediated crisis intervention
-
-### Mitigations
-- Always include human escalation path
-- Clear disclaimers about AI limitations
-- Regular human review of conversations
-- Insurance and legal consultation
-
----
-
-## 10. Key Citations
-
-1. Deng et al. (2025). "Evaluating Large Language Models in Crisis Detection: A Real-World Benchmark from Psychological Support Hotlines." arXiv:2506.01329. PsyCrisisBench.
-
-2. Wiest et al. (2024). "Detection of suicidality from medical text using privacy-preserving large language models." British Journal of Psychiatry, 225(6), 532-537.
-
-3. Holmes et al. (2025). "Applications of Large Language Models in the Field of Suicide Prevention: Scoping Review." J Med Internet Res, 27, e63126.
-
-4. Levkovich & Omar (2024). "Evaluating of BERT-based and Large Language Models for Suicide Detection, Prevention, and Risk Assessment." J Med Syst, 48(1), 113.
-
-5. Shukla et al. (2026). "Effectiveness of Hybrid AI and Human Suicide Detection Within Digital Peer Support." J Clin Med, 15(5), 1929.
-
-6. Qi et al. (2025). "Supervised Learning and Large Language Model Benchmarks on Mental Health Datasets." Bioengineering, 12(8), 882.
-
-7. Liu et al. (2025). "Enhanced large language models for effective screening of depression and anxiety." Commun Med, 5(1), 457.
-
----
-
-## Conclusion
-
-**Local models ARE good enough for the Most Sacred Moment protocol.**
-
-The research is clear:
-- Crisis detection F1 scores of 0.88-0.91 are achievable
-- Fine-tuned small models (1.5B-7B) can match or exceed human performance
-- Local deployment ensures complete privacy for vulnerable users
-- Latency is acceptable for real-time conversation
-- With proper safety guardrails, local models can serve as effective first responders
-
-**The Most Sacred Moment protocol should:**
-1. Use Qwen2.5-7B or similar as primary conversational model
-2. Implement Llama-Guard3 as safety filter
-3. Build in immediate 988 Lifeline escalation
-4. Maintain human oversight and review
-5. Fine-tune on crisis-specific data when possible
-6. Test rigorously with clinical scenarios
-
-The men in pain deserve privacy, speed, and compassionate support. Local models deliver all three.
-
----
-
-*Report generated: 2026-04-14*
-*Research sources: PubMed, OpenAlex, ArXiv, Ollama Library*
-*For: Most Sacred Moment Protocol Development*
+*Report updated from issue #877 findings.*
+*Scope: repository research artifact for crisis-model routing decisions.*

tests/fixtures/holographic_recall_matrix.json

@@ -1,56 +0,0 @@
-{
-  "facts": [
-    {
-      "content": "Alexander Whitestone aka Rockachopa.",
-      "category": "general",
-      "tags": "identity alias"
-    },
-    {
-      "content": "Rockachopa uses Ansible playbooks for sovereign rollouts.",
-      "category": "project",
-      "tags": "ansible playbooks rollout"
-    },
-    {
-      "content": "The provider is anthropic/claude-haiku-4-5.",
-      "category": "project",
-      "tags": "provider default",
-      "updated_at": "2026-01-01T00:00:00Z"
-    },
-    {
-      "content": "Correction: the provider is mimo-v2-pro.",
-      "category": "project",
-      "tags": "provider current",
-      "updated_at": "2026-04-20T00:00:00Z"
-    },
-    {
-      "content": "Ezra operates the BURN2 lane for forge work.",
-      "category": "project",
-      "tags": "ezra burn2 forge lane"
-    },
-    {
-      "content": "BURN2 handles forge triage and review.",
-      "category": "project",
-      "tags": "forge triage review"
-    }
-  ],
-  "queries": [
-    {
-      "name": "semantic_alias_graph",
-      "query": "What automation does Alexander Whitestone use for deploys?",
-      "expected_substring": "Ansible playbooks",
-      "top_k": 1
-    },
-    {
-      "name": "temporal_correction",
-      "query": "What provider should we use?",
-      "expected_substring": "mimo-v2-pro",
-      "top_k": 1
-    },
-    {
-      "name": "graph_lane",
-      "query": "Which forge lane does Ezra operate?",
-      "expected_substring": "BURN2 lane",
-      "top_k": 1
-    }
-  ]
-}

tests/plugins/memory/test_holographic_retrieval.py

@@ -1,116 +0,0 @@
-"""Tests for multi-path holographic retrieval fusion and traceability."""
-
-from __future__ import annotations
-
-import json
-import sys
-from pathlib import Path
-
-sys.path.insert(0, str(Path(__file__).resolve().parents[3]))
-
-from plugins.memory.holographic import HolographicMemoryProvider
-from plugins.memory.holographic.retrieval import FactRetriever, format_benchmark_report
-from plugins.memory.holographic.store import MemoryStore
-
-_FIXTURE_PATH = Path(__file__).resolve().parents[2] / "fixtures" / "holographic_recall_matrix.json"
-
-
-def _fixture() -> dict:
-    return json.loads(_FIXTURE_PATH.read_text())
-
-
-def _seed_store(tmp_path) -> MemoryStore:
-    store = MemoryStore(db_path=tmp_path / "memory_store.db")
-    for fact in _fixture()["facts"]:
-        fact_id = store.add_fact(fact["content"], category=fact["category"], tags=fact.get("tags", ""))
-        if fact.get("updated_at"):
-            store._conn.execute(
-                "UPDATE facts SET created_at = ?, updated_at = ? WHERE fact_id = ?",
-                (fact["updated_at"], fact["updated_at"], fact_id),
-            )
-    store._conn.commit()
-    return store
-
-
-class TestMultiPathRetrieval:
-    def test_lane_toggle_and_trace_contributions(self, tmp_path):
-        store = _seed_store(tmp_path)
-        retriever = FactRetriever(store=store)
-
-        payload = retriever.search_with_trace(
-            "Which forge lane does Ezra operate?",
-            limit=3,
-            lanes=["lexical", "graph"],
-        )
-
-        assert payload["trace"]["lanes_run"] == ["lexical", "graph"]
-        assert payload["results"]
-        top = payload["results"][0]
-        assert "BURN2 lane" in top["content"]
-        assert "graph" in top["lane_contributions"]
-        assert set(top["lane_contributions"]).issubset({"lexical", "graph"})
-
-    def test_trace_available_for_failed_recall(self, tmp_path):
-        store = _seed_store(tmp_path)
-        retriever = FactRetriever(store=store)
-
-        payload = retriever.search_with_trace(
-            "nonexistent memory topic xyz123",
-            limit=3,
-            lanes=["lexical", "semantic", "graph", "temporal"],
-        )
-
-        assert payload["results"] == []
-        assert payload["trace"]["fused_count"] == 0
-        assert payload["trace"]["lane_hits"]["lexical"] == 0
-        assert payload["trace"]["lane_hits"]["semantic"] == 0
-
-    def test_benchmark_prompt_matrix_shows_gain_over_baseline(self, tmp_path):
-        store = _seed_store(tmp_path)
-        retriever = FactRetriever(store=store)
-        report = retriever.benchmark_prompt_matrix(_fixture()["queries"], limit=3)
-
-        assert report["fused_top1_hits"] > report["baseline_top1_hits"]
-        assert report["improvement"] > 0
-
-        rendered = format_benchmark_report(report)
-        assert "Prompt matrix benchmark" in rendered
-        assert "semantic_alias_graph" in rendered
-        assert "improvement" in rendered.lower()
-
-
-class TestHolographicProviderTrace:
-    def test_prefetch_records_trace_and_trace_action_returns_it(self, tmp_path):
-        provider = HolographicMemoryProvider(
-            config={
-                "db_path": str(tmp_path / "provider.db"),
-                "retrieval_lanes": ["lexical", "semantic", "graph", "temporal"],
-                "enable_rerank": True,
-            }
-        )
-        provider.initialize("test-session")
-
-        seed_store = _seed_store(tmp_path / "seed")
-        rows = seed_store.list_facts(min_trust=0.0, limit=20)
-        for row in rows:
-            provider._store.add_fact(row["content"], category=row["category"], tags=row.get("tags", ""))
-            if row["content"].startswith("The provider is anthropic"):
-                provider._store._conn.execute(
-                    "UPDATE facts SET created_at = ?, updated_at = ? WHERE content = ?",
-                    ("2026-01-01T00:00:00Z", "2026-01-01T00:00:00Z", row["content"]),
-                )
-            elif row["content"].startswith("Correction: the provider is mimo"):
-                provider._store._conn.execute(
-                    "UPDATE facts SET created_at = ?, updated_at = ? WHERE content = ?",
-                    ("2026-04-20T00:00:00Z", "2026-04-20T00:00:00Z", row["content"]),
-                )
-        provider._store._conn.commit()
-
-        block = provider.prefetch("What provider should we use?")
-        assert "Holographic Memory" in block
-        assert "mimo-v2-pro" in block
-
-        trace_payload = json.loads(provider.handle_tool_call("fact_store", {"action": "trace"}))
-        assert trace_payload["trace"]["query"] == "What provider should we use?"
-        assert trace_payload["trace"]["rerank_applied"] in {True, False}
-        assert trace_payload["trace"]["lane_hits"]["temporal"] >= 1

tests/test_research_local_model_crisis_quality.py

@@ -0,0 +1,16 @@
+from pathlib import Path
+
+
+REPORT = Path(__file__).resolve().parent.parent / "research_local_model_crisis_quality.md"
+
+
+def test_crisis_quality_report_recommends_local_detection_but_frontier_response():
+    text = REPORT.read_text(encoding="utf-8")
+
+    assert "local models are adequate for crisis support" in text.lower()
+    assert "not for crisis response generation" in text.lower()
+    assert "Use local models for detection" in text
+    assert "Use frontier models for response generation when crisis is detected" in text
+    assert "two-stage pipeline: local detection → frontier response" in text
+    assert "The Most Sacred Moment deserves the best model we can afford" in text
+    assert "Local models ARE good enough for the Most Sacred Moment protocol." not in text