Update backend/README.md

backend/README.md

@@ -18,10 +18,10 @@ HaramGuard/
 ├── agents/                    ← One file per agent
 │   ├── __init__.py
 │   ├── perception_agent.py    ← YOLO tracking + Guardrails GR1/GR2
-│   ├── risk_agent.py          ← Sliding window + count-based scoring
+│   ├── risk_agent.py          ← Clip segmentation + sliding K-window density scoring
 │   ├── reflection_agent.py    ← Self-critique design pattern (Bias detection)
 │   ├── operations_agent.py    ← Event-driven playbook + rate-limit guardrail
-│   └── coordinator_agent.py   ← GPT-4o + output validation guardrails GR-C1..5
+│   └── coordinator_agent.py   ← openai/gpt-oss-120b + output validation guardrails GR-C1..5
 │
 └── outputs/                   ← Auto-created at runtime
     ├── hajjflow_rt.db         ← Main SQLite database
@@ -37,16 +37,16 @@ Frame
 PerceptionAgent   →  FrameResult   (YOLO detect + track)
   │
   ▼
-RiskAgent         →  RiskResult    (count-based risk score)
+RiskAgent         →  RiskResult    (clip segmentation + K-window density score)
   │
   ▼
 ReflectionAgent   →  reflection{}  (bias check + correction)
   │
   ▼
 OperationsAgent   →  Decision      (event-driven, P0/P1/P2)
-  │ (P0 only)
+  │
   ▼
-CoordinatorAgent  →  plan{}        (GPT-4o action plan)
+CoordinatorAgent  →  plan{}        (openai/gpt-oss-120b action plan, all priorities)
   │
   ▼
 pipeline.state    →  Dashboard
@@ -58,9 +58,10 @@ pipeline.state    →  Dashboard
 |----|-------|-------------|
 | GR1 | PerceptionAgent | Person count capped at 500 |
 | GR2 | PerceptionAgent | Density score capped at 50 |
-| GR3 | RiskAgent | Risk score clamped to [0, 1] |
-| GR4 | OperationsAgent | P0 rate-limited (1 per 5 min per zone) |
-| GR-C1..5 | CoordinatorAgent | LLM output validation (fields, threat level, confidence, consistency, Arabic fallback) |
+| GR3 | RiskAgent | Risk score clamped to [0, 1] (density_pct / 100) |
+| GR3b | RiskAgent | level_changed suppressed during K-window warmup (first 17 frames per clip) |
+| GR4 | OperationsAgent | P0 rate-limited (1 per 5 min per zone); resets on pipeline restart |
+| GR-C1..5 | CoordinatorAgent | LLM output validation (fields, threat_level, confidence, score-level consistency, Arabic fallback) |
 | RF1..3 | ReflectionAgent | Chronic LOW bias, rising trend ignored, count-risk mismatch |
 
 ## 📊 Run Evaluation
@@ -81,9 +82,9 @@ Rubric coverage:
 - ✅ Error analysis methodology       (Section 6)
 - ✅ Evidence of iterative improvement (Section 7)
 
-## 🚀 Run Dashboard
+## 🚀 Run Backend API
 
 ```bash
 pip install -r requirements.txt
-streamlit run dashboard.py
-```
+python api.py
+```