diff --git a/CLAUDE.md b/CLAUDE.md
index a8846e2..92d0781 100644
--- a/CLAUDE.md
+++ b/CLAUDE.md
@@ -4,7 +4,7 @@ This file provides guidance to Claude Code (claude.ai/code) when working with co
 
 ## Project Overview
 
-Bishop State Student Success Prediction - Full-stack ML + web application predicting student outcomes for Bishop State Community College. Uses 5 ML models to generate retention predictions, early warnings, time-to-credential estimates, credential type forecasts, and GPA predictions for ~4K students.
+Bishop State Student Success Prediction - Full-stack ML + web application predicting student outcomes for Bishop State Community College. Uses 6 ML models to generate retention predictions, time-to-credential estimates, credential type forecasts, gateway math/English success predictions, and first-semester low-GPA predictions for ~4K students. Authoritative AI/ML inventory: `codebenders-dashboard/content/ai-transparency.ts`.
 
 ## Tech Stack
 
@@ -15,14 +15,14 @@ Bishop State Student Success Prediction - Full-stack ML + web application predic
 | Charts | Recharts |
 | UI Components | shadcn/ui (Radix UI) |
 | Database | Postgres (Supabase), pg driver |
-| AI Features | OpenAI (natural language query analysis) |
+| AI Features | OpenAI gpt-4o-mini for: NL query → SQL analyzer (`codebenders-dashboard/app/api/analyze`), query result summarizer (`codebenders-dashboard/app/api/query-summary`), course-pairing explainer (`codebenders-dashboard/app/api/courses/explain-pairing`). Rule-based fallback at `codebenders-dashboard/lib/prompt-analyzer.ts`. Authoritative list: `codebenders-dashboard/content/ai-transparency.ts`. |
 | Infrastructure | Docker Compose, Vercel |
 
 ## Key Directories
 
 | Directory | Purpose |
 |-----------|---------|
-| `ai_model/` | Python ML pipeline - 5 models (XGBoost + Random Forest) |
+| `ai_model/` | Python ML pipeline - 6 models (XGBoost + Random Forest). Authoritative inventory: `codebenders-dashboard/content/ai-transparency.ts`. |
 | `codebenders-dashboard/` | Next.js web application |
 | `codebenders-dashboard/app/` | App Router pages and API routes |
 | `codebenders-dashboard/components/` | React components (shadcn/ui based) |
@@ -87,6 +87,7 @@ Check these files for detailed information on specific topics:
 
 | Topic | File |
 |-------|------|
+| AI / ML surface inventory (models, LLM routes, data flows) | `codebenders-dashboard/content/ai-transparency.ts` |
 | Architectural patterns | `.claude/docs/architectural_patterns.md` |
 | Project overview | `README.md` |
 | Quick start guide | `QUICKSTART.md` |
diff --git a/README.md b/README.md
index 35f124d..5768964 100644
--- a/README.md
+++ b/README.md
@@ -17,23 +17,26 @@ A comprehensive machine learning pipeline for predicting student success outcome
 
 ## 🎯 Overview
 
-This project implements five machine learning models to predict various aspects of student success:
+This project implements six machine learning models to predict various aspects of student success (authoritative names and data flows: [`codebenders-dashboard/content/ai-transparency.ts`](codebenders-dashboard/content/ai-transparency.ts)):
 
 1. **Retention Prediction** - Will the student be retained?
-2. **Early Warning System** - Is the student at risk?
-3. **Time-to-Credential** - How long until graduation?
-4. **Credential Type** - What credential will they earn?
-5. **Course Success** - What will their GPA be?
+2. **Time-to-Credential** - How long until credential completion?
+3. **Credential Type** - What credential will they earn?
+4. **Gateway Math Success** - Will the student succeed in gateway math?
+5. **Gateway English Success** - Will the student succeed in gateway English?
+6. **First-Semester Low-GPA Prediction** - Is the student at risk of a low first-semester GPA?
 
 The models use demographic, academic preparation, enrollment, and course performance data to generate actionable predictions for student support services.
 
+The Next.js dashboard adds **natural language query (NLQ)** features: three OpenAI `gpt-4o-mini` API routes (`codebenders-dashboard/app/api/analyze/route.ts`, `codebenders-dashboard/app/api/query-summary/route.ts`, `codebenders-dashboard/app/api/courses/explain-pairing/route.ts`), a **rule-based fallback** in `codebenders-dashboard/lib/prompt-analyzer.ts`, and (when not using direct database mode) an **external data API** at `schools.syntex-ai.com`. See the same `ai-transparency.ts` file for the full inventory.
+
 ## 📁 Project Structure
 
 ```
 codebenders-datathon/
 ├── ai_model/                          # Machine learning models and scripts
 │   ├── __init__.py                    # Package initialization
-│   ├── complete_ml_pipeline.py        # Main ML pipeline (5 models)
+│   ├── complete_ml_pipeline.py        # Main ML pipeline (6 models)
 │   ├── generate_bishop_state_data.py  # Synthetic data generation
 │   └── merge_bishop_state_data.py     # Data merging script
 │
@@ -59,11 +62,11 @@ codebenders-datathon/
 
 ### Prediction Capabilities
 
-- **Retention Risk Assessment**: Identify students at risk of not returning
-- **Early Warning Alerts**: Four-level alert system (URGENT, HIGH, MODERATE, LOW)
+- **Retention Risk Assessment**: Retention probability and risk categories; dashboard alert views (URGENT / HIGH / MODERATE / LOW) are driven by these signals
 - **Graduation Timeline**: Predict time to credential completion
 - **Credential Path**: Forecast credential type (Certificate, Associate's, Bachelor's)
-- **Academic Performance**: Predict expected GPA and identify over/underperformers
+- **Gateway Success**: Predict gateway math and English completion outcomes
+- **Early Academic Risk**: First-semester low-GPA risk prediction
 
 ### Technical Features
 
@@ -139,7 +142,7 @@ python complete_ml_pipeline.py
 This will:
 1. Test database connection
 2. Load and preprocess data
-3. Train all 5 models
+3. Train all 6 models
 4. Generate predictions for all students
 5. Save results to **Postgres database** (or CSV files as fallback)
 6. Save model performance metrics to database
@@ -195,47 +198,31 @@ For more details, see [operations/README.md](operations/README.md).
 
 ## 🤖 Models
 
-### 1. Retention Prediction Model
-
-**Algorithm**: XGBoost Classifier
-**Target**: Binary (Retained / Not Retained)
-**Features**: 40+ demographic, academic, and performance features
-
-**Output**:
-- `retention_probability`: Probability of retention (0-1)
-- `retention_prediction`: Binary prediction (0/1)
-- `retention_risk_category`: Risk level (Critical/High/Moderate/Low)
-
-### 2. Early Warning System
+Authoritative descriptions (inputs, algorithms, data flow): [`codebenders-dashboard/content/ai-transparency.ts`](codebenders-dashboard/content/ai-transparency.ts). The summaries below match that inventory.
 
-**Algorithm**: Composite Risk Score
-**Target**: Binary (At Risk / Not At Risk)
-**Approach**: Combines retention probability with performance metrics
+### 1. Retention Prediction
 
-**Risk Factors**:
-- Retention probability (50% weight)
-- GPA performance (20% weight)
-- Course completion rate (20% weight)
-- Credit progress (10% weight)
+**Algorithm**: XGBoost classifier (model family selected in `ai_model/complete_ml_pipeline.py`)
+**Target**: Binary (Retained / Not Retained)
+**Features**: Demographic, enrollment, year-one performance, and program signals
 
-**Output**:
-- `risk_score`: Comprehensive risk score (0-100)
-- `at_risk_alert`: Alert level (URGENT/HIGH/MODERATE/LOW)
-- `at_risk_probability`: Risk probability (0-1)
-- `at_risk_prediction`: Binary prediction (0/1)
+**Output** (examples):
+- Retention probability and binary prediction
+- Retention risk category (Critical / High / Moderate / Low)
+- Dashboard risk alerts combine retention and related metrics
 
-### 3. Time-to-Credential Model
+### 2. Time-to-Credential Prediction
 
-**Algorithm**: XGBoost Regressor
+**Algorithm**: Random Forest regressor
 **Target**: Continuous (Years to credential)
 
 **Output**:
 - `predicted_time_to_credential`: Years to completion
 - `predicted_graduation_year`: Expected graduation year
 
-### 4. Credential Type Model
+### 3. Credential Type Prediction
 
-**Algorithm**: Random Forest Classifier
+**Algorithm**: Random Forest multi-class classifier
 **Target**: Multi-class (No Credential / Certificate / Associate's / Bachelor's)
 
 **Output**:
@@ -243,14 +230,26 @@ For more details, see [operations/README.md](operations/README.md).
 - `predicted_credential_label`: Text label
 - `prob_no_credential`, `prob_certificate`, `prob_associate`, `prob_bachelor`: Class probabilities
 
-### 5. Course Success Model
+### 4. Gateway Math Success Prediction
 
-**Algorithm**: Random Forest Regressor
-**Target**: Continuous (GPA 0-4 scale)
+**Algorithm**: XGBoost classifier
+**Target**: Binary (success in gateway math)
 
-**Output**:
-- `predicted_gpa`: Expected GPA (0-4 scale)
-- `gpa_performance`: Performance vs. expected (Above/Below/As Expected)
+**Output**: Probability and prediction fields written to `student_predictions` (see data dictionary and pipeline outputs).
+
+### 5. Gateway English Success Prediction
+
+**Algorithm**: XGBoost classifier
+**Target**: Binary (success in gateway English)
+
+**Output**: Probability and prediction fields written to `student_predictions`.
+
+### 6. First-Semester Low-GPA Prediction
+
+**Algorithm**: XGBoost classifier
+**Target**: Binary (low first-semester GPA risk)
+
+**Output**: Probability and prediction fields written to `student_predictions`.
 
 ## 📊 Data
 
@@ -299,6 +298,7 @@ If database connection fails, predictions are saved to CSV:
 
 ## 📚 Documentation
 
+- **[codebenders-dashboard/content/ai-transparency.ts](codebenders-dashboard/content/ai-transparency.ts)**: Authoritative inventory of ML models, OpenAI/NLQ routes, rule-based fallback, and external data API surfaces
 - **[DATA_DICTIONARY.md](DATA_DICTIONARY.md)**: Detailed descriptions of all data fields
 - **[ML_MODELS_GUIDE.md](ML_MODELS_GUIDE.md)**: In-depth guide to machine learning models
 - **[DOCKER_SETUP.md](DOCKER_SETUP.md)**: Docker Compose setup for local Postgres
diff --git a/codebenders-dashboard/DASHBOARD_README.md b/codebenders-dashboard/DASHBOARD_README.md
index 1a25f44..bf013f5 100644
--- a/codebenders-dashboard/DASHBOARD_README.md
+++ b/codebenders-dashboard/DASHBOARD_README.md
@@ -4,6 +4,8 @@
 
 A modern, interactive dashboard for visualizing student success metrics and predictive analytics for Bishop State Community College.
 
+**Authoritative AI/ML inventory** (six trained models, three OpenAI `gpt-4o-mini` routes, rule-based NLQ fallback, external data API when not on direct DB): [`content/ai-transparency.ts`](content/ai-transparency.ts) (path from repo root: `codebenders-dashboard/content/ai-transparency.ts`).
+
 ## Features
 
 ### 📊 Executive Dashboard (Home Page - `/`)
@@ -38,9 +40,9 @@ Color-coded from red (critical) to green (low risk).
 ### 🔍 SQL Query Interface (`/query`)
 
 Advanced query interface for custom data analysis:
-- Natural language to SQL conversion
+- Natural language to SQL via OpenAI `gpt-4o-mini` (`app/api/analyze/route.ts`), optional result summarization (`app/api/query-summary/route.ts`), and course-pairing explanations (`app/api/courses/explain-pairing/route.ts`), with **rule-based fallback** in `lib/prompt-analyzer.ts` when the LLM path is off or disabled
 - Support for multiple institutions (Bishop State, University of Akron, Cal State San Bernardino, Thomas More)
-- Direct database or API mode
+- Direct database or **external data API** (`schools.syntex-ai.com`) when not in direct DB mode (see `content/ai-transparency.ts`)
 - Interactive visualizations (line, bar, pie charts, tables)
 - Query plan visualization
 
@@ -154,11 +156,19 @@ Returns retention risk categories:
 ### Query APIs
 
 #### `POST /api/analyze`
-Analyzes natural language prompts and generates SQL queries.
+Analyzes natural language prompts and generates SQL (OpenAI `gpt-4o-mini`, with rule-based fallback in `lib/prompt-analyzer.ts`).
+
+#### `POST /api/query-summary`
+Summarizes query results in natural language (OpenAI `gpt-4o-mini`).
+
+#### `POST /api/courses/explain-pairing`
+Explains course-pairing recommendations (OpenAI `gpt-4o-mini`).
 
 #### `POST /api/execute-sql`
 Executes SQL queries directly against the database.
 
+Full route list and data-flow notes: [`content/ai-transparency.ts`](content/ai-transparency.ts).
+
 ## Environment Variables
 
 Create a `.env.local` file:
@@ -252,6 +262,7 @@ See `/DASHBOARD_VISUALIZATIONS.md` for a comprehensive list of additional visual
 
 ## References
 
+- **AI / ML surface inventory**: `codebenders-dashboard/content/ai-transparency.ts` (authoritative list of models, LLM routes, and integrations)
 - **Visualization Guide**: `/DASHBOARD_VISUALIZATIONS.md`
 - **Schema Documentation**: `codebenders-dashboard/env.example`
 - **Project PRD**: `/AI_Powered_Student_Success_PRD.md`