Prefrontal

Evolutionary psychology explainer — trace modern fears to ancestral origins.

A high-fidelity RAG-powered platform designed to decode the biological legacy of our ancient brain using Groq LLMs and Pinecone vector search.

Installation →  |  Architecture →  |  Tech Stack →

---

Overview

Prefrontal is your personal AI-powered companion for tracing modern anxieties, behaviors, and cravings back to their survival origins. Whether it's the paralyzing fear of social rejection (ancestral exile) or the irresistible urge for sugar (ancient scarcity), Prefrontal provides a clinical yet accessible explanation by bridging modern neuroscience with evolutionary biology.

What makes this different:

• RAG-First Intelligence — Every answer is grounded in curated evolutionary data and academic texts (Sapolsky, Dawkins, Lieberman).
• Sub-100ms Inference — Leverages Groq's Llama-3.3-70b for near-instantaneous streaming explanations.
• Biometric UI — A premium, glassmorphic interface designed with a clinical aesthetic and "neural" loading states.
• Traceable Context — Users can see exactly which survival patterns and academic sources were used to construct the explanation.

---

🖥️ Clinical Interface & Card Telemetry

Prefrontal Exploration Interface: Natural language prompt input, suggestion chips (Doomscrolling, Sugar Cravings), and neural pathway analysis loader.

Evolutionary Insight Card (Digital Information Overload): Ancestral survival mechanisms, brain chemistry breakdown, Sapolsky/Dawkins citations, and core takeaway insight.

---

Architecture

graph TD
    User((User)) -->|Query| Frontend[Next.js Frontend]
    Frontend -->|POST /api/explain| Backend[FastAPI Backend]
    
    subgraph "RAG Pipeline"
        Backend -->|Semantic Search| Embedder[Sentence Transformer]
        Embedder -->|Vector| Pinecone[(Pinecone Vector DB)]
        Pinecone -->|Context Chunks| Backend
    end
    
    subgraph "Inference Layer"
        Backend -->|System Prompt + Context| Groq[Groq Cloud]
        Groq -->|Streaming Llama-3.3-70b| Backend
    end
    
    Backend -->|Server-Sent Events| Frontend
    Frontend -->|Render| Explanation[Explanation Card]

Loading

The system employs a Retrieval-Augmented Generation (RAG) architecture. When a user asks about a behavior, the backend performs a semantic search against a Pinecone index containing thousands of vectors from evolutionary psychology papers and curated survival patterns. This context is then piped into Llama-3.3-70b on Groq to generate a scientifically grounded response.

---

Features

Feature	Detail
Streaming Explanations	Real-time SSE streaming for instant feedback, even for complex queries.
Semantic Knowledge Base	Vector search across PDF books and curated seed data using BGE-base-en-v1.5.
Neural Loader	A custom-built CSS animation that visualizes "neural firing" during retrieval.
Context Metadata	Displays retrieval latency and the number of survival patterns analyzed.
Cross-Platform Scripts	One-click .bat scripts for Installation, Running, Testing, and Uninstallation.
Suggestion Chips	Pre-defined queries for common behaviors like "Sugar Cravings" and "Fear of the Dark".

---

Tech Stack

Layer	Technology	Purpose
Frontend	Next.js 15 (App Router)	High-performance, SEO-friendly UI with React 19.
Backend	FastAPI (Python 3.11+)	Asynchronous API for low-latency RAG processing.
LLM Engine	Groq (Llama-3.3-70b)	Ultra-fast inference with 70B parameter precision.
Vector DB	Pinecone	Serverless vector index for semantic retrieval.
Embeddings	BAAI/bge-base-en-v1.5	State-of-the-art local embeddings via Sentence-Transformers.
Orchestration	LangChain	Pipeline management for retrieval and LLM integration.
Styling	Vanilla CSS + JSX	Premium glassmorphic design without utility overhead.

---

Project Structure

Prefrontal/
├── backend/                # Python FastAPI Service
│   ├── api/                # Route handlers and schemas
│   ├── llm/                # Groq/LangChain integration
│   ├── rag/                # Embedding and retrieval logic
│   ├── main.py             # Entry point
│   └── requirements.txt    # Python dependencies
├── frontend/               # Next.js Application
│   ├── src/app/            # App Router pages
│   ├── src/components/     # UI/Neural components
│   └── package.json        # Node dependencies
├── knowledge_base/         # RAG Source Materials
│   ├── books/              # PDF academic texts (ingested)
│   └── curated/            # JSON seed patterns
├── scripts/                # Ingestion and evaluation tools
│   ├── build_index.py      # Pinecone ingestion script
│   └── eval_rag.py         # Performance testing
├── install.bat             # One-click environment setup
├── run.bat                 # One-click project launch
└── test.bat                # Comprehensive test suite

---

Installation

The project includes a robust automation suite to handle environment setup.

Prerequisites

• Python 3.11+
• Node.js 18+
• Groq API Key
• Pinecone API Key

Quick Start

1. Clone the repository:

   git clone https://github.com/Ares19v/Prefrontal.git
   cd Prefrontal

2. Configure Environment: Create a .env file in the backend/ directory based on .env.example.

3. Automatic Installation: Run the included installation script to set up the Python venv, install Node modules, and build dependencies.

   ./install.bat

4. Ingest Knowledge Base: Populate your Pinecone index with the evolutionary knowledge base.

   python scripts/build_index.py

5. Run the Project:

   ./run.bat

---

Docker Deployment

Prefrontal can be fully orchestrated using Docker Compose. This starts both the FastAPI backend and Next.js frontend in production mode.

1. Create your backend/.env file.
2. Build and run the containers:

   docker-compose up --build -d

3. The app is now running at http://localhost:3000.

---

Configuration

Variable	Description
GROQ_API_KEY	Your Groq API key for Llama-3.3 inference.
PINECONE_API_KEY	Your Pinecone API key.
PINECONE_INDEX_NAME	The name of your serverless index (default: prefrontal-knowledge).

---

© 2026 Devansh Tyagi (Ares19v). All Rights Reserved.