Synaptica AI: Neurosymbolic Orchestration Framework

Synaptica AI is an advanced, high-performance neurosymbolic orchestration engine. It efficiently routes user queries across tiered expert LLMs (Small, Medium, Large) by using "Logic-as-Loss". This allows it to dynamically cascade from highly efficient, low-cost API models to larger, more capable ones based on domain validation.

By decoupling the routing logic from complex LLM parameters, Synaptica achieves high precision orchestration, drastically reducing API costs and latency without sacrificing output quality.

Features

• Logic Router Engine: Analyzes symbolic linguistic structure to route queries to Math, Code, or General experts.
• Tiered Escalation: Automatically attempts a "Small/Cheap" expert. If output fails the Validator, it cascades effortlessly to "Medium" or "Large".
• Modern REST API Backend (FastAPI): Optimized, asynchronous orchestration endpoints.
• Stunning Next.js Visualizer: Real-time Glassmorphism UI to observe step-by-step routing, decision confidence, validation paths, and system cost/efficiency metric comparisons.

Quick Start (Docker)

1. Clone the Repo:

   git clone https://github.com/AbhijeetP21/SynapticaAI.git
   cd SynapticaAI

2. Configure API Keys: Create a .env in the backend/ directory:

   cp backend/.env.example backend/.env

   Add your GROQ_API_KEY, GEMINI_API_KEY or others.

3. Run with Docker Compose:

   docker-compose up --build

4. Experience the Demo:

   • Open your browser to http://localhost:3000
   • Test out mathematical queries involving equations, coding tasks like "write a python script for quicksort", or general factual questions to see the Orchestrator dynamically identify the domain and tier.

Architecture

• backend/: FastAPI application handling OrchestrationPipeline. Includes api_expert.py for snappy integrations with litellm.
• frontend/: Next.js 14 App Router project styled with TailwindCSS and Framer Motion for premium portfolio-worthy demonstration.
• orchestrator/: Contains the specific neurosymbolic routers (LogicRouter, HybridRouter).

Customization

You can switch the CloudAPIExpert models (e.g. Gemini Flash, Groq Llama3) in backend/config.yaml to experiment with open-source endpoints or different providers.