Meet Nexa AI.
WhatsApp's Personal AI.
A dual-personality WhatsApp agent that weaves personalized vector memories, routes tasks dynamically across LLMs, and acts as both a chill conversational companion and an executive scheduler.
About Nexa AI
A Double Agent in Your Pocket
Nexa is a tailored WhatsApp artificial intelligence trained with an organic, human-centric Malayalam-English (Manglish) personality. Operating continuously in the background, it coordinates scheduling, takes messages, and controls IoT nodes.
Dual-Personality Orchestration
Unlike generic, rigid AI agents that start every sentence with "As an AI model...", Nexa features a dynamic persona router. It automatically adjusts its linguistic tone and utility set based on who is asking and what they need.
Nexa Engine
Active: Chill Chat Core
Autonomous Capabilities
Equipped with Autonomous Tools
Nexa isn't just a text bot. It acts on the environment through integrated APIs and node libraries, converting simple chat statements into workflows.
Core Service Architecture
End-to-End Execution Pipeline
Nexa processes incoming chat streams through an ultra-low latency, decoupled pipeline. It leverages localized databases and remote APIs in parallel.
1. Ingress Packet Stream
WhatsApp webhooks intercept incoming chat packets in real-time. System performs active authentication check against stored session tokens.
2. Vector Context Fusion
Computes user intent vector. Performs vector similarity queries in Supabase database to fetch relevant memory records and PDF context.
3. HMP & ORA Execution
Decision engine parses the request, executes autonomous tool calls, manages key rotations, and executes generation passes via LLM.
4. Egress Stream
Validates response safety boundaries, parses custom Malayalam/Manglish filters, and transmits final reply back to the WhatsApp client.
Secure Encrypted Credentials
Google service accounts operate using local, high-security credentials keyfiles. Session caches store encrypted metadata keys directly in standard local directory layouts.
Core Architecture
Nexa Core System Stack
Nexa runs on a highly specialized pipeline. Here is the breakdown of its custom components and the newly introduced VASP protocol.
UCF // User Context Fabric
Dynamic Context Assembler & Memory Weaving
UCF intercept incoming streams, queries the vector embeddings db, and pulls patient/user metadata profiles (affinity scores, active mood, relationship notes). It merges these data points into a cohesive 'Master Fabric' context template, ensuring every response feels highly contextual and remembers previous events.
Nexa Live Access
Try Nexa Directly
Nexa is running live. Test using our sandbox console below, or view details for the official online WhatsApp release!
heyyy! Nexa here. I'm your personal WhatsApp assistant. How can I help you today? Ask me to search, book a slot, or generate images.
Nexa v0.1 Launch
Official release next Monday. Scan or click once the countdown ends to connect your WhatsApp!