Asist.IA

A production-ready, enterprise-grade AI agent platform that enables businesses to build intelligent conversational agents with RAG, tool execution, multi-tenancy, and extensive customization options.

Commercial Project
In Development
December 22, 2025

Technologies Used

NestJS
React
TanStack Start
TypeScript
PostgreSQL
pgvector
Drizzle ORM
Better Auth
BullMQ
Redis
Socket.io
Tailwind CSS
Zustand
TurboRepo

Project Description

Asistia is a comprehensive AI agent platform designed for businesses that need to deploy intelligent conversational assistants across multiple channels. Unlike simple chatbot builders, Asistia provides a complete ecosystem for building production-ready AI agents with enterprise features.

The platform handles the entire lifecycle of AI agent management: from creating and configuring bots, uploading knowledge bases, connecting communication channels, to monitoring conversations and tracking costs.

Core Architecture

Multi-Tenant Design

Every tenant operates in complete isolation with their own:

  • Bots and configurations
  • Knowledge bases and documents
  • Conversations and message history
  • Users and permissions
  • Billing plans and quotas

AI Conversation Engine

The heart of Asistia is the conversation engine that orchestrates:

  • Model Registry - Pluggable AI providers (OpenAI, Claude)
  • RAG Service - Semantic search across knowledge bases
  • Prompt Builder - Dynamic system prompts with context injection
  • Tool Executor - External service integrations
  • Cost Tracking - Per-token cost calculation
  • Content Filter - Input/output moderation

Plugin System

Extend functionality without modifying core code:

  • Channels - Add new communication platforms
  • Models - Integrate new AI providers
  • Embedders - Add embedding providers
  • Parsers - Support new document formats
  • Tools - Create custom service integrations
  • Billing - Add payment providers

Key Features

Knowledge Base & RAG

Upload documents and build a searchable knowledge base:

  • Intelligent chunking strategies for optimal retrieval
  • Vector embeddings with pgvector
  • Semantic search with citation tracking
  • Support for PDF, DOCX, TXT, and more

Channel Integrations

Deploy bots across multiple platforms:

  • Web Chat - Embeddable widget with customizable UI
  • Telegram - Full Bot API integration
  • Extensible - Add custom channels via plugins

Enterprise Features

Built for production environments:

  • Granular permission system with 50+ permissions
  • Role-based access control (Super Admin, Client, Agent)
  • Audit logs for compliance
  • Analytics and reporting
  • Rate limiting and quota management

Technical Stack

The project uses a modern monorepo architecture with Turbo:

  • Backend: NestJS with PostgreSQL and Redis
  • Frontend: React with TanStack ecosystem
  • Real-time: Socket.io for live updates
  • Jobs: BullMQ for background processing
  • Storage: S3-compatible (MinIO/AWS)

Current Status

Asistia is in active development with core features implemented. The focus is on stabilizing the conversation engine and expanding channel integrations before the first production release.

Project Metrics

159
Commits
81,730
Lines of Code
2 weeks
Development Time
Dec 2025
Started
1
Contributors
0
Open Issues

Key Highlights

  • Multi-channel support: Web, Telegram, and custom channels
  • Multi-tenancy with billing plans (Free, Starter, Professional, Enterprise)
  • RAG integration with pgvector for semantic search
  • Extensible plugin system for channels, models, tools, and parsers
  • 50+ granular permissions with role-based access control
  • Real-time WebSocket support for live conversations
  • AI cost tracking per conversation and tenant
  • Document processing with PDF, DOCX, TXT support
  • Built-in integrations: Google Calendar, Google Sheets
  • Content moderation with PII detection

Challenges & Solutions

Problem

Multi-tenant data isolation - Ensuring complete separation between tenant data

Solution

Row-level security with tenant ID in all queries and middleware validation

Problem

Plugin architecture complexity - Making the system extensible without core modifications

Solution

Manifest-based plugin system with dependency injection and clear interfaces

Problem

RAG performance at scale - Vector search slowing down with large document sets

Solution

pgvector with HNSW indexes and intelligent chunking strategies

Problem

Real-time conversation sync - Keeping multiple clients in sync during conversations

Solution

Socket.io rooms per conversation with optimistic updates

Why This Stack?

NestJS

Enterprise-grade Node.js framework with excellent DI and modularity

TanStack Start

Full-stack React framework with type-safe routing

Drizzle ORM

Type-safe ORM with excellent PostgreSQL support

pgvector

Native vector similarity search in PostgreSQL

BullMQ

Robust job queue for document processing and async tasks

Better Auth

Modern authentication with session management

Future Improvements

  • Additional channel integrations (WhatsApp Business, Slack)
  • More AI model providers (Gemini, Mistral)
  • Advanced analytics dashboard
  • White-label customization
  • Marketplace for community plugins