MeerBot

Problem

Businesses want AI chatbots that answer questions based on their own knowledge base, not generic GPT responses. Existing solutions are either too expensive (enterprise), too limited (simple wrappers), or require technical expertise to set up. There was no affordable, self-hosted option that combines a web dashboard, Telegram bot, and multi-model AI with proper knowledge base management.

Solution

Built a two-part platform: AI Core (self-hosted LLM engine) and Platform (SaaS dashboard + Telegram bot). AI Core: Next.js 16 API with OpenAI-compatible endpoint, pgvector for embeddings (1536-dimensional, cosine similarity), document parsing (PDF, DOCX, TXT, URL scraping). Platform: Next.js 16 dashboard with Prisma/PostgreSQL, Grammy Telegram bot with Redis-backed sessions, YooKassa billing, per-request token counting with 1.4x markup. Knowledge base pipeline: upload document, auto-chunk (1000 chars, 200 overlap), generate embeddings via text-embedding-3-small, store in pgvector, retrieve top-5 relevant chunks per query.

Results

• +Multi-model support: GPT-4o, Claude, Gemini via LiteLLM proxy
• +RAG pipeline: document upload to vector search in seconds
• +Streaming responses (SSE) for real-time chat experience
• +Telegram bot onboarding: business profile setup in 3 steps
• +Billing: per-token pricing with cost tracking per user/model/day
• +Currently in development, demo bot live at @meerAgentAIbot

Role & timeline

Co-founder & developer. Started with a free demo bot to validate the idea. Building AI Core, Platform, Telegram bot with a partner