Local AI Agent for CRM — Secure Django Integration

An autonomous AI agent for Django CRM systems powered by a local LLM (Qwen via Ollama): natural-language queries, pgvector semantic search, permission-based security, and human-in-the-loop confirmation — with data that never leaves the customer's servers.

Django

Django REST Framework

Docker

FastAPILangGraph

Neo4j

Ollamapgvector

PostgreSQL

Django CRM ma'lumotlarini tabiiy tilda so'rash va boshqarish uchun mustaqil AI agent; barcha ishlov lokal modellarda (Ollama orqali Qwen) — ma'lumot mijoz serveridan chiqmaydi.
LangGraph asosidagi ko'p-tugunli agent workflow va 40+ ixtisoslashgan tool: tashkilotlar, hamkorlar, ombor, hisob-fakturalar, bank operatsiyalari, xodimlar va analitika.
pgvector semantik qidiruv (BGE-M3 embedding, IVF indeks) — millionlab yozuv va kod bo'yicha ma'noli izlash.
Kod analizatori: AST + runtime introspeksiya orqali Django loyihalarini (URL, view, model, template) tahlil qilish; Neo4j bog'liqlik grafi.
Multi-tenant xavfsizlik: har foydalanuvchi uchun tashkilot bo'yicha avtomatik filtr, row-level ruxsat nazorati va to'liq audit log.
Risklarni (o'chirish/yangilash) HITL tasdiqlash (timeout + egalik tekshiruvi), foydalanuvchi fikri asosida o'z-o'zini o'rgatuvchi tizim va SSE real-time chat.