No description
Find a file
2026-06-03 02:11:45 +08:00
data-model.md docs: initial publish — architecture documentation 2026-05-31 23:50:07 +08:00
deployment.md docs: sync from chatbot@f02809d 2026-06-03 02:11:45 +08:00
design-decisions.md docs: initial publish — architecture documentation 2026-05-31 23:50:07 +08:00
infrastructure.md docs: sync from chatbot@f02809d 2026-06-03 02:11:45 +08:00
models.md docs: AI models reference — Ollama settings, KV cache, VRAM planning 2026-06-02 11:22:40 +08:00
overview.md docs: sync from chatbot@6e0cdbc 2026-05-31 23:53:09 +08:00
README.md docs: initial publish — architecture documentation 2026-05-31 23:50:07 +08:00

Chatbot Platform — Architecture Documentation

Technical reference for the Chatbot Platform, a self-hosted AI-powered sentiment collection system built for government and enterprise clients.

Documents

Document Description
Overview System architecture, conversation flow, and implementation state
Data Model Every table, column, constraint, and index
Design Decisions The why behind key technical choices
Infrastructure Stack components, environment variables, OTP supervision tree
Deployment Provisioning, hardware sizing, monitoring, upgrade and backup procedures

Stack

  • Runtime: Elixir / Phoenix LiveView / Bandit
  • Database: PostgreSQL 18 (UUIDv7 primary keys)
  • LLM: Ollama (Qwen2.5:32b) — fully self-hosted, no external API calls
  • Vector store: Qdrant — per-campaign document embeddings (BGE-M3)
  • Session cache: Redis 7
  • Channels: WhatsApp Cloud API, Telegram Bot API