Knowledge Base
Upload any policy, FAQ, or reference document. Kuram and the Social Media Triage bot will read it once, write wiki pages, and ground every answer in these documents.
Wiki Pages
Click any page to view its full contents
Talk to Kuram
Kuram is trained on your uploaded knowledge base and follows the published tone guide. Every answer cites the page it used. Try asking about the Emergency Fund, safety, confidentiality, or anything TechHer supports.
Social Media DM Simulator
Pretend you're a member of the public sending TechHer a DM. The AI triage bot will classify it, draft a reply if appropriate, and log it to the review queue — all in real time.
Review Queue
Live view of what AI has triaged. Every row is a human-reviewable audit record.
🚨 Fund Applications — needs human review
💬 Social Media Triage Log
What you're looking at
This is a live AI lab connected to TechHer NG's own server. Everything here runs on infrastructure you control, at n8n.nuton.app.
How each tab works
- 📚 Knowledge Base — upload a document. The AI reads it once, writes structured wiki pages, and updates the catalog. No vector databases, no embeddings — just markdown files that humans can read and edit.
- 💬 Talk to Kuram — a real chat with the AI, grounded entirely in your knowledge base. Try asking something it does know (like emergency fund documents) and something it doesn't (like the weather). Notice it says "I don't have documented guidance on this" rather than making something up.
- 📱 Simulate DM — send a fake social-media DM and watch the triage bot classify, draft a reply, and log it. This is exactly what would happen if someone messaged your real Instagram or WhatsApp.
- 📊 Review Queue — the audit log. Every automated decision logged with its reasoning. Nothing silent, nothing untraceable.
Real Telegram bot
There's also a live Telegram bot: search for @KuramSupportBot on Telegram. Message it and it runs through the same triage workflow.
Under the hood
n8n workflows on a self-hosted VPS · Groq gpt-oss-120b for all LLM calls · Wiki stored as markdown on the server · Data persistence via n8n native Data Tables · DNS/SSL via Traefik + Let's Encrypt · Zero third-party analytics · Zero third-party data processors except Groq (which does not retain inputs).