n8n Workflows

HypnoELP uses n8n (self-hosted) for automation workflows that connect AI services, handle email, and orchestrate multi-step processes. Three main workflows run on OVH VPS servers.

Workflow Summary

1. AI Voice Chat

Handles the Mentor AI feature in the app. Users speak, the workflow transcribes, processes, and returns spoken audio.

Flow

1. User speaks in app
2. Whisper/Voxtral transcribes audio to text (STT)
3. Mistral LLM generates therapeutic response
4. Orpheus or OpenAI TTS converts response to audio
5. Audio streamed back to app

Details

• Credit-based: users spend credits per interaction
• Free weekly allowance (weekly_limit_credits) plus purchased credits (bank_limit_credits)
• Cost tracking logged to HypnoELP-AI Supabase project (yxudnnytcjuafeclsofg)
• 8 data payloads exchanged between app and workflow (see AI Voice Chat overview docs)

2. Custom Course Generation

Creates personalized hypnotherapy audio sessions from user-provided information.

Flow

1. User fills in: issue, goal, how they feel, why they want change
2. Mistral LLM generates a hypnotherapy script based on input
3. ElevenLabs TTS converts script to professional audio
4. Audio stored in Cloudflare R2
5. Course unlocked in user’s app (purchasable at $20)

Details

• Product ID: X99ENB (Custom Audio)
• Consumable purchase — user can create multiple custom courses
• Script generation uses trained Mistral model with hypnotherapy knowledge

3. Email Auto-Reply (Influencer)

Automated email handling for influencer/partner inquiries.

Flow

1. Check Inbox (IMAP poll every 1 minute)
2. Store email data (from, subject, body, message-id)
3. AI Categorize (Mistral) — classifies as SPAM, BUSINESS, or REPLY
4. SPAM/BUSINESS → forwarded to Sean
5. REPLY → AI Agent generates response using knowledge base
6. If AI can answer → sends reply via Resend (threaded in same conversation)
7. If AI cannot answer → escalates to human (forwards to Sean)

Knowledge Base

The AI can answer questions about:

• Portal URL and login instructions
• Referral code location
• Splash screen specs (1080x1920px, 9:16, PNG/JPG)
• Approval timeline (24-48 hours)
• Re-uploading rejected images

Must Escalate

• Money, payments, earnings, commissions
• Custom referral codes
• Account deletion
• Technical bugs
• Anything not in knowledge base

Configuration

• IMAP: Currently sean@omelasai.com (production will be info@omelasai.com)
• AI: Mistral API for categorization and response generation
• Send: Resend API from marketing@hypnoelp.com
• Conversation memory: Per-sender session maintained across replies
• Email threading: Uses In-Reply-To/References headers for Gmail thread continuity

Local Backup

Workflow JSON exported to n8n-workflows/email-auto-reply.json

Server Access

Status

All three workflows operational. Email auto-reply blocked from production switch (need IMAP credentials for info@omelasai.com from Chas).