ION AI Management System (Discord Bot + Dashboard)

Project Brief

Client: ION Agency (Internal Automation Project)
Timeline: March 31, 2026 → April 7, 2026 (8 days active development)
Status: 🟡 In Progress — Backend 90% complete, Frontend in development
Budget: Internal resource allocation
Primary Developer: team-lexomee
Technical Support: team-krush91
Project Oversight: team-monte-cristos

Objective

Build an AI-powered management system to automate ION Agency’s internal operations and client relationship management. The system consists of two core components: (1) Employee Management — task tracking, workload monitoring, project board automation, and (2) Client Management — CRM automation, deal tracking, financial management. The goal is to eliminate manual administrative overhead and create a self-documenting, intelligent workspace using Discord as the interface layer.

Scope

In Scope — Employee Management Side

AI Discord bot for task tracking and workload monitoring
Automated project board creation and updates
Employee mention system for @notifications
Real-time progress queries (“What is [employee] working on?”)
Cross-channel posting capability
Intelligent follow-up question system
Date/time awareness for context
Concise, optimized responses (token efficiency)
Memory system using Supabase API

In Scope — Client Management Side

Deal tracking automation (closed deals, pending deals)
Client onboarding workflow
CRM data storage (client name, location, acquisition source, value, service, timeline)
Financial tracking (deal values, payment status)
Automated deal summary generation
Cross-channel announcement posting

In Scope — Technical Infrastructure

FastAPI backend
Discord bot integration
Supabase memory/database layer
NGINX reverse proxy
Cloudflare proxy for HTTPS
Web dashboard (HTML/CSS/JS)
API subdomain: api.ionagency.in

Out of Scope (Current Phase)

Public client-facing portal (future consideration)
Payment gateway integration
Multi-language support
Mobile app interface

Technical Architecture

Stack Overview

Backend:

FastAPI — Python web framework for API endpoints
Uvicorn — ASGI server
Discord.py — Discord bot library
Groq API — LLM inference (switched from Claude due to cost)
Supabase — Database + memory storage

Frontend:

HTML/CSS/JavaScript — Dashboard interface
Fetch API — Backend data retrieval

Infrastructure:

NGINX — Reverse proxy (attempted, facing configuration issues)
Cloudflare — DNS + HTTPS proxy
VIP2.HeavenCloud.in — Server hosting
Subdomain: api.ionagency.in

System Architecture Diagram

┌─────────────────┐
│  Discord Server │
│  (ION Agency)   │
└────────┬────────┘
         │
         │ Bot Commands / Messages
         ▼
┌─────────────────┐
│   Discord Bot   │
│   (main.py)     │
└────────┬────────┘
         │
         │ POST /update
         ▼
┌─────────────────┐
│  FastAPI Backend│
│  (backend.py)   │
│  Port 2040      │
└────────┬────────┘
         │
         ├─── Groq API (LLM Inference)
         │
         ├─── Supabase (Memory/Database)
         │
         └─── Dashboard (HTML) via /dashboard endpoint
                    ▲
                    │ HTTPS Fetch Request
                    │
              ┌─────┴──────┐
              │  Browser   │
              └────────────┘

Request Flow (Current Implementation)

User (Discord) 
    → Discord Bot (main.py)
        → FastAPI Backend (backend.py)
            → Groq API (LLM processing)
                → Supabase (memory/data persistence)
                    → Discord Bot (response)
                        → User (Discord)

Dashboard (Browser)
    → HTTPS Fetch → api.ionagency.in (Cloudflare Proxy)
        → vip2.heavencloud.in:2040 (HTTP)
            → FastAPI /dashboard endpoint
                → JSON Response

Key Technical Decisions

Decision	Rationale	Alternatives Considered	Date	Owner
Groq API over Claude	Cost efficiency (Claude expensive, Groq free tier)	Gemini, Claude	2026-04-02	Lexomee
FastAPI for backend	Async support, modern Python, API-first	Django, Flask	2026-04-02	Lexomee
Supabase for memory	Managed database, easy integration, PostgreSQL backend	MongoDB, Custom DB	2026-04-02	Lexomee
Discord as interface	Team already uses Discord, zero friction adoption	Slack, Custom web UI	2026-04-01	Monte_Cristos
Subdomain `api.ionagency.in`	Professional, branded API endpoint	Direct IP access	2026-04-06	Lexomee
Skip NGINX (temporarily)	Configuration issues blocking progress	Caddy, direct uvicorn	2026-04-07	Lexomee + Krush91

Development Timeline

gantt
    title ION AI Management System Development
    dateFormat YYYY-MM-DD
    section Planning
    Initial concept : 2026-03-31, 1d
    section Bot Development
    Test #1-3 (Basic functionality) : 2026-04-02, 1d
    Test #4 (Channel awareness) : 2026-04-03, 1d
    Test #5 (Cross-posting, mentions) : 2026-04-05, 1d
    Test #6 (Progress tracking) : 2026-04-05, 1d
    section Backend
    FastAPI setup : 2026-04-02, 1d
    Supabase integration : 2026-04-02, 1d
    API endpoints : 2026-04-06, 2d
    section Infrastructure
    Cloudflare proxy setup : 2026-04-06, 1d
    NGINX troubleshooting : 2026-04-07, 1d
    section Frontend
    Dashboard HTML/CSS : 2026-04-07, 1d

Bot Capabilities (Testing Evolution)

Test #3 → Test #4 (April 2-3, 2026)

Flaws Identified:

❌ Mentions live channel list every message
❌ Cannot post text to other channels

Fixes Implemented:

✅ Knows all channels passively (no system prompt hardcoding)
✅ Knows current conversation context

Test #4 → Test #5 (April 3-5, 2026)

Additional Issues:

❌ No date/time awareness
❌ Cannot @mention members
❌ Excessive output tokens (verbose responses)
❌ Cannot interpret “right now” or time-related keywords

Test #5 Final Capabilities (April 5, 2026)

Flaws Fixed:

✅ Can send messages to other channels (cross-posting)
✅ Understands difference between channels and project posts
✅ Knows date/time of messages
✅ Interprets temporal keywords (“right now”, “just now”, etc.)
✅ Can @mention employees
✅ Tracks employee activity and prompts for documentation
✅ Creates project board posts automatically
✅ Can query other employees and summarize responses
✅ Concise, optimized responses (token efficiency)

Test #6 Goals (April 5, 2026)

Targeted Improvements:

Accurate work progress reporting for any team member
Enhanced cross-employee communication

Deal Tracking System

Deal Intake Flow (Implemented)

The bot conducts a structured interview when a deal is announced:

User: “we secured a deal lets fucking goooo”

Bot Sequence:

“What is the client’s name?”
“Where is [client] from, or how did they find us?”
“How did we acquire this client?”
“How much is [client] going to pay us?”
“What service is [client] looking to hire us for?”
“When did [client] first contact us?”
“What’s the expected due date?”
“Who is going to handle the work?”
“Is this a one-time deal or monthly retainer?”

Deal Summary Template

───────────────────────────────
✅ CLOSED DEAL SUMMARY
───────────────────────────────
Client: [Name]
Location: [City/Country]
Acquired via: [Method]
Brought in by: [Team Member]
Deal value: [Amount]
Service: [Service Type]
First contact: [Date]
Due date: [Timeline]
Handled by: [Assignee]
Deal type: [One-time/Retainer]
───────────────────────────────

Example Deal (Test Case)

Client: Daniyal
Location: India
Acquired via: Cold call through website
Brought in by: Lexomee
Deal value: 45-50k INR (potentially up to 1 lakh)
Service: Music video production
First contact: April 2, 2026
Due date: Next month (May 2026)
Handled by: TBD (Lexomee or Sidey — both do VFX)
Deal type: One-time

Technical Implementation Details

API Keys & Credentials (REDACTED IN PRODUCTION)

APIs Used:

Discord Bot Token: [REDACTED]
Gemini API (tested, rejected): [REDACTED]
Claude API (tested, too expensive): [REDACTED]
Groq API (production): [REDACTED]

Rationale for Groq:

“claude goated but im gareeb” — Lexomee (budget constraints favored Groq’s free tier)

FastAPI Backend Structure

File: backend.py

Endpoints:

POST /update          # Discord bot message handling
GET  /dashboard       # Dashboard data retrieval
GET  /health          # Server health check

Port: 2040 (non-standard, requires NGINX/proxy for HTTPS)

Deployment Command:

uvicorn backend:app --host 0.0.0.0 --port 2040

Discord Bot Structure

File: main.py

Core Functions:

Message listener (on_message event)
Channel detection
Member mention system
Cross-channel posting
Project board thread creation
API communication (POST to backend)

Supabase Integration

Purpose: Persistent memory storage for bot context

Implementation:

Stores conversation history
Maintains employee activity logs
Tracks deal data
Enables context-aware responses

Example Query:

-- Store employee activity
INSERT INTO employee_activity (name, project, timestamp, notes)
VALUES ('Lexomee', 'AI Management System', NOW(), 'Backend 90% complete');

Infrastructure Challenges & Solutions

Challenge 1: HTTPS to HTTP Communication

Problem:

Dashboard (HTTPS) cannot fetch from backend (HTTP) due to mixed content blocking
Browsers block HTTP requests from HTTPS pages for security

Attempted Solution:

Browser (HTTPS)
    → api.ionagency.in (Cloudflare HTTPS Proxy)
        → vip2.heavencloud.in:2040 (HTTP Backend)

Blocker:

NGINX not configured on server
Cannot listen on port 443 (HTTPS default)
Forced to use port 2040 with HTTP

Workaround:

Temporarily allow mixed content in browser settings
Direct HTTP requests (bypasses NGINX/Cloudflare chain)
Plan to migrate to proper server with NGINX support

Challenge 2: Cloudflare CNAME Limitations

Problem:

Cloudflare CNAME records cannot specify ports
Example: api.ionagency.in → vip2.heavencloud.in:2040 (INVALID)

Solution Required:

NGINX must listen on default HTTPS port 443
Reverse proxy to backend on port 2040
Cloudflare proxies api.ionagency.in → vip2.heavencloud.in:443

Current Status: Bypassing NGINX entirely for testing

Challenge 3: Server Architecture Limitations

Ideal Flow:

Website (HTTPS)
    → Fetch data from api.ionagency.in (HTTPS)
        → Cloudflare proxies to vip2.heavencloud.in (HTTPS)
            → NGINX listens on port 443
                → Forwards to backend (uvicorn) on port 2040
                    → Backend processes request
                        ← Response flows backward

Current Reality:

NGINX unavailable on current server
Cannot configure SSL/TLS termination
Forced to use HTTP on non-standard port

Future Solution (Per Team Discussion):

“bad me fir poora backend + discord bot ek acche server me host krenge” — Lexomee
Translation: “Later we’ll host the entire backend + Discord bot on a proper server”

Dashboard Development

Dashboard Requirements (April 7, 2026)

Lexomee’s Specifications:

Incorporate all features from demo.html
Better UI design (premium aesthetic)
Comprehensive management system panels:
- Employee workload overview
- Active project status
- Client deal pipeline
- Financial metrics
- Task distribution charts

Current Status:

Dashboard HTML file created
Backend integration pending
Fetch API implementation in progress

Frontend Architecture

File: dashboard.html

Fetch Logic:

// On page load, fetch data from backend
fetch('http://vip2.heavencloud.in:2040/dashboard')
  .then(response => response.json())
  .then(data => {
    // Populate dashboard with employee data, deals, tasks
    renderEmployeePanel(data.employees);
    renderDealPipeline(data.deals);
    renderTaskMetrics(data.tasks);
  })
  .catch(error => console.error('Backend fetch failed:', error));

Security Note (Per Krush91’s Feedback):

“secure it with login etc"
"also inspect se bypass na ho aisa banana”
Translation: Add authentication layer + prevent inspect element bypass

Team Collaboration & Dynamics

Developer Velocity

Lexomee’s Work Ethic:

Multiple all-night development sessions (5:00 AM - 7:00 AM active coding)
Rapid iteration (Test #3 → Test #5 in 3 days)
Backend 90% completion in 6 days
Self-directed problem-solving (infrastructure debugging at 5 AM)

Team Recognition:

“Nvm I take it back your boss is strong fuck it we doing it” — Monte_Cristos (April 7, 5:10 AM)

Cross-Team Support

Krush91’s Contributions:

Infrastructure consultation (NGINX, Caddy suggestions)
Security recommendations (authentication, inspect blocking)
CSS/frontend guidance
Late-night problem-solving support (5:00 AM responses)

Wana_be_sidey’s Role:

QA testing observations
VFX work assignment (potential handler for Daniyal music video)
Team morale (concerned about all-night work sessions)

Communication Patterns

Hindi-English Code-Switching:

Technical discussions in English
Troubleshooting in Hindi
Casual banter mixed throughout
Example: “nga abhi backend 90% done h” (Translation: “Yeah, backend is 90% done now”)

Problem-Solving Style:

Live debugging in Discord
Stream-of-consciousness technical explanations
Visual architecture diagrams shared mid-conversation
Immediate peer review and feedback loops

Known Issues & Bugs

Active Bugs (April 7, 2026)

Bug: Screenshot showing error condition (details not transcribed)
- Status: Under investigation
- Priority: Unknown (visual evidence only)

Resolved Issues

Channel List Spam (Test #4)
- Issue: Bot mentioned live channel list in every message
- Fix: Passive channel awareness without explicit listing
No Cross-Channel Posting (Test #4)
- Issue: Bot confined to single channel
- Fix: Multi-channel posting capability implemented
Temporal Awareness (Test #5)
- Issue: No understanding of “right now”, “just now”, dates, times
- Fix: Date/time context injection + temporal keyword parsing
Token Inefficiency (Test #5)
- Issue: Verbose, generic AI responses wasting tokens
- Fix: Concise, context-aware prompting + output optimization
Member Mentions (Test #5)
- Issue: Could not @mention team members
- Fix: Discord mention syntax integration

Deliverables

Completed

In Progress

NGINX reverse proxy configuration
HTTPS/HTTP communication fix
Dashboard UI implementation
Frontend-backend data integration
Authentication system

Pending

Dashboard authentication (login system)
Inspect element bypass protection
Production server migration
Client-side management portal (visual + trustworthy design)
Testing channel for bot development (per Monte_Cristos request)
Code/prompt documentation for team collaboration

Dependencies

tech-discord — Bot platform
tech-fastapi — Backend framework
tech-groq — LLM inference API
tech-supabase — Database + memory
tech-nginx — Reverse proxy (pending configuration)
tech-cloudflare — DNS + HTTPS proxy
tech-uvicorn — ASGI server
process-client-onboarding — CRM workflow automation

Blockers

NGINX Configuration: Cannot listen on port 443 (HTTPS), blocking Cloudflare CNAME integration
Server Limitations: Current VPS lacks SSL/TLS termination capability
Mixed Content Blocking: Browser security prevents HTTPS → HTTP requests
Authentication Not Implemented: Dashboard currently unprotected

Next Actions

Immediate (< 48 hours)

Resolve NGINX configuration (try Caddy as alternative)
Complete dashboard UI design
Implement authentication layer
Add inspect element bypass protection

Short-term (1 week)

Migrate to production server with proper NGINX support
Deploy HTTPS-enabled backend
Create dedicated testing channel for bot development
Document prompts and code for team collaboration
Build client-facing management portal

Medium-term (2-4 weeks)

Implement financial tracking dashboard
Add automated reporting (weekly summaries)
Expand bot capabilities (Test #6 improvements)
Integrate payment tracking
Create employee performance analytics

Technical Debt

Security: No authentication on dashboard endpoints
Infrastructure: Temporary workaround for HTTPS/HTTP (browser settings hack)
Error Handling: Limited error recovery in bot
Documentation: Code/prompts not yet documented for team
Testing: No dedicated testing channel (development in production)

Lessons Learned

What Worked

Groq API Selection: Cost-effective alternative to Claude without sacrificing quality
Discord Interface: Zero friction adoption — team already uses platform daily
Iterative Testing: Test #3 → #5 rapid iteration identified and fixed core issues quickly
Structured Deal Intake: Bot-driven interview ensures consistent CRM data
FastAPI Choice: Async support critical for Discord bot + web dashboard hybrid

What Didn’t Work

NGINX Setup: Infrastructure complexity underestimated, blocked progress
Initial API Choice: Gemini tested and rejected, Claude too expensive
Server Selection: Current VPS lacks NGINX/SSL support, forcing migration plan

Future Improvements

Infrastructure-First Planning: Validate server capabilities before development
Authentication from Day 1: Security should be integrated, not added later
Testing Environment: Dedicated channel for bot development (reduce production noise)
Documentation Culture: Document prompts/code in parallel with development

team-lexomee — Primary developer
team-krush91 — Infrastructure consultant
team-monte-cristos — Project oversight
team-wana-be-sidey — QA testing, VFX assignee
tech-discord — Bot platform
tech-fastapi — Backend framework
tech-groq — LLM API
tech-supabase — Database
tech-nginx — Reverse proxy
tech-cloudflare — DNS + HTTPS
process-client-onboarding — CRM automation

Notes

Development Hours: Extensive all-night sessions (evidence: 5:00 AM - 7:00 AM active development)

Team Culture: High-velocity execution, immediate problem-solving, peer support despite unconventional hours

Budget Constraints: API selection driven by cost (“im gareeb” — chose Groq over Claude)

Technical Ambition: Building production-grade AI automation system with 8 days of development time

Monte_Cristos’ Initial Skepticism Reversed:

Initial directive: “make another channel for testing”
Later: “Nvm I take it back your boss is strong fuck it we doing it” (recognition of execution quality)

Infrastructure Philosophy: “Backend first, frontend later” — prioritize core functionality over UI polish

Future Vision: Dual system (employee management + client management) with visual, trustworthy client-facing portal

Status Summary: Project is in active development with backend 90% complete. Core bot functionality operational (cross-posting, mentions, deal tracking, memory). Infrastructure challenges (NGINX, HTTPS) require server migration. Frontend dashboard in progress. Team velocity high, execution quality recognized. Expected completion: 1-2 weeks pending infrastructure resolution.

This project represents ION’s first major internal automation effort — if successful, establishes template for AI-driven workflow optimization across all agency operations.

WIKI—BOND

Explorer

ION AI Management System (Discord Bot + Dashboard)

ION AI Management System (Discord Bot + Dashboard)

Project Brief

Objective

Scope

In Scope — Employee Management Side

In Scope — Client Management Side

In Scope — Technical Infrastructure

Out of Scope (Current Phase)

Technical Architecture

Stack Overview

System Architecture Diagram

Request Flow (Current Implementation)

Key Technical Decisions

Development Timeline

Bot Capabilities (Testing Evolution)

Test #3 → Test #4 (April 2-3, 2026)

Test #4 → Test #5 (April 3-5, 2026)

Test #5 Final Capabilities (April 5, 2026)

Test #6 Goals (April 5, 2026)

Deal Tracking System

Deal Intake Flow (Implemented)

Deal Summary Template

Example Deal (Test Case)

Technical Implementation Details

API Keys & Credentials (REDACTED IN PRODUCTION)

FastAPI Backend Structure

Discord Bot Structure

Supabase Integration

Infrastructure Challenges & Solutions

Challenge 1: HTTPS to HTTP Communication

Challenge 2: Cloudflare CNAME Limitations

Challenge 3: Server Architecture Limitations

Dashboard Development

Dashboard Requirements (April 7, 2026)

Frontend Architecture

Team Collaboration & Dynamics

Developer Velocity

Cross-Team Support

Communication Patterns

Known Issues & Bugs

Active Bugs (April 7, 2026)

Resolved Issues

Deliverables

Completed

In Progress

Pending

Dependencies

Blockers

Next Actions

Immediate (< 48 hours)

Short-term (1 week)

Medium-term (2-4 weeks)

Technical Debt

Lessons Learned

What Worked

What Didn’t Work

Future Improvements

Related Entities

Notes

Graph View

Table of Contents

Backlinks