CloseIQ

AI-Powered Real-Time Sales Coaching Platform

CloseIQ is an intelligent sales copilot that listens to live sales conversations, analyzes prospect behavior, retrieves relevant knowledge from uploaded documents, and provides real-time coaching recommendations to help sales representatives close more deals.

The platform combines:

• Real-time transcription
• Retrieval-Augmented Generation (RAG)
• Deal-state tracking
• Prospect avatar profiling
• Close probability prediction
• AI-powered coaching suggestions

---

Features

Real-Time Sales Intelligence

• Live conversation analysis
• Prospect behavior detection
• Deal-stage tracking
• Objection identification
• Close probability estimation
• Next-action recommendations

AI Coaching Engine

CloseIQ generates:

• What To Say
• Why It Works
• Next Move
• Close Probability

during ongoing conversations.

Knowledge-Based Coaching

Upload sales playbooks, scripts, SOPs, objection-handling guides, and training materials.

The platform:

1. Processes documents
2. Creates vector embeddings
3. Stores them in ChromaDB
4. Retrieves relevant context during conversations
5. Grounds AI recommendations using company knowledge

Prospect Avatar Intelligence

The system continuously builds prospect profiles by analyzing:

• Communication style
• Buying signals
• Objection patterns
• Engagement indicators
• Decision-making tendencies

Document Management

Supported file formats:

• PDF
• TXT
• DOCX

Features:

• Upload documents
• Automatic text extraction
• Chunking and embedding generation
• Vector storage
• Metadata tracking
• Secure deletion with vector cleanup

---

Architecture

┌───────────────────────────┐
│   React / Electron UI     │
│ Real-Time Transcription   │
└─────────────┬─────────────┘
              │
              ▼
┌───────────────────────────┐
│      FastAPI Backend      │
└─────────────┬─────────────┘
              │
      ┌───────┴────────┐
      │                │
      ▼                ▼

┌──────────────┐  ┌──────────────┐
│  RAG Engine  │  │ Deal Engine  │
└──────┬───────┘  └──────┬───────┘
       │                 │
       ▼                 ▼

┌──────────────┐  ┌──────────────┐
│  ChromaDB    │  │ Intelligence │
│ Vector Store │  │   Modules    │
└──────────────┘  └──────────────┘

              │
              ▼

       OpenRouter LLM

---

Repository Structure

backend/
├── app/
│   ├── core/
│   ├── models/
│   ├── routers/
│   ├── services/
│   │   ├── deal_engine/
│   │   └── intelligence/
│   └── main.py
│
├── file_metadata/
├── user_files/
├── vector_store/
└── requirements.txt

realtime-transcription-app/
├── public/
├── src/
└── package.json

---

Core Components

RAG Engine

Retrieval-Augmented Generation powers the coaching system.

Workflow:

Uploaded Documents
        ↓
Text Extraction
        ↓
Chunking
        ↓
Embeddings
        ↓
ChromaDB
        ↓
Similarity Search
        ↓
LLM Context Injection
        ↓
Sales Recommendation

---

Deal Engine

The deal engine maintains conversation state across the sales call.

Tracks:

• Deal stage
• Objection level
• Payment discussions
• Prospect engagement
• Session memory

Modules:

deal_engine/
├── perception.py
├── decision.py
├── session_store.py
└── state.py

---

Prospect Intelligence Layer

Advanced behavioral analysis modules.

Avatar Profiling

Creates dynamic prospect personas.

Avatar Signals

Extracts:

• Buying intent
• Trust indicators
• Resistance patterns
• Decision triggers

Close Probability

Predicts likelihood of closing a deal based on:

• Transcript signals
• Deal progression
• Prospect engagement
• Objection severity

---

Technology Stack

Backend

• FastAPI
• LangChain
• LangGraph
• ChromaDB
• OpenRouter
• OpenAI-Compatible APIs
• Pydantic
• Uvicorn

AI & Machine Learning

• GPT Models via OpenRouter
• OpenAI Embeddings
• Vector Search
• Retrieval-Augmented Generation (RAG)

Frontend

• React
• Electron
• TailwindCSS

Authentication

• Supabase Auth

Real-Time Services

• Deepgram Speech-to-Text

Deployment

• Docker
• Docker Compose

---

API Endpoints

Health Check

GET /

Response:

{
  "status": "ok"
}

---

Upload Knowledge Document

POST /upload

Form Data:

file
user_id

---

List User Files

GET /user/{user_id}/files

---

Delete File

DELETE /user/{user_id}/files/{filename}

Deletes:

• Physical file
• Metadata
• Chroma vectors

---

Generate Sales Coaching

POST /query

Example Request:

{
  "user_id": "sales_user",
  "conversation_transcript": "...",
  "prospect_transcript": "...",
  "closer_transcript": "..."
}

Example Response:

{
  "what_to_say": "...",
  "why_it_works": "...",
  "next_move": "...",
  "deal_stage": 70,
  "close_probability": 0.82,
  "sources": []
}

---

Installation

Clone Repository

git clone https://github.com/yourusername/closeiq.git
cd closeiq

---

Backend Setup

cd backend

python -m venv .venv

# Windows
.venv\Scripts\activate

# Linux/Mac
source .venv/bin/activate

pip install -r requirements.txt

---

Configure Environment Variables

Create:

backend/.env

Example:

OPENROUTER_API_KEY=your_key

OPENAI_API_BASE=https://openrouter.ai/api/v1

LLM_MODEL=openai/gpt-4o-mini

EMBEDDING_MODEL=text-embedding-3-small

SUPABASE_URL=your_supabase_url

SUPABASE_SERVICE_KEY=your_supabase_service_key

---

Run Backend

uvicorn app.main:app --reload

Backend:

http://localhost:8000

---

Frontend Setup

cd realtime-transcription-app

npm install

Start:

npm start

---

Docker Deployment

Build and run:

docker compose up --build

Services:

Service	Port
Frontend	8080
Backend	8000

---

External Services

OpenRouter

Used for:

• LLM responses
• Embeddings

Website:

https://openrouter.ai

---

Deepgram

Used for:

• Real-time transcription

Website:

https://deepgram.com

---

Supabase

Used for:

• Authentication

Website:

https://supabase.com

---

Future Roadmap

Phase 1

• Real-time coaching
• RAG integration
• File knowledge base

Phase 2

• CRM integrations
• Salesforce connector
• HubSpot connector

Phase 3

• Multi-user teams
• Analytics dashboard
• Call scoring system

Phase 4

• Mobile applications
• React Native deployment
• Push notifications

Phase 5

• Fine-tuned sales models
• Organization-wide knowledge graphs
• Enterprise deployment

---

Security Considerations

• API keys stored server-side
• Environment-based configuration
• Vector store isolation per user
• Metadata tracking
• Secure document deletion
• No provider keys exposed to clients

---

License

This project was developed as part of a Mobile Application Development project and serves as a foundation for real-time AI-powered sales enablement systems.

Feel free to fork, modify, and extend for educational or research purposes.

---

Documentation

Additional project documentation:

• SERVICE_SETUP.md
• CLOSEIQ_Mobile_Project_Article.md
• CLOSEIQ_Mobile_ReactNative_Deployment.md

---

CloseIQ — Helping Sales Teams Close Smarter, Faster, and More Consistently.