📊 ConcallIQ — Earnings Call Intelligence Platform

Ask questions about any earnings call transcript using RAG + LLM. Get live news and StockTwits sentiment. Built for analysts, by a developer.

Features • Demo • Tech Stack • Setup • Usage

---

🚀 What is ConcallIQ?

ConcallIQ is an AI-powered earnings call analysis tool that lets you:

• 📄 Upload any concall PDF and instantly ask questions about it
• 🤖 Get precise, cited answers grounded in the actual transcript — no hallucinations
• 📰 Fetch live news sentiment about any stock using NewsData.io + LLM scoring
• 📈 Analyze retail investor sentiment from StockTwits (no API key needed)
• 📊 Compare multiple companies side by side

This is the kind of tool used by hedge funds and trading firms — built from scratch with open-source tools and free APIs.

---

✨ Features

Feature	Description	Tech
📤 PDF Upload & Index	Upload concall transcripts, auto-chunked and embedded	PyPDF2 + pdfplumber
💬 RAG Q&A	Ask natural language questions, get cited answers	LangChain + ChromaDB + Groq
📋 Auto Summary	One-click comprehensive concall summary	LLaMA3 70B
📰 News Sentiment	Live news fetched and scored -1 to +1	NewsData.io + LLM
📈 StockTwits Sentiment	Retail investor bullish/bearish sentiment	StockTwits API (Free)
📊 Multi-Company Compare	Ask same question across multiple concalls	Multi-doc RAG
📈 Sentiment Charts	Visual gauge charts for sentiment scores	Plotly

---

🏗️ How It Works

User uploads Concall PDF/Transcript
            ↓
RAG chunks + indexes it (ChromaDB + Local Embeddings)
            ↓
User asks → "What did CEO say about margins?"
            ↓
RAG retrieves relevant chunks → LLM answers with citations
            +
NewsData.io fetches latest Indian & global news sentiment
            +
StockTwits retail investor sentiment (bullish/bearish)
            ↓
Final Answer = Concall Insight + Market Sentiment

---

🛠️ Tech Stack

Frontend          →  Streamlit
RAG Framework     →  LangChain
Vector Store      →  ChromaDB (local, free)
LLM               →  Groq (LLaMA3 70B / Mixtral) — Free tier
Embeddings        →  all-MiniLM-L6-v2 (local, free)
PDF Parser        →  pdfplumber / PyPDF2
News Sentiment    →  NewsData.io + Groq LLM
Social Sentiment  →  StockTwits — Free, no API key needed
Visualization     →  Plotly

---

📁 Project Structure

ConcallIQ/
│
├── src/
│   ├── ingestion/
│   │   ├── pdf_loader.py       # Load & parse concall PDFs
│   │   └── chunker.py          # Smart text chunking with overlap
│   │
│   ├── rag/
│   │   ├── embeddings.py       # Local HuggingFace embeddings
│   │   ├── vector_store.py     # ChromaDB setup & retrieval
│   │   └── retriever.py        # Full RAG pipeline
│   │
│   ├── sentiment/
│   │   ├── news_sentiment.py   # NewsData.io + LLM scoring
│   │   └── reddit_sentiment.py # StockTwits retail sentiment
│   │
│   └── utils/
│       ├── logger.py
│       └── exception.py
│
├── data/concalls/              # Store uploaded PDFs
├── app.py                      # Streamlit UI (4 tabs)
├── requirements.txt
├── .env                        # API keys (never commit!)
└── README.md

---

⚡ Quick Start

1. Clone the repo

git clone https://github.com/YOUR_USERNAME/ConcallIQ.git
cd ConcallIQ

2. Create virtual environment

python -m venv venv
source venv/bin/activate        # macOS/Linux
venv\Scripts\activate           # Windows

3. Install dependencies

pip install -r requirements.txt

4. Set up API keys

Create a .env file:

OPENAI_API_KEY=your_groq_key_here           # Get free at console.groq.com
NEWSDATA_API_KEY=your_newsdata_key_here     # Get free at newsdata.io
OPENAI_BASE_URL=https://api.groq.com/openai/v1

# StockTwits — no API key needed!

5. Run the app

streamlit run app.py

Open http://localhost:8501 🎉

---

🔑 API Keys (All Free!)

Service	Cost	Link	Used For
Groq	✅ Free	console.groq.com	LLM inference
NewsData.io	✅ Free (200 req/day)	newsdata.io	Live Indian & global news
StockTwits	✅ Free (no key needed)	stocktwits.com	Retail investor sentiment
ChromaDB	✅ Free (local)	Built-in	Vector storage
HuggingFace Embeddings	✅ Free (local)	Built-in	Text embeddings

Total cost to run: $0 🎉

---

📖 Usage Guide

Upload & Index a Concall

1. Go to Upload & Index tab
2. Upload a concall PDF (BSE/NSE filings, investor relations pages)
3. Enter the company name (e.g. Infosys)
4. Click 🚀 Index Concall — done in ~30 seconds

Ask Questions

1. Go to Q&A tab
2. Select company or ask across all
3. Use preset questions or type your own:
   • "What was Q3 revenue growth?"
   • "What did the CEO say about AI investments?"
   • "What are the key risks for next quarter?"

Get Sentiment

1. Go to Sentiment tab
2. Type any company/stock name or NSE ticker (e.g. Netweb or NETWEB)
3. Get live news + StockTwits retail sentiment scores with visual gauges

Where to find Concall PDFs

• BSE India → bseindia.com → Search company → Announcements
• NSE → nseindia.com → Company page → Transcripts
• Company investor relations pages

---

🤖 Supported LLM Models (via Groq — Free)

Model	Speed	Best For
llama3-8b-8192	⚡ Fastest	Quick Q&A
llama3-70b-8192	🎯 Best quality	Detailed analysis
mixtral-8x7b-32768	⚡ Fast	Long transcripts (32K context)
gemma2-9b-it	⚡ Fast	Balanced performance

---

🙋 FAQ

Q: Does it work with Hindi/regional language concalls?
A: It works best with English transcripts. Mixed language may reduce accuracy.

Q: How many PDFs can I index?
A: Unlimited — ChromaDB is local and only limited by your disk space.

Q: Is my data sent to any server?
A: PDF text is sent to Groq for LLM inference. ChromaDB and embeddings run 100% locally.

Q: Can I use OpenAI instead of Groq?
A: Yes — set OPENAI_BASE_URL to https://api.openai.com/v1 and use gpt-4o as the model.

---

⚠️ Disclaimer

This tool is for educational and research purposes only.
It is not financial advice. Always do your own research before making investment decisions.

---

Built with ❤️ using Streamlit, LangChain, and Groq