https://github.com/prisha03/swot-rag-pipeline
A Streamlit app that turns transcripts into boardroom-ready SWOTs and forecasts, using Ollama LLMs and real financial data.
https://github.com/prisha03/swot-rag-pipeline
arimax financial-data forecasting llm ollama rag streamlit swot-analysis transcript-analysis
Last synced: about 1 month ago
JSON representation
A Streamlit app that turns transcripts into boardroom-ready SWOTs and forecasts, using Ollama LLMs and real financial data.
- Host: GitHub
- URL: https://github.com/prisha03/swot-rag-pipeline
- Owner: prisha03
- Created: 2025-06-18T23:22:22.000Z (4 months ago)
- Default Branch: main
- Last Pushed: 2025-06-19T00:06:02.000Z (4 months ago)
- Last Synced: 2025-06-19T00:26:40.819Z (4 months ago)
- Topics: arimax, financial-data, forecasting, llm, ollama, rag, streamlit, swot-analysis, transcript-analysis
- Language: Python
- Homepage:
- Size: 7.71 MB
- Stars: 0
- Watchers: 0
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# Competitor Intelligence Suite
This project started as a capstone, but evolved into something closer to a product.
The **Competitor Intelligence Suite** is a complete analytics tool that reads through corporate transcripts, pulls official financial data, monitors stock trends, and delivers both strategic insights and forward-looking forecasts. It’s designed for analysts and decision-makers who want to move from raw data to real answers — fast.
I built it to mirror what an in-house analytics product at a consulting firm or investment team might do, without needing multiple tools or licenses.
## What it does
- **Turns unstructured transcripts into boardroom-ready SWOT insights**
- **Connects narrative data to financial performance** (via SEC and Yahoo Finance)
- **Uses statistical forecasting (ARIMAX) to predict next-quarter returns**
- **Builds downloadable reports (PDF & CSV) for instant sharing**
- All through a **clean Streamlit interface** that works offline
## Why it matters
Companies say a lot in their earnings calls. Some of it is noise.
But some of it foreshadows movement — in performance, in stock price, in strategy.
This app doesn’t just summarise what was said. It:
- Measures sentiment over time
- Tracks mentions of key business units (buzz)
- Tags risks, opportunities, and strategic moves using RAG + LLMs
- Checks which variables (text, sentiment, financials) correlate with actual stock movement
Then it models the signal and forecasts forward.
In other words, **it connects the story to the stock.**
## How it works (in plain English)
1. **You upload a transcript** (PDF or TXT)
2. The system breaks it into readable chunks
3. It asks smart questions like “What are the strengths?” and uses a local AI model to answer them
4. It finds supporting quotes, so you're never guessing where the insight came from
5. It saves everything in a polished PDF
6. Meanwhile, it also pulls the company’s financials (from the SEC), stock history, and text-based sentiment
7. It tests what variables might actually influence returns
8. Finally, it builds a predictive ARIMAX model and shows a forecast
## Built with
- **LLM + Retrieval**: SentenceTransformers, FAISS, Ollama (LLaMA 3)
- **Forecasting**: statsmodels, ARIMAX, Granger causality
- **Data ingestion**: SEC API, Yahoo Finance, custom transcript parsers
- **Interface**: Streamlit
- **Reporting**: FPDF for downloadable PDFs
## Example use case
Imagine uploading `UNH_Q1_2024.pdf`
→ You get:
- 3–4 focused SWOT bullets per category, backed by exact quotes
- A clean PDF you could walk into a client meeting with
- A sentiment/buzz trend showing concern or confidence over time
- Financials pulled straight from SEC filings
- A predictive chart showing the next 4 quarters of expected stock movement
It’s not just analysis. It’s explanation, direction, and justification — in one tool.
## How to run it
### Prerequisites
- Python 3.9+
- [Ollama](https://ollama.com) installed locally (to run LLaMA 3 offline)
- Internet for financial data APIs
### Setup
```bash
git clone https://github.com/your-username/competitor-intelligence-suite.git
cd competitor-intelligence-suite
python3 -m venv venv
source venv/bin/activate
pip install -r requirements.txt