📊 PSX AI Assistance — AI-Powered Financial Report & Market Analysis

PSX AI Assistance is an intelligent financial analytics app that integrates Pakistan Stock Exchange (PSX) data, company fundamentals, and financial reports with LLM-powered summarization, embedding-based retrieval (RAG), and interactive visualizations.

Built with Streamlit, Ollama, and ChromaDB, it enables users to:

• Fetch real-time market and company data from PSX APIs
• Download and process company financial PDFs
• Extract tables and text for semantic understanding
• Summarize complex financial data into human-readable insights
• Query embedded knowledge for interactive AI-assisted analysis

---

1112.mov

🚀 Features

✅ Real-Time Market Data
Fetches and displays live data for indices (KSE100, KSE30, PSXDIV20, ALLSHR)

✅ Company Financial Details
Retrieves fundamentals, dividends, and K-Line data from PSX APIs

✅ PDF Financial Reports Parsing
Automatically downloads recent company PDFs from PSX
Extracts text and tables using pdfplumber

✅ AI Summarization
Summarizes structured and unstructured financial data using Ollama models
Supports multiple LLM backends (Qwen, DeepSeek, Kimi, GPT-OSS)

✅ RAG-Enabled Conversational Assistant
Uses ChromaDB for local vector storage of summaries
Users can chat with embedded financial knowledge

✅ Data Visualization
Interactive candlestick charts and metrics using Plotly and Streamlit

---

🧠 System Architecture

📊 PSX API + 📄 PDF Reports
  │
  ▼
Data Collection Layer
(Requests, BeautifulSoup)
  │
  ▼
PDF Parsing & Structuring
(pdfplumber, table extraction)
  │
  ▼
LLM Summarization (Ollama)
Functions used:

• summarize_table()
• summarize_ticker_detail()
• summary_ai_assistance()
    │
    ▼
  Embedding (ChromaDB + embeddinggemma)
    │
    ▼
  RAG Query & AI Response
  Functions used:
• ai_assistance()
• embedd_prompt()
    │
    ▼
  Streamlit UI
  Modules:
• charts, metrics, chat interface

---

🛠️ Tech Stack

Category	Technology
Frontend	Streamlit, Plotly
Backend	Python 3.10+, Requests, BeautifulSoup, asyncio
AI Models	Ollama (Qwen, DeepSeek, Kimi, GPT-OSS), embeddinggemma
Vector DB	ChromaDB
PDF Parsing	pdfplumber, PyPDF
Data Sources	PSX Terminal API, DPS PSX

---

⚙️ Installation

1️⃣ Clone the repository

git clone https://github.com/<your-username>/psx-ai-assistance.git
cd psx-ai-assistance

Create and activate a virtual environment

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

Run the Streamlit app

streamlit run main.py

🧩 Configuration

You may need to configure Ollama or adjust models: response = ollama.chat( model="qwen3-coder:latest", # or "deepseek-v3.1:671b-cloud" messages=[...] )

ChromaDB persistent collection is initialized automatically:

client = chromadb.PersistentClient() collection = client.get_or_create_collection("my_db")

🧾 Core Functions Overview

Function	Purpose
makeMarketRequest()	Fetch index data (KSE100, KSE30, etc.)
makeApiRequest(ticker)	Fetch company fundamentals, dividends, and K-Line data
getPdfFiles(ticker, session_data)	Download and parse company PDFs
summarize_ticker_detail()	Summarize financial reports using LLMs
embedd_text()	Store summaries in ChromaDB
embedd_prompt()	Retrieve context embeddings and answer user queries
ai_assistance()	LLM reasoning based on embedded context
root()	Streamlit entry point that loads both data columns and UI

🧠 Example Workflow

1. Select a symbol (e.g., OGDC) in the sidebar
2. App fetches fundamentals, K-Line, and dividends
3. Downloads the company’s latest financial PDFs
4. Extracts tables and text, summarizes, and embeds them
5. Ask natural language questions like: “What is OGDC’s profit growth over the last year?” “Summarize the company’s recent dividend trends.”
6. The app answers using embedded summaries and AI reasoning

🖼️ UI Preview