FinBot

Cornell Fintech Club Spring '25 Project

About

We built a chatbot that answers questions about institutional investor behavior by analyzing SEC Form 13F filings. These filings, submitted quarterly by investment managers with over $100 million in assets, disclose their holdings in U.S. exchange-traded securities, including position sizes, market values, and asset types.

Our system uses a database of 13F filings data, a LangChain SQL Agent to interpret natural language queries, and a Streamlit frontend to deliver insights. Users can ask questions in plain English and receive data-driven answers about institutional trading trends.

Example questions it can answer:

• "Which top 5 firms increased their Tesla holdings the most over the last two quarters?"
• "Has BlackRock shifted from financials to healthcare in the past year?"
• "What new positions did Berkshire Hathaway add this quarter?"

This information can help investors understand large firms' strategies. For example, if a major firm suddenly puts a lot more money into healthcare, it could indicate that they expect that sector to grow, maybe by anticipating policy changes in that area. This can inform your own portfolio strategy. It can also give you insights into the risk exposure of the firm itself.

Contributors

TPM: Niti Goyal
PM: Kelly Lambert
APM: Samuel Hecht
SWE: Jane Tenecota-Villa
SWE: Adeeb Khan
SWE: Noah Lingo
SWE: Saanvi Jain

FinBot Local Development Setup

This guide walks you through:

1. Cloning the repo
2. Creating a Python virtual environment
3. Installing dependencies
4. Setting up PostgreSQL locally
5. Populating the database
6. Running tests

---

1. Clone the repository

git clone git@github.com:niti-go/FinBot.git
cd FinBot

---

2. Create and activate a Python virtual environment

python3 -m venv venv
# macOS/Linux:
source venv/bin/activate
# Windows PowerShell:
# .\\venv\\Scripts\\activate

---

3. Install Python dependencies

Install requirements.txt with:

pip install -r requirements.txt

---

4. Set up PostgreSQL locally

A. Starting PostgreSQL server manually

• macOS (Homebrew):

  brew services start postgresql

• Linux (systemd):

  sudo systemctl start postgresql

• Windows (pgAdmin4 bundled service):

  1. Open pgAdmin 4.
  2. In the Browser panel, right-click Servers → Create → Server... and configure your local server connection if not present.

B. Creating the database via pgAdmin 4

1. In pgAdmin's Browser, expand your server, right-click Databases → Create → Database...
2. In the General tab:
   • Database: finbot
   • Owner: your user (e.g. postgres)
3. In the Definition tab:
   • Template: template0
4. Click Save.

C. Applying the schema

1. Right-click the new finbot database → Query Tool.
2. Open or paste the contents of init_db.sql into the editor.
3. Click Execute/Refresh (▶️) to run the script and create tables.

Verify tables via the Query Tool:

\dt
-- should list: investmentmanagers, filings, securities, holdings

---

5. Configure environment variables

Create a .env file:

DB_NAME=FinBot (OR WTVR YOU NAMED YOUR DB)
DB_USER=postgres (OR YOUR SERVER USERNAME
DB_PASSWORD=YOURPASSWORD
DB_HOST=localhost
DB_PORT=5432

---

6. Populate the database

# Full run (7.6k CIKs) - may take 1-2 hours
python populate_db.py

# Quick run (first 10 CIKs)
python populate_db.py --limit 10

---

7. Running tests

python Tests/test_insert.py
python Tests/test_get_filings.py

---

8. Inspect data

In pgAdmin4, view tables by clicking through Schemas --> Tables --> Right click desired table --> View/Edit Data --> All Rows

Or through query tool:

SELECT COUNT(*) FROM InvestmentManagers;
SELECT COUNT(*) FROM Filings;
SELECT COUNT(*) FROM Securities;
SELECT COUNT(*) FROM Holdings;