Stanford AIxLaw Hackathon

Stanford AI x Law Hackathon (9/8/2024, Stanford Campus)

• Team Name: Segfault Security Solutions
• Team Members: Caleb Ditchfield (github: kryptoklob)
• Repository: Stanford-AIxLaw-Hackathon_Google-Docs-Dispute-Resolution
• Whiteboarding Start Time: 5:51pm PST (end: 6:16pm PST)
• Code Start Time: 6:24pm PST
• Code Deadline: 7:00pm PST

Project: Google Docs Automated Dispute Resolution

[https://docs.google.com/document/d/17lCyWnT9TnhfsiEuNxbsEIYWEhc57eboPlnFx3xerIk/edit?usp=sharing](Example Google Doc)

Built With:

• Langchain
• NodeJS
• Typescript
• Google Docs Integrations (tbd)

---

Everything below this line is the original forked readme - probably no time to update it given that I only have 12 minutes left.

---

install

Prerequisites

This assumes you're on a Unix system. (WSL is fine)

If on Windows...why?

First, make sure you are using Node v18. I recommend using nvm for Node & npm version management.

nvm use 18

Install Packages

cd spearbot-node && npm install

Export OpenAI API Key

Create an API key here if you don't have one.

You must have access to the GPT-4 model to use this tool.

(If you don't know if you have it, you probably don't.)

export OPENAI_API_KEY=<your-api-key-here>

1. Put the files to be audited (supports markdown and solidity) in ./spearbot-node/put_files_to_audit_here.

   Note: Data ingestion via codebase url (ie github codearena link for the contest) will be coming soon.

2. Run the summarizer. This will generate summaries of all markdown & solidity files, and output by default to summarization-results.json

   ./summarize.sh --dir spearbot-node/put_files_to_audit_here
   

   Note: coming shortly will be a tool that will automatically generate markdown documentation of the entire codebase using the output from this step.

3. Run the embedder. This will create a vector database of the summarized code & the code itself. Outputs by default to spearbot-node/vectore/embeddings/

   ./embedder.sh --in summarization-results.json
   

4. Query the code via the analyzer chatbot.

   ./analyzer.sh "What does <somefunction> in <somecontract> do?"
   
   <natural language results>