fire-bot

Description:

Prompt queryable database for departmental guidlines.

---

Feature Ideas

Semantic Search

• Return relevant info for a prompt using the OP stack
• Vector queryable datebase with semantic search
• Use the sentence transformer library to generate embeddings
• Use pinecone database library to store and search embeddings
• Insert top k responses from query as context for gpt4 model
• Extract relevant information using chat completion
• return top result and gpt response

Fine tuning

• Fine-tune GPT davinci model on policy manuels
• Standard prompting to ask questions on fine-tuned models

Future tasks

UI/Website

• React native application
• Backend Flask server
• Integrate React and Flask
• Use render for web hosting