FinAI - React / Express / NodeLlamaCPP Financial App (get financial predictions based on a LLM)

A Web App built with React as a Frontend and Express as a Backend. It uses Node LlamaCPP as a way to run an LLM locally and get financial predictions based on data from different APIs. Data store: MongoDB.

Frontend: React (Vite + Typescript)

Backend: Express + MongoDB + Node LLAMA CPP LLM

Getting Started:

You need to setup the LLM you want to be using in the backend. The LLM should be saved as a .gguf file in the backend/models folder.

Validate LLM:
npx --no node-llama-cpp chat --model PATH-TO-MODEL-DIR npx --no node-llama-cpp chat --model c:/projects/finai/backend/models/codellama-13b.Q3_K_M.gguf
Or do a test request:

curl --location 'localhost:9000/api/llm' \
--header 'Content-Type: application/json' \
--data '{ "messages": "Hello there" }'

Or call the express API endpoint localhost:9000/api/llm to see the result

Some examples for models and formats: LLMTypeDefinitions.json

Then you can run the below commands from the FinAI (main) directory and start the project:

npm i
npm start

This installs and starts both the FE and BE using the npm tool 'concurrently'. Or you can run the commands separately in the frontend / backend folders to have them running in separate instances/terminals.

Built With:

[✔] React (Vite, Typescript)
[✔] Express API
[✔] Node LlamaCPP
[✔] TradingView API / Widgets
[✔] Axios
[✔] MongoDB

Features:

Free Chat / Text Inputs
Analysis Prompts - specific prompts for analyzing financial data
Price / Ticker Inputs - a structured way of serving financial data to the LLM
Data Visualisation for Stocks / Crypto - Charts / Diagrams / Tickers with live price updates from the TradingView API

Not implemented yet / In Progress:

Auth0 Auth and User Management
LLM Fine-Tuning and general Model-related options

Based on the Node Llama CPP Runtime/Library:

https://github.com/withcatai/node-llama-cpp

ivaaak/FinAI