llama-rag

This is a university project for use a RAG middleware between Home Assisant and llama.

Install dependecies

yarn install

Run with memory vector store

yarn start

For select the RAG data set check:

localhost:3000/rag

Use m1.txt for task generator model
Use m2.txt for syntax generator model

Performance testing

• First select the datasets for models
• Run pre-defined test for performance metrics.

yarn test

Use --verbose attribute for log the response of failed test runs.

Create new test case

Create a new test data files in test/data folder.
Export the userPrompt, systemPrompt and response variables.
Import them in test/runner.js and run with await runTest('Test-Name', systemPrompt, userPrompt, response, verboseError);