Repository files navigation

Developer sample written in Angular demonstrating how developers might ingest developer docs and make the content accessible to Gemini via RAG (Retrieval Augmented Generation). The ingested content is accessible in the sample through an Angular chatbot.

This project was generated withAngular CLI and usesDocsAgent, to impliment RAG (Retrieval Augmented Generation) with Gemini and create a domain-specific expertise chatbot. This sample uses Googles Semantic Revtriever API and Generative Language APIs as well as an AQA (Attributed Questions and Answer) model with Gemini Pro.

For more information on Angular, visitangular.dev.

1. Create a personal fork of the project on GitHub, then clone the fork on your local machine.
2. Runnpm run i to install the dependencies required to run the server.
3. [IMPORTANT!!] This demo needs a Gemini API to run. Go toGoogle AI Studio to get an API key then add it to the Firebase Function infunctions/.env. This demo simulates how you might store and protect a private Gemini API key in a real world app.
4. [IMPORTANT!!] This demo relies on aCORPUS_NAME from Docs Agent, then authenticates with aservice_account_key.json. SeeDocsAgent Set Up guide to set up your own corpus and authentication. You'll need to replacefunctions/service_account_key.json with the one provided to you by Google Cloud, and then make sure to march theCORPUS_NAME infunctions/ to your uploaded corpus id.
5. Runng run angular-chatbot:serve to run the server. Since we're using Firebase Functions, you'll need to run our functions and the app in a Firebase Emulator, this command does this automatically!
6. Open a browser tab tohttp://localhost:4200. The app will automatically reload if you change any of the source files.