RAG API Service

This directory contains the code for the Retrieval-Augmented Generation (RAG) API service.

Running the Server

There are two modes to run the server:

1. Normal Mode (Database Based)

This mode requires a running Spanner database.

make run-history-api

This command uses ./configs/chrome-internal.json by default.

2. In-Memory Mode

This mode loads the index snapshot directly from GCS and serves queries from memory. It does not require a running Spanner database.

make run-history-api-in-memory

By default, it uses the date 2026-04-20. You can override the date by setting the INDEX_DATE environment variable:

INDEX_DATE=2026-04-20 make run-history-api-in-memory

Note: Ensure that your configuration file (e.g., ./configs/demo.json) has the gcs_bucket field set.

Configuration

Configuration files are located in the configs directory. Key parameters include:

• spanner_config: For database mode.
• gcs_bucket: For in-memory mode.
• query_embedding_model: Gemini model for embeddings.
• summary_model: Gemini model for summaries.

Environment Variables

Ensure you have the following environment variables set for Gemini API access:

• GEMINI_API_KEY (if running locally with an API key)
• Or GEMINI_PROJECT and GEMINI_LOCATION if using Google Cloud project credentials.