Gemini RAG Overview & Setup

Gemini RAG turns your documents into a managed knowledge system built on Google Gemini's File Search API. Import files, documentation repositories, or entire websites into isolated File Stores, organize them with categories and metadata, then ask questions over exactly the documents you choose.

The same knowledge base can power private research chats and two independent public experiences: a fast, model-free Website Search and a citation-backed Website Assistant. Search runs against your App's RDBMS without calling Gemini, while Assistant responses stay grounded in the Gemini index and can include citations that lead readers back to the original source.

Crucially, a knowledge base does not need to mirror a single site. Combine saved imports from multiple websites, repositories and file collections in one File Store, then give customers one place to search or ask questions across all of them. Each Assistant and Search deployment can still apply its own metadata scope, while Search ranking controls decide which titles, headings, document types and newer content should win across the combined corpus.

What you can build​

Enable and configure​

Gemini RAG is a built-in AI Chat extension. It installs automatically when AI Chat can resolve both:

1. a Gemini API key; and
2. the App's IDbConnectionFactory, used by OrmLite to persist File Stores, documents, imports, Search indexes and analytics, Assistants, conversations, and citations.

Create an API key in Google AI Studio and add either variable to the App's environment:

GOOGLE_API_KEY=your_api_key
# or
GEMINI_API_KEY=your_api_key

AI Chat resolves GOOGLE_API_KEY, then GEMINI_API_KEY, then the API key on its configured google provider. You can also provide the key programmatically:

services.AddPlugin(new ChatFeature {
    Variables = {
        ["GEMINI_API_KEY"] = context.Configuration["Gemini:ApiKey"]!,
    },
});

The Gemini integration is a native AI Chat extension and talks directly to Gemini's HTTP APIs. No Google client SDK or additional Gemini package is required.

Restart the App after changing its environment. You also need at least one Google Gemini chat model configured in AI Chat; the Gemini model selector only lists compatible Google chat models. Without an API key or database connection, the extension logs why and disables its routes and UI.

To disable it explicitly:

services.AddPlugin(new ChatFeature {
    DisableExtensions = ["gemini"],
});

Upload tuning and MIME type overrides​

The upload worker normally lets Gemini infer a file's type. If Gemini rejects or misidentifies a particular extension, map it explicitly with GEMINI_UPLOAD_MIME_TYPES:

# Comma-separated extension:mime/type pairs; do not include the leading dot
GEMINI_UPLOAD_MIME_TYPES="mdx:text/markdown,cshtml:text/html,ss:text/markdown"

The built-in mapping is mdx:text/markdown,cshtml:text/html. Setting the environment variable replaces that mapping, so retain any defaults you still need. Restart the App after changing it. Use an override only for extensions that need one; forcing the wrong type can reduce indexing and retrieval quality.

Large imports use bounded concurrency and retry transient provider failures with exponential backoff. Their defaults can also be overridden from the App's environment:

GEMINI_UPLOAD_CONCURRENCY=4
GEMINI_UPLOAD_MAX_RETRIES=4

Two model selections can also be overridden. GEMINI_ASSISTANT_MODEL is the default model for published Assistants that don't select their own, and GEMINI_PROBE_MODEL is used by the metadata filter capability probe:

GEMINI_ASSISTANT_MODEL=gemini-flash-latest
GEMINI_PROBE_MODEL=gemini-flash-latest

Write operations can be restricted to a role with GEMINI_WRITE_ROLE.

File Stores​

A File Store is an isolated searchable corpus. It is also the practical ownership and lifecycle boundary: documents that should be searched, administered, and deleted together belong in the same store.

The Gemini home page lists each store with its document count and size. Create a store with a clear, stable name such as docs.example.com, or use the chat icon to immediately query an existing one.

Opening a store exposes four deep-linkable workspaces:

• Explore - browse categories, search documents, apply metadata filters, edit metadata, inspect coverage, monitor uploads, and start grounded chats.
• Import - upload files, synchronize folders, or stage a website crawl.
• Assistants - design, publish, and review website chat Assistants grounded in the store.
• Search - tune and publish model-free website Search widgets backed by the local index.

The selected workspace, Import subsection, Explorer category, saved crawl, Assistant, Search widget, and conversation are preserved in the URL, so a reload or shared link returns to the same view.

Quick start​

1. Open Gemini from the left toolbar and create a File Store.
2. Open Import → Upload files and drop in one or more documents.
3. Optionally choose a destination category and add metadata.
4. Upload the files, then follow View uploads to watch Gemini indexing progress.
5. In Explore, choose New Chat to query the whole store-or apply filters and choose Ask about this.
6. Expand Sources below an answer to inspect the evidence Gemini retrieved.
7. Optionally open Search, test the local index, and publish a standalone Search widget.

That is the smallest useful RAG workflow. The guides above show how to turn it into a repeatable, curated knowledge pipeline.

Storage economics​

Google currently charges for embeddings when documents are indexed, whilst file storage and query-time embeddings are free. Retrieved document tokens are billed as normal model context, and stores persist until deleted, with project capacity scaling by usage tier.

That removes the recurring vector-storage charge common in managed RAG architectures - an organization can maintain durable knowledge bases and pay primarily when content is indexed and when users actually query it. Content-addressed deduplication reinforces this: the same document uploaded twice is stored and indexed once.

Where to next​

Import your first real source with Importing Documents, or read Metadata & Source URLs first if you want retrieval scoping and durable citations in place before the initial import.