Skip to main content
This recipe shows you how to build an app that combines multiple Vector Stores for a powerful RAG (Retrieval-Augmented Generation) workflow. Learn to integrate Document Vector Store and GitHub Vector Store with query optimization for comprehensive knowledge base Q&A.

Overview

Purpose: Learn how RAG works by combining multiple Vector Stores.
NodeRoleModel
Document Vector StoreVectorize PDF/text documentstext-embedding-3-small
GitHub Vector StoreVectorize code repositorytext-embedding-3-small
App EntryQuestion input-
Generator x2Optimize search queriesgpt-5-nano
Query Node x2Search for relevant information-
GeneratorGenerate answer from search resultsclaude-haiku-4.5
What you’ll learn: RAG (Retrieval-Augmented Generation), Document Vector Store, GitHub Vector Store, query optimization, integrating multiple Query nodes

Workflow Diagram

Node Configuration Details

1. Document Vector Store

ItemValue
Node NameDocument Knowledge Base
Embedding ModelOpenAI text-embedding-3-small
Pre-loaded Files:
FileSource URLContent
docs.giselles.ai-llms-full.txthttps://docs.giselles.ai/llms-full.txtDocumentation (Text)
giselles-terms.pdfhttps://giselles.ai/legal/termsTerms of Service
giselles-privacy.pdfhttps://giselles.ai/legal/privacyPrivacy Policy

2. GitHub Vector Store

ItemValue
Node NameGitHub Knowledge Base
Repositorygiselles-ai/giselle
Include CodeON
Include IssuesON
Include Pull RequestsON
Embedding ModelOpenAI text-embedding-3-small

3. App Entry

ItemValue
Node NameQuestion Input
Node Input Namequestion
Prompt- (User input only)

4. Generator Node (Doc Query)

ItemValue
Node NameDoc Query Generator
Modelgpt-5-nano
ThinkingOFF
Web Search-
Prompt:
Convert the following question into an optimized search query for documentation:

Question: @Question Input

Generate a query with relevant keywords and concepts for searching documentation.
Output only the query, nothing else.

5. Generator Node (Code Query)

ItemValue
Node NameCode Query Generator
Modelgpt-5-nano
ThinkingOFF
Web Search-
Prompt:
Convert the following question into an optimized search query for code:

Question: @Question Input

Generate a query with function names, file patterns, and technical terms for searching code.
Output only the query, nothing else.

6. Query Node (Document)

ItemValue
Node NameDocument Search
Source@Document Knowledge Base
Query@Doc Query Generator
Max Results10
Similarity Threshold0.3

7. Query Node (GitHub)

ItemValue
Node NameCode Search
Source@GitHub Knowledge Base
Query@Code Query Generator
Max Results10
Similarity Threshold0.3

8. Generator Node (Answer)

ItemValue
Node NameAnswer Generator
Modelclaude-haiku-4.5
ThinkingON
Web SearchOFF
Prompt:
Answer the following question based on the provided content:

Question: @Question Input

Documentation:
@Document Search

Code & Repository:
@Code Search

Requirements:
- Answer based on both documentation and code repository content
- Reference specific files or passages when applicable
- If the answer is not found, clearly state that
- Be concise but thorough