RAG Quickstart : with default settings
Getting you RAG application built in under 30 seconds.
1. Create a New RAG Application
From the RAG Applications dashboard, click on the Create RAG App button.
You will see two options:
Create with Default Settings — Recommended for your first app. This option auto-applies standard ingestion, chunking, and retrieval defaults so you can focus only on naming your app, selecting an LLM, and providing data.
Create with Custom Settings — Gives you full control over ingestion, chunking, vectorization, and retrieval. Best for advanced use cases.
For now, select Create with Default Settings.
2. Name Your Application
Provide a name for your RAG application. The name must use only underscores (_) and alphanumeric characters.
Tip: Use descriptive names like policy_app_v1 or review_retrieval_rag so that it’s easier to manage multiple RAG applications later.
3. Select Your LLM
Next, select the Large Language Model (LLM) your application will use. If you haven’t configured an LLM yet, click the LLM configuration link (+) to set one up.
Supported providers include:
OpenAI (e.g., GPT-4o, GPT-4.1, GPT-3.5 Turbo)
Azure OpenAI Service
Amazon Bedrock
Hosted models such as LLaMA
Gemini (Google) and others depending on your deployment
Each LLM can be configured with parameters such as:
Max Tokens — upper limit of generated text
Temperature — controls creativity (lower = factual, higher = creative)
Response Length — caps reply size
Presence/Frequency penalties — reduce repetition
Once configured, select your preferred LLM.
4. Add Your Data Source
The next step is to provide the data that the RAG app will index and retrieve from. You have three main options:
Pick from Data Storage — Use already connected storage like AWS S3, GCS, or internal databases. Supported file formats: PDF, Word, PowerPoint, Excel, Markdown, Plain Text.
Upload from Computer — Drag and drop or select files from your local machine. Size limit: 5MB per file. Source limit: 15 files.
Links and Web Crawling — Provide up to 3 URLs to crawl. This feature extracts text from public or authenticated web pages (useful for policy docs, FAQs, or intranet knowledge).
Tip: Start with a few representative documents to validate your workflow before uploading large volumes.
5. Review and Create
Once your data source and LLM are configured, click Create App.
Your app will now appear in the RAG Applications dashboard with its name, creator, and status indicators for Ingestion, Vectorization, and Retrieval.
Ingestion: Whether your data was successfully pulled in
Vectorization: Whether embeddings were created
Retrieval: Whether queries can be run successfully
If any step has not started, you can click View Details to re-run ingestion or fix issues.
Example Use Case: Policy Document Retrieval
Imagine you’re building a policy assistant for employees. By uploading company policy PDFs into your RAG application:
The LLM (for example, GPT-4o) handles natural language understanding.
The retrieval pipeline fetches only relevant sections (for example, “What’s the leave policy for probation employees?”).
Employees get precise, contextual answers instead of having to skim long PDFs.
This use case highlights how quickly RAG can turn static documents into searchable, conversational knowledge bases.
Key Notes
Default settings handle chunking, embedding, and retrieval strategies automatically. You can later switch to Custom Settings for fine-tuning.
Be mindful of free-tier limits (file size, web crawling cap). For production, connect your own S3, MongoDB, or vector database.
Sensitive data should not be uploaded to the default workspace—configure secure storage first.
Last updated