Interactive Query Demo: Q&A vs Search+Chunk

See how different intents work together to power AI workflows

Compare corpus-wide discovery (Search, Q&A) with single-document precision (Chunk) using the same question. Understanding when to use each intent is key to building efficient RAG systems.

Select a sample question:

Demonstrates corpus-wide search vs single-document chunk extraction

🔍

Use Search When

• Agents need to discover what content exists without external search engines
• Finding multiple relevant documents across your corpus
• Building a list of candidates to process
• Enabling direct discovery instead of relying on Google/Bing

💬

Use Q&A When

• You want a direct answer to a question
• Need synthesis across multiple sources
• Attribution/citations are important
• End-user experience (chatbots)

✂️

Use Chunk When

• You already know which document to process (good for using after getting Search results)
• Building your own RAG system
• Need text + vectors together
• Precision grounding with citations

🔑 Understanding the Relationships

Embeddings vs Chunk:

• Same vectors: Both use identical 1536-dimensional embeddings for semantic representation

• Embeddings: Returns ALL chunks blindly (comprehensive coverage). Use when you want to index everything for your own vector database.

• Chunk: Returns only RELEVANT chunks (query-filtered) with both text AND vectors. Use when you know what you're looking for and want precision with grounding.

Think: "give me all the pieces to index" vs "give me just the pieces that answer my question"

Q&A vs Search + Chunk:

• Q&A Pros: Complete synthesized answer, multi-document reasoning, ready for end-users, includes citations automatically

• Q&A Cons: Black box synthesis, higher latency, no control over answer construction, can't inspect reasoning

• Search + Chunk Pros: Full transparency, custom RAG pipeline control, efficient multi-step workflow, can inspect/modify at each stage

• Search + Chunk Cons: Requires building your own answer generation, multi-step process, need to manage context window

When to use what: Use Q&A for chatbots and direct user-facing answers. Use Search + Chunk when building custom RAG systems where you want control over retrieval, ranking, and generation.