# Advanced RAG Settings Fine-tune how your AI retrieves and uses knowledge from your documents When you add knowledge sources (documents, websites, etc.) to your Chipp app, the AI uses **Retrieval-Augmented Generation (RAG)** to find and use relevant information. The Advanced RAG Settings let you fine-tune this process. ## How Your AI Uses Your Documents Before diving into settings, here's how your AI actually finds and uses information from your uploaded documents. ### What is RAG? **RAG (Retrieval-Augmented Generation)** is how your AI "reads" your documents to answer questions. Instead of memorizing everything, the AI: 1. **Searches** your documents for relevant information when a user asks a question 2. **Retrieves** the most relevant sections 3. **Generates** a response using that information Think of it like a research assistant who quickly scans your files for relevant passages before answering, rather than trying to memorize every document. ### What is a "Chunk"? When you upload a document, Chipp automatically breaks it into smaller pieces called **chunks**. A chunk is typically a paragraph or a few paragraphs of text. **Why chunks?** - AI models have limits on how much text they can process at once - Smaller pieces let the AI find the exact relevant section, not the whole document - It's faster and more accurate than searching full documents **Example:** A 50-page PDF might be split into 200 chunks. When a user asks a question, the AI searches all 200 chunks and retrieves only the 5 most relevant ones. ### How Search Works When a user asks a question, Chipp converts both the question and your document chunks into mathematical representations (called "embeddings"). It then finds chunks whose meaning is closest to the question's meaning. This is **semantic search** - it understands meaning, not just keywords. So "How do I cancel my subscription?" will match a chunk about "ending your membership" even if those exact words aren't used. ### Standard Search vs. Hybrid Search **Standard Search (Default)** Looks at each chunk independently and finds the ones most similar to the user's question. - Works great for: FAQ-style content, focused topics, smaller knowledge bases - Best when: Each chunk contains self-contained answers **Hybrid Search** Combines two search strategies: 1. **Chunk search**: Which individual chunks match the question best? 2. **Document search**: Which overall documents are most relevant? Then it boosts chunks that come from highly relevant documents, even if those chunks aren't the top individual matches. - Works great for: Multiple documents on similar topics, long documents, broad questions - Best when: Context from the same document matters, or you have 10+ files **Example:** A user asks about "return policies." Standard search might find one great chunk from your FAQ and another from a random product page. Hybrid search recognizes that your "Returns & Refunds" document is highly relevant overall, so it prioritizes chunks from that document. ## Accessing RAG Settings 1. Go to your app in the Chipp dashboard 2. Navigate to **Build** tab 3. Scroll to **Knowledge Sources** 4. Click **Advanced RAG Settings** to expand ## Settings Reference ### Relevance Threshold **Default: 0.15** | Range: 0.00 - 1.00 Controls how closely a document chunk must match the user's query to be included in the AI's context. | Value | Behavior | | ------------------------ | ---------------------------------------------------------- | | **Lower (0.00 - 0.15)** | More results included, may include loosely related content | | **Medium (0.15 - 0.30)** | Balanced approach (recommended starting point) | | **Higher (0.30 - 1.00)** | Stricter matching, only highly relevant content | **When to adjust:** - **Lower the threshold** if users report the AI doesn't find relevant information that exists in your documents - **Raise the threshold** if responses include too much irrelevant information or the AI seems confused by conflicting sources ### Max Chunks **Default: 5** | Range: 1 - 20 The maximum number of document chunks retrieved for each user query. Each chunk is a section of your uploaded content. | Value | Behavior | | ---------------------- | -------------------------------------------- | | **Fewer chunks (1-3)** | Focused responses, lower token usage, faster | | **Moderate (4-7)** | Balanced context (recommended) | | **More chunks (8-20)** | Comprehensive coverage, higher token usage | **When to adjust:** - **Use fewer chunks** for simple Q&A where answers are typically in one place - **Use more chunks** when questions require synthesizing information from multiple sources, or for complex topics with scattered information **Note:** More chunks means more context for the AI, but also increases token usage and may slow responses. ### Hybrid Search **Default: Off** When enabled, combines two search strategies: 1. **Chunk-level search**: Finds individual chunks most similar to the query 2. **Document-level search**: Considers which documents overall are most relevant This helps when the best answer comes from a document that's highly relevant overall, even if no single chunk perfectly matches the query. **When to enable:** - You have multiple documents on similar topics - Users ask broad questions that span document sections - You notice the AI missing context from relevant documents **When to keep off:** - Simple, focused knowledge bases - When chunk-level matching works well ### Document Weight **Default: 0.30** | Range: 0.00 - 1.00 | Only visible when Hybrid Search is enabled Controls the balance between chunk-level and document-level relevance when Hybrid Search is on. | Value | Behavior | | -------------------------- | ---------------------------------------------------------- | | **Lower (0.00 - 0.25)** | Prioritize individual chunk matches (chunk focus) | | **Balanced (0.25 - 0.50)** | Equal weight to both signals | | **Higher (0.50 - 1.00)** | Prioritize chunks from relevant documents (document focus) | **When to adjust:** - **Lower values** when your documents cover very different topics and chunk relevance is most important - **Higher values** when you have long documents where context from the same document is valuable ## Recommended Configurations ### Customer Support Bot ``` Relevance Threshold: 0.20 Max Chunks: 5 Hybrid Search: Off ``` Good for FAQ-style knowledge bases where answers are self-contained. ### Research Assistant ``` Relevance Threshold: 0.10 Max Chunks: 10 Hybrid Search: On Document Weight: 0.40 ``` Retrieves more context for synthesizing comprehensive answers. ### Technical Documentation ``` Relevance Threshold: 0.25 Max Chunks: 7 Hybrid Search: On Document Weight: 0.30 ``` Balances precision with coverage for technical queries. ### Legal/Compliance Bot ``` Relevance Threshold: 0.15 Max Chunks: 8 Hybrid Search: On Document Weight: 0.50 ``` Ensures relevant policy documents are well-represented. ## Troubleshooting ### AI doesn't find information that exists in my documents 1. Lower the **Relevance Threshold** to 0.10 2. Increase **Max Chunks** to 8-10 3. Enable **Hybrid Search** if you have multiple related documents ### Responses include irrelevant information 1. Raise the **Relevance Threshold** to 0.25-0.35 2. Decrease **Max Chunks** to 3-4 3. Review your document organization ### Token usage is too high 1. Reduce **Max Chunks** to 3-4 2. Raise **Relevance Threshold** to reduce low-quality matches 3. Consider splitting large documents into focused topics ### AI gives conflicting information 1. Raise **Relevance Threshold** to be more selective 2. If using Hybrid Search, increase **Document Weight** to favor cohesive document sources 3. Review documents for outdated or contradictory content ## Resetting to Defaults Click **Reset to defaults** in the Advanced RAG Settings panel to restore all settings to their default values. This is useful if you've experimented and want to start fresh.