### Answer --- Here are several things you can try to improve accuracy when retrieving vehicle info by stock number: 1. Improve Content Structure Prefer

.md

(Markdown) files over plain text or raw HTML. [(1) ]() Well-structured Markdown is chunked and retrieved more accurately. Use clear, descriptive headings and keep each section focused and reasonably short. For tabular or image-heavy content (like vehicle listings) , retrieval may be less reliable—consider adding explanatory text so related information stays in the same chunk. 2. Avoid Ambiguity Include specific identifiers like stock numbers, names, and details directly in the text. Avoid ambiguous pronouns like "it" or "this" because prior chunks may not be present during retrieval. 3. Adjust Retrieval Settings After connecting a knowledge base to your agent, you can configure how many chunks are retrieved via "Adjust KB Retrieval Chunks and Similarity": - **Chunks to retrieve**: Max number of chunks retrieved (range 1-10, default 3) . - **Similarity Threshold**: Adjust how strict the matching is. A higher setting gives fewer but more similar matches (default 0.6) . If your vehicle data is detailed, you may want to increase chunks to retrieve slightly, but note that increasing this can also interfere with generation quality. *4. Use Node-Level Knowledge Base (for Conversation Flow Agents) * You can assign a specific knowledge base to a specific node in your conversation flow. This allows the agent to retrieve more accurate answers based on context and respond with lower latency. [(1) ]() [(2) ]() 5. Consider Content Format If your website is dynamically generated or heavily relies on JavaScript, the content may not be scraped effectively. You mig

ht get better results by exporting vehicle data to a structured

.md

or

.csv

file and uploading it directly. [(1) ]() --- Summary: The knowledge base uses the transcript so far to find the most relevant chunks—it doesn't perform a live lookup on the website during the call. If the stock number and vehicle details aren't clearly associated in the scraped content, retrieval will be inconsistent. Restructuring your data with clear headings and explicit stock number references should help.

​ To learn more, see the sources I considered: [[1] Knowledge Base](https://docs.retellai.com/build/knowledge-base) [[2] Platform Changelogs | Retell AI](https://www.retellai.com/changelog) ​

Feel free to tag @AI Evy with additional questions. I also appreciate feedback, it helps me improve.