Learning Timeline
Key Insights
Optimizing with Q&A
Use the Q&A section specifically for the top 10 questions that account for the majority (e.g., 90%) of your support volume. Hard-coding these ensures the chatbot gives the exact approved answer every time.
Using Text Snippets for Rules
For rigid procedures like 'Refund Requests,' do not rely solely on crawled pages. Paste the exact procedure into the 'Text' section to ensure the AI follows specific logic steps.
Step by Step
Initializing the Website Crawler
- Click the 'Build your AI agent' button on the Chatbase homepage.
- Select the 'Website' data source option.
- Paste the target URL into the provided text field.
- Set the 'Use Case' to 'Customer Support Agent'.
- Click the 'Continue' (or 'Fetch Links') button to initiate the crawler.
- Wait for the process to finish scanning the site (e.g., finding ~58 links).
- Review the list of populated URLs displayed on the screen.
Refining Data and Training the Agent
- Identify irrelevant pages in the crawled link list.
- Click the exclude/delete icon next to any link you wish to remove from the training data.
- Add specific URLs manually if the crawler missed a required page.
- (Optional) Navigate to the 'Files' tab to upload PDF or Word documents containing internal knowledge.
- (Optional) Navigate to the 'Q&A' tab to manually input high-frequency questions and answers.
- (Optional) Navigate to the 'Text' tab to paste specific snippets (like refund policies).
- Click the 'Train and continue' button to finalize the data ingestion.
