Victor Joseph
Last updated: 13 December 2024
Are there any limitations to the sources we can add to the AI Agent?
Only publicly accessible web URLs can be added as a source; private URLs are not supported.
In addition to web URLs, you can also upload files in the following formats: .txt, .pdf, .docx, .json, .doc, .md, and .html. Please note that files must be less than 50 MB.
Additionally, you can select “Thena requests” as a source, allowing the AI Agent to analyze existing requests and provide answers based on them.
How often is the data updated or re-scraped when a URL is added?
For web URLs, we provide an option to manually trigger the rescraping process. You should re-scrape whenever your public documentation is updated.
Rescraping is not available for file uploads, as files do not change after being uploaded.
For Thena requests, the data is automatically updated every 24 hours.
How is the data handled?
The scraped data is stored in OpenAI’s vector store, which manages embeddings.
When a user asks a question, the query is sent to the OpenAI assistant along with relevant context from the vector store. The assistant retrieves the relevant data from the vector store to answer the question and constructs the final response.
Each customer has a separate assistant and a corresponding vector store.
A diagram is attached to visually explain this process.
