How can we add the whole Vendor Catalog in a chat without the character limit ?
I have 10K ASINs + Item names , which is way over the current limit
Hi @laxmie and welcome to the community!
I have a few workarounds for your case. The first is to take your vendor catalog (as a CSV or Excel file) and upload it into a Space. Once the file is indexed, you can focus that Space in your chat conversation and the agent will ideally be able to search and retrieve from the full catalog without any character restrictions. The second approach you can do is to load it as a dataset, if your catalog is structured (ASIN, Item Name, category, price, etc.), you can upload it as a quick sight dataset connected through SPICE. Once it’s in SPICE, you can optionally create a Topic on top of it, which lets the chat agent query your full 10K+ rows using natural language.
I would also like to add that you can add ‘Specific Data’ to a chat agent. If you go to your chat agent and expand the ‘All’ dropdown, you have the option to tailor it to a specific dataset.
Hope this helps!
I am trying to create a chat agent which can give recommendations for Item Names. For a list of 7k+ ASINs, the chat is giving me report for 70 ASINs only. Any way to fix this?
Hi @laxmie,
The 70-ASIN cap you are seeing is not the character limit anymore, it is the way a chat agent surfaces query results. When the agent runs against your dataset or Topic, it works with a limited result set rather than enumerating thousands of rows, so a single request that asks it to generate recommendations across all 7k+ ASINs will keep coming back truncated no matter how the catalog is loaded. In other words, the agent is designed to reason over your data and answer targeted questions, not to produce a full row-by-row report of the entire catalog in one pass.
One way to work with this is to narrow each ask so the agent returns a focused set rather than the whole catalog at once. For example, you can request a recommendations for a specific category, brand, or a top-N slice (top 10 for example), or pass a smaller batch of ASINs per query and iterate. That keeps each response within what the agent can return cleanly.