Vectara’s free tier; what can you *really* build?
TLDR; With Vectara’s free tier, you can add LLM-powered search to almost any single website on the internet or index the equivalent of 7 copies of Tolstoy’s War and Peace… Read more »
January 23, 2023 by Ed Albanese
TLDR; With Vectara’s free tier, you can add LLM-powered search to almost any single website on the internet or index the equivalent of 7 copies of Tolstoy’s War and Peace or 17,000 minutes of audio transcription.
Vectara is priced using two simple metrics; “Queries per Month” and “Account Size”. Every Vectara account starts with 50MB of Account Size and 15,000 free Queries per Month for free, forever.
“Queries per Month” A query is defined as a discrete request made to Vectara’s API. Vectara counts an API call against a single corpus as a single query. The total number of those discrete requests over a full 30 day period equals “Queries per Month” for any single account.
“Account Size” is the total available disk capacity allocated for storing extracted and compressed text from documents including JSON, PDF, Microsoft Office, HTML, Markdown and more. Every Vectara Free Tier account starts with 50 MB of Account Size.
So, what does 50 MB of text mean in the real world?
To give you real world examples of the usable capacity with Vectara’s Free tier, we created 3 corpora with the following size metrics.
- We indexed the full 1,225 pages of Leo Tolstoy’s War and Peace. Total corpus size was 6.9 MB. With Vectara’s free tier, users get enough capacity to index around 9,000 pages of text, or a bit more than 7 full copies of War and Peace.
- We indexed transcripts of 100 public company earnings calls with a total of 1,000 pages. The average length for the call recordings was 35 minutes with a transcript length of 10 pages and 5,000 words per transcript. Total corpus size was 10.8 MB. With Vectara’s free tier, users get enough capacity to index over 17,000 minutes of call recordings
- We indexed 150 web pages from Vectara.com and docs.vectara.com using our Web Crawler. Total corpus size was 1.4 MB. With Vectara’s free tier, users get enough storage capacity to index roughly 5,357 web pages. The average website is under 20 pages. Very few sites exceed 1000 pages. >99% of unique domains on the internet could be indexed and searched using Vectara’s free tier.
What uses up your available Account Size?
- Extracted and compressed raw text.
- Optional client-provided metadata (during indexing) such as “original_url” (original file location), “document_owner” (document owner user ID), and “has_pii” (whether indexed content contains personally identifiable information).
- Metadata automatically attached by Vectara during ingest such as language type and time of index.
What does not use up your available Account Size?
- File formatting specifications such as cross-reference (xref) tables, figures, and field codes.
What happens if I exceed the available Account Size for Vectara’s free tier?
Subscribers are unable to add additional files. All prior successfully added files are retained and maintained, but Account Size is fixed and must be expanded by directly requesting additional capacity.
How do you secure additional Account Size capacity?
- Growth Plan subscribers can add a credit card at any time to purchase units as small as 5 MB of additional Account Size, which include 1000 additional Queries per Month as part of that same bundle.
- Scale Plan subscribers can contact [email protected] anytime to quickly add additional capacity.
How are the 15,000 free queries allocated within Vectara’s free tier?
The 30 day period starts when a Vectara Free user signs up. For example, if a Vectara user signs up on January 5th, 2023, the 30 day period starts on January 5th, 2023 and the 15,000 Queries per Month quota resets on February 6th, 2023.
If you have any questions about Vectara’s plans, please visit our Pricing page or contact us directly at [email protected]
Happy Indexing and Searching!