introducing Zai #4
Description
Zai, or Zenodeo AI, is an AI-enhanced interface to the treatments database. Here is how it works.
You can ask it a question, for example:
What is Saigona baiseensis?
Does Rhinolophus sinicus live in caves?
Where does Laephotis botswanae roost?
What distinguishes Choanolaimus sparsiporus?
What is the etymology of Gammanema lunatum?
What distinguishes Cynodon gibbus from other cofamilial genera?
Zai can also describe more than 760K species by their binomen. Just type "Describe " like so:
Describe Saigona sinicola
Describe Carvalhoma malcolmae
Describe Amnestus sinuosus
With a minimum of three letters typed after "Describe …", Zai returns an autosuggest dropdown of matching binomens. Choose one and get a (hopefully useful) description. For every query, Zai returns an answer formulated by the LLM using the fulltext of a relevant treatment, and also displays related images.
The query itself is parsed into searchterms (green) and stopwords (lightgrey). Zai first performs a search on the searchterms, finds the matching treatments, and uses the fulltext of the top-ranked treatment (bm25() ranking) as the context to ask the LLM to generate an answer.
The answers are stored in a semantic cache that can match similiar queries, for example, Describe Saigona baiseensis vs. What is Saigona baiseensis?
I have been working on this app now for almost a couple of months. Now it is time to reveal to you all as a test. There are several known bugs already, and as you use it, more problems will be discovered. But I am very happy with how all its components have turned out, and the future for such an app is very exciting.
Please use the app and open an issue if you encounter any problem.
cc:
@tcatapano
@myrmoteras
@flsimoes
@Carol-Sokolowicz
@gsautter
@jugiora
@juwingert
@slint
@lnielsen