A typed search by meaning. The app sends your words and the search it built; Opensolr turns the words into a query vector and runs the search on your own photos index. Only the results come back.
POST https://api.opensolr.com/solr_manager/api/photos_selectA JSON body with email, api_key, index_name, vector_text (your words, 2 to 300 characters), top_k and params, the search as name and value pairs.
It runs once per typed search with the AI switch on, on a plan with vector search. The answer is kept in the phone's search cache for a while.
vectorQuery is added by Opensolr; alpha follows the balance in search tuning, and mm its Lexical matching step.
- The index must be a photos index of the caller's account.
- Refused parameters: a vector of the caller's own, the response format, the request handler, shards, streaming, commits, and joins or graphs that would read another index.
- The vector is made on Opensolr's side and never returned.
Solr's answer, as it is. A tenth of an AI request when the vector had to be made; the same words searched again cost nothing.
VECTOR_NOT_ALLOWEDThe plan has no vector search.
ERROR_AI_MONTHLY_QUOTA_EXCEEDEDThis month's AI requests are used up, with the date they reset.
VECTOR_UNAVAILABLENo vector could be made at that moment.
In each case the app runs the same search on the words alone, straight on your index, and does not keep that answer in its cache.