Create an Upstash Search index and run semantic or full-text search over documents

Documented steps

1. Product-status check first: the Upstash Search getstarted page currently steers new projects toward Redis Search ('We now have Redis Search in Upstash. If you are starting a new project, we recommend using Redis Search'). Confirm Upstash Search is still the right choice before building on it.
2. There is no dedicated data-plane index-creation endpoint — a Search index is created implicitly on first upsert via the SDK: client.index('<name>') then index.upsert(...).
3. To create/list/delete an index explicitly, use the Management API: POST https://api.upstash.com/v2/search with HTTP Basic auth (email:management_api_key) and body {"name":...,"region":"eu-west-1|us-central1","type":"free|payg|fixed"}. The response includes endpoint, token and read_only_token.
4. Install the SDK: pip install upstash-search (Python) or npm install @upstash/search (TypeScript).
5. Set UPSTASH_SEARCH_REST_URL and UPSTASH_SEARCH_REST_TOKEN from the console Connect tab, then Search.from_env() (Python) / Search.fromEnv() (TypeScript) and index = client.index('films').
6. Document shape: {"id":"<string>","content":{<searchable JSON fields>},"metadata":{<optional, NOT indexed>}}. Example: index.upsert(documents=[{'id':'movie-0','content':{'title':'Star Wars','overview':'Sci-fi space opera','genre':'sci-fi'},'metadata':{'poster':'https://...'}}]).
7. content fields are indexed and are both searchable and filterable; metadata is documented as not indexed — use it only for display data.
8. Search: index.search({'query':'space opera','limit':5,'reranking':false,'filter':"genre = 'sci-fi'",'semanticWeight':0.75,'inputEnrichment':true,'keepOriginalQueryAfterEnrichment':false}). Defaults are limit 5, reranking false, semanticWeight 0.75, inputEnrichment true.
9. semanticWeight ranges 0-1: higher favors conceptual/semantic matching, lower favors exact keyword matching.
10. Reranking is opt-in ('reranking': true) and adds an AI rerank pass on top of hybrid results, billed at $1 per 1,000 reranked documents.
11. Filters on content fields use SQL-like expressions similar in style to Upstash Vector's filter syntax.
12. The response includes a results array; scores are normalized 0-1 with 1 most relevant. The SDKs also expose index.fetch/delete plus client.list_indexes() and client.delete_index(name).
13. Official docs: https://upstash.com/docs/search/overall/getstarted | https://upstash.com/docs/search/features/content-and-metadata | https://upstash.com/docs/search/features/reranking | https://upstash.com/docs/search/features/advanced-settings