Use Upstash Vector namespaces and built-in embedding models to upsert and query raw text without a separate embedding call

Documented steps

1. Create the index WITH an embedding model configured — via the console dropdown or the Management API create_index call with "embedding_model":"<model>" and, for hybrid/sparse, "sparse_embedding_model" plus "index_type".
2. Documented dense/hybrid embedding models: BAAI/bge-large-en-v1.5 (1024 dims, 512 sequence length), BAAI/bge-base-en-v1.5 (768 dims, 512), BAAI/bge-small-en-v1.5 (384 dims, 512), BAAI/bge-m3 (1024 dims, 8192).
3. Sparse indexes accept only BAAI/bge-m3 or BM25 — the dense-only bge-large/base/small models are not valid sparse models.
4. /upsert-data and /query-data work only on an index created with an embedding model; they will not work against a plain vector-only index.
5. Upsert raw text: POST {UPSTASH_VECTOR_REST_URL}/upsert-data (or /upsert-data/{namespace}) with 'Authorization: Bearer $UPSTASH_VECTOR_REST_TOKEN' and body {"id":"<string>","data":"<raw text>","metadata":{...optional...}} — or an array for batch. Upstash embeds server-side and returns {"result":"Success"}.
6. Query with raw text: POST {UPSTASH_VECTOR_REST_URL}/query-data (or /query-data/{namespace}) with body {"data":"<query text>","topK":10,"includeMetadata":true,"includeVectors":false,"includeData":false,"filter":"<filter string>"}.
7. Namespace addressing is a path segment: /upsert-data/my-namespace, /query-data/my-namespace. Omitting it targets the default namespace (empty-string name). One request reads or writes exactly one namespace.
8. Namespaces need no explicit creation — they exist on first upsert. Indexes created before namespaces existed keep working and implicitly use the default namespace.
9. For hybrid indexes (dense plus sparse/BM25), tune result combination in /query-data with "weightingStrategy":"IDF", "fusionAlgorithm":"RRF" or "DBSF", and "queryMode":"HYBRID"|"DENSE"|"SPARSE".
10. Response shape from /query-data: {"result":[{"id":...,"score":...,"metadata":{...},"vector":[...],"sparseVector":{"indices":[...],"values":[...]},"data":"..."}]} with optional fields present only when requested.
11. Official docs: https://upstash.com/docs/vector/features/embeddingmodels | https://upstash.com/docs/vector/api/endpoints/upsert-data | https://upstash.com/docs/vector/api/endpoints/query-data | https://upstash.com/docs/vector/features/namespaces