== 1. longest chunk and batch size
longest chunk: 2093 bytes -> embed_total_tokens=371 (model limit 2,048 tokens per input)
97 documents in one upsert: ApiError: [400 INVALID_ARGUMENT] Number of documents (97) exceeds the maximum limit of 96 documents per upsert request. This index embeds a field server-side, and one request is one embedding call, so it is limited to the embedding batch size rather than the usual document limit. Split the documents across smaller requests.

== 2. text far over the model limit (default truncate=END)
text: 63600 bytes
over-length upsert: OK DocumentUpsertUsage(embed_total_tokens=2050)

== 3. delete filters
delete with text-match filter: ApiError: [400 INVALID_ARGUMENT] Text match operators ($match_phrase, $match_all, $match_any) are only supported in query requests
delete with $eq on url (nothing matches): OK 0

== 4. id prefix
documents.list(prefix='7a') in 'docs': OK ['7a4cf5cb7fa2ff5f', '7a98e091a559c766']
delete(prefix=...): TypeError: Documents.delete() got an unexpected keyword argument 'prefix'

== 5. namespaces
describe_namespace on a name never written: NotFoundError: [404] No namespace: 'never-written' found in index kb-docs
search in a namespace never written: OK 0

== 6. schema is fixed at creation
create the same index name again: ConflictError: [409 ALREADY_EXISTS] Resource  already exists
upsert with a vector on an embed field: ApiError: [400 INVALID_ARGUMENT] Document with id 'v': field 'chunk_text' embeds its text server-side, so it must be a string. Supply the text and Pinecone derives the vector; a caller-supplied vector is not accepted on this field.
upsert with an empty string on an embed field: ApiError: [400 INVALID_ARGUMENT] Document with id 'e': field 'chunk_text' is empty, so there is no text to embed. Provide the text to embed.

== 7. other hosted models on a Starter account
gemini-embedding-2: ForbiddenError: [403 PERMISSION_DENIED] Schema field 'chunk_text' uses model 'gemini-embedding-2', which is not available on the Starter plan. Upgrade to the Builder, Standard or Enterprise plan to use 'gemini-embedding-2', or choose another embedding model.
processing_location=us: ApiError: [400 INVALID_ARGUMENT] Invalid 'embed' configuration on schema field 'chunk_text': Invalid field 'processing_location' - Unsupported field 'processing_location'. Model llama-text-embed-v2 has no processing location. Remove 'processing_location'.
indexes now: ['kb-docs-byo', 'kb-docs']

probe namespace deleted