Cloudflare/vectorize

@webframp/cloudflare/vectorizev2026.10.21.1· Apache-2.0· 5d agoBEHAVIORS

01README

Cloudflare Vectorize — vector indexes, insert/query/delete operations

02Release Notes

2026.10.21.1

Changed: The README and the apiToken description now lead with a vault. Store the token with swamp vault put cloudflare API_TOKEN and set apiToken to ${{ vault.get("cloudflare", "API_TOKEN") }}. The CLOUDFLARE_API_TOKEN environment variable still works as a fallback, but swamp does not redact it, so it is no longer recommended. The missing-token error now says the same.

Fixed: get_vectors_by_id wrote to a resource the model does not declare, so it always failed. It now writes the resource the model declares.

Fixed: Create methods whose response carries no id wrote every result to one instance named created, so each call overwrote the previous one. They now name the instance by the response's id when present, else by the path parameter that scopes the call, else by a hash of the request arguments (secret arguments left out). Affected: create_vectorize_index, delete_vectors_by_id, create_metadata_index, delete_metadata_index and create_vectorize_query_vector.

Upgrade note: Results earlier versions wrote to the created instance stay there; new runs write under the new names.

03Behaviors1

@webframp/cloudflare/vectorizev2026.10.21.1cloudflare/vectorize.ts

Global Arguments

Argument	Type	Required	Description
apiToken?	string	no	Cloudflare API token. Wire it with ${{ vault.get("cloudflare", "API_TOKEN") }}; never pass the literal token. The CLOUDFLARE_API_TOKEN environment variable is read as a fallback, but it is not redacted and is not recommended.
accountId	string	yes	Cloudflare account ID

fn list_vectorize_indexes()

List Vectorize Indexes

fn create_vectorize_index(config: unknown, description?: unknown, name: unknown)

Create Vectorize Index

Argument	Type	Required
config	unknown	yes
description?	unknown	no
name	unknown	yes

fn get_vectorize_index(index_name: string)

Get Vectorize Index

Argument	Type	Required	Description
index_name	string	yes

fn delete_vectorize_index(index_name: string)

Delete Vectorize Index

Argument	Type	Required	Description
index_name	string	yes

fn delete_vectors_by_id(index_name: string, ids?: array)

Delete Vectors By Identifier

Argument	Type	Required	Description
index_name	string	yes
ids?	array	no	A list of vector identifiers to delete from the index indicated by the path.

fn get_vectors_by_id(index_name: string, ids?: array)

Get Vectors By Identifier

Argument	Type	Required	Description
index_name	string	yes
ids?	array	no	A list of vector identifiers to retrieve from the index indicated by the path.

fn get_vectorize_index_info(index_name: string)

Get Vectorize Index Info

Argument	Type	Required	Description
index_name	string	yes

fn vectorize_insert_vector(index_name: string, unparsable_behavior?: enum)

Insert Vectors

Argument	Type	Required	Description
index_name	string	yes
unparsable_behavior?	enum	no

fn list_vectors(index_name: string, count?: number, cursor?: string)

List Vectors

Argument	Type	Required
index_name	string	yes
count?	number	no
cursor?	string	no

fn create_metadata_index(index_name: string, indexType: enum, propertyName: string)

Create Metadata Index

Argument	Type	Required	Description
index_name	string	yes
indexType	enum	yes	Specifies the type of metadata property to index.
propertyName	string	yes	Specifies the metadata property to index.

fn delete_metadata_index(index_name: string, propertyName: string)

Delete Metadata Index

Argument	Type	Required	Description
index_name	string	yes
propertyName	string	yes	Specifies the metadata property for which the index must be deleted.

fn list_metadata_indexes(index_name: string)

List Metadata Indexes

Argument	Type	Required	Description
index_name	string	yes

fn create_vectorize_query_vector(index_name: string, filter?: object, returnMetadata?: enum, returnValues?: boolean, topK?: number, vector: array)

Query Vectors

Argument	Type	Required	Description
index_name	string	yes
filter?	object	no	A metadata filter expression used to limit nearest neighbor results.
returnMetadata?	enum	no	Whether to return no metadata, indexed metadata or all metadata associated wi...
returnValues?	boolean	no	Whether to return the values associated with the closest vectors.
topK?	number	no	The number of nearest neighbors to find.
vector	array	yes	The search vector that will be used to find the nearest neighbors.

fn vectorize_upsert_vector(index_name: string, unparsable_behavior?: enum)

Upsert Vectors

Argument	Type	Required	Description
index_name	string	yes
unparsable_behavior?	enum	no

Resources

vectorize_indexes(infinite)— List Vectorize Indexes

vectorize_index(infinite)— Create Vectorize Index

delete_vectors_by_id(infinite)— Delete Vectors By Identifier

vectors_by_id(infinite)— Get Vectors By Identifier

vectorize_index_info(infinite)— Get Vectorize Index Info

vectorize_insert_vector(infinite)— Insert Vectors

list_vectors(infinite)— List Vectors

metadata_index(infinite)— Create Metadata Index

delete_metadata_index(infinite)— Delete Metadata Index

list_metadata_indexes(infinite)— List Metadata Indexes

vectorize_query_vector(infinite)— Query Vectors

vectorize_upsert_vector(infinite)— Upsert Vectors

04Previous Versions12

2026.10.14.1Apache-2.0(was none)Oct 14, 2026

2026.10.14.1

Fixed: Methods that write a singleton resource failed on current swamp with Data name 'latest' is reserved for internal use.

Changed: Singleton get and action methods now write their data under vectorize_insert_vector-current, vectorize_upsert_vector-current instead of latest. Anything that read data.latest("<model>", "latest") or queried name == "latest" must use the new name.

Upgrade note: Data previously stored under latest is not migrated; re-run the method to populate the new name. No attribute schemas changed.

Added: The manifest now declares license: Apache-2.0 (SPDX). The registry quality rubric scores a declared license separately from the bundled LICENSE.md.

2026.09.18.1Sep 20, 2026

2026.09.18.1

Upgrade note: Normalized npm:zod dependency version to 4.6.5 across the repo. No behavioral changes in this extension.

2026.09.17.2Sep 17, 2026

2026.09.17.2

No changes to this extension's methods. Regenerated to refresh the shared _lib/api.ts helper library after PR #427 review fixes to cfApiPaginatedCursor (used by r2, kv, durable-objects, and workers-ai, but emitted into every extension's api.ts since it's part of the shared template).

2026.09.15.1Sep 15, 2026

2026.09.15.1

Changed: Bump zod 4.4.3 → 4.6.5

2026.08.28.2

Hardened codegen: instance names sanitized against path traversal; 429 rate-limit retry with Retry-After; string schema patterns emit regex validation; output schemas passthrough unknown fields; apiToken now optional and vault-wireable with CLOUDFLARE_API_TOKEN env fallback.

2026.08.28.2Aug 28, 2026

2026.08.28.2

updated labels

2026.08.28.1Aug 28, 2026

2026.08.28.1

Changed: Normalized the extension license to Apache-2.0 and corrected the copyright holder to "Sean Escriva". Extensions that previously shipped an MIT LICENSE.md are now Apache-2.0, consistent with the repository root and every other extension. No code or behavioral changes.

Upgrade note: License text only. No API, schema, or runtime behavior changed.

2026.08.26.2

Fixed: Restored inline npm:zod@4.4.3 import specifiers so the registry quality scorer can resolve dependencies and score the extension. An earlier release used a bare "zod" import-map specifier, which published but scored as unscored.

Changed: Retained explicit compilerOptions.strict in deno.json. No behavioral or schema changes.

2026.08.26.2Aug 27, 2026

2026.08.26.2

Changed: Retained explicit compilerOptions.strict in deno.json. No behavioral or schema changes.

2026.08.25.1Aug 26, 2026

2026.08.25.1

Changed: Updated labels for improved extension discoverability. Added cross-cutting category labels (security, observability, finops, infrastructure, networking, compliance, devops, ai, incident-response) where applicable.

updated labels

2026.08.24.1Aug 24, 2026

2026.08.24.1

Added: Output metadata attributes for observability.

durationMs: Method execution duration in milliseconds.
collectedBy: Extension name that produced the data.
fetchedAt: ISO 8601 timestamp when data was fetched (added to resources that previously lacked it).

2026.08.21.1Aug 22, 2026

2026.08.21.1

Changed: Errors and validation now say what went wrong and where.

Cloudflare API failures (HTTP errors and success: false API responses) now name the HTTP method and path that was attempted, in addition to the original status/message. Previously the error read only Cloudflare API error: <message>, with no indication of which endpoint the method call was hitting — now it reads Cloudflare API request failed: GET /accounts/<account>/vectorize/v2/indexes/<name>: <message>. This applies to every method on this model, since they all share the same request helper.
index_name is now validated as non-empty before any request is made. Passing an empty string previously sent a request to a malformed URL and produced a confusing 404 from Cloudflare; it now fails fast with index_name must not be empty.

Upgrade note: No method was added, removed, or renamed. Existing callers that always pass a non-empty index_name see no behavioral change beyond clearer error text.

2026.07.27.1Jul 27, 2026

2026.07.27.1

Fixed: Regenerated from scripts/cloudflare-codegen after two generator bugs were repaired (webframp/swamp-extensions#284).

Methods referencing an undeclared path parameter did not compile. The generator derived a method's arguments schema and execute signature from the OpenAPI parameters list, but built the request URL from the path template. Where the Cloudflare spec omits a declaration for a {placeholder} — which it does in several places — the result was a method with arguments: z.object({}) and an unused _args parameter whose body still interpolated args.<name>. Those methods failed type checking and were uncallable even if they had compiled, because the argument was never declared. Path-template placeholders are now unioned into the declared parameters, so the schema, the signature, and the body agree.
Generated tests could request a URL the mock server did not serve. Test arguments merged the request-body fixture over the path-parameter values, so a body property sharing a name with a path parameter (commonly id) substituted its own example value into the URL. The request then missed the mock and failed with Cloudflare API error: Not found. Path parameters now take precedence, matching what the generated model already does by excluding path-parameter names from the request body.

Upgrade note: No API surface change and no method was added or removed. If this extension type-checked and tested cleanly before, its behavior is unchanged and only the version moved. Extensions that previously failed deno check or deno task test now pass.

2026.07.19.1Jul 19, 2026