payloadcms-vectorize@1.0.1

Patch Changes

• #65 d8fea7a Thanks @techiejd! - Defer adapter config validation to call-time instead of throwing at construction.

  The mongodb and cf adapter factories previously threw on missing config (e.g. uri, dbName, binding) the moment they were called, which happens while the Payload config is being built. That broke payload generate:types and generate:importmap in environments without runtime variables — such as CI that builds payload-types to publish as a separate package. Validation now runs when an adapter method is actually invoked, so config-time codegen no longer requires runtime credentials. Valid configurations behave exactly as before.

  Closes #64.

@payloadcms-vectorize/pg@1.0.1

Patch Changes

• #65 d8fea7a Thanks @techiejd! - Defer adapter config validation to call-time instead of throwing at construction.

  The mongodb and cf adapter factories previously threw on missing config (e.g. uri, dbName, binding) the moment they were called, which happens while the Payload config is being built. That broke payload generate:types and generate:importmap in environments without runtime variables — such as CI that builds payload-types to publish as a separate package. Validation now runs when an adapter method is actually invoked, so config-time codegen no longer requires runtime credentials. Valid configurations behave exactly as before.

  Closes #64.

• Updated dependencies [d8fea7a]:

  • payloadcms-vectorize@1.0.1

@payloadcms-vectorize/mongodb@1.0.1

Patch Changes

• #65 d8fea7a Thanks @techiejd! - Defer adapter config validation to call-time instead of throwing at construction.

  The mongodb and cf adapter factories previously threw on missing config (e.g. uri, dbName, binding) the moment they were called, which happens while the Payload config is being built. That broke payload generate:types and generate:importmap in environments without runtime variables — such as CI that builds payload-types to publish as a separate package. Validation now runs when an adapter method is actually invoked, so config-time codegen no longer requires runtime credentials. Valid configurations behave exactly as before.

  Closes #64.

• Updated dependencies [d8fea7a]:

  • payloadcms-vectorize@1.0.1

@payloadcms-vectorize/cf@1.0.1

Patch Changes

• #65 d8fea7a Thanks @techiejd! - Defer adapter config validation to call-time instead of throwing at construction.

  The mongodb and cf adapter factories previously threw on missing config (e.g. uri, dbName, binding) the moment they were called, which happens while the Payload config is being built. That broke payload generate:types and generate:importmap in environments without runtime variables — such as CI that builds payload-types to publish as a separate package. Validation now runs when an adapter method is actually invoked, so config-time codegen no longer requires runtime credentials. Valid configurations behave exactly as before.

  Closes #64.

• Updated dependencies [d8fea7a]:

  • payloadcms-vectorize@1.0.1

payloadcms-vectorize@1.0.0

Minor Changes

• #57 19ea3a0 Thanks @techiejd! - Add optional rerank callback on EmbeddingConfig for per-pool reranking. When configured, the plugin fetches Math.floor(limit * multiplier) candidates from the adapter, hands them to the user-supplied callback ((query, results) => Promise<results>), and trims the callback's output back down to the caller's limit. Provider-agnostic — bring your own Voyage / Cohere / local cross-encoder. multiplier must be a finite number >= 1; invalid configs are rejected at plugin init. Callback errors propagate to the caller.

@payloadcms-vectorize/pg@1.0.0

Minor Changes

• #57 19ea3a0 Thanks @techiejd! - Add optional rerank callback on EmbeddingConfig for per-pool reranking. When configured, the plugin fetches Math.floor(limit * multiplier) candidates from the adapter, hands them to the user-supplied callback ((query, results) => Promise<results>), and trims the callback's output back down to the caller's limit. Provider-agnostic — bring your own Voyage / Cohere / local cross-encoder. multiplier must be a finite number >= 1; invalid configs are rejected at plugin init. Callback errors propagate to the caller.

Patch Changes

• Updated dependencies [19ea3a0]:
  • payloadcms-vectorize@1.0.0

@payloadcms-vectorize/mongodb@1.0.0

Minor Changes

• #57 19ea3a0 Thanks @techiejd! - Add optional rerank callback on EmbeddingConfig for per-pool reranking. When configured, the plugin fetches Math.floor(limit * multiplier) candidates from the adapter, hands them to the user-supplied callback ((query, results) => Promise<results>), and trims the callback's output back down to the caller's limit. Provider-agnostic — bring your own Voyage / Cohere / local cross-encoder. multiplier must be a finite number >= 1; invalid configs are rejected at plugin init. Callback errors propagate to the caller.

Patch Changes

• Updated dependencies [19ea3a0]:
  • payloadcms-vectorize@1.0.0

@payloadcms-vectorize/cf@1.0.0

Minor Changes

• #57 19ea3a0 Thanks @techiejd! - Add optional rerank callback on EmbeddingConfig for per-pool reranking. When configured, the plugin fetches Math.floor(limit * multiplier) candidates from the adapter, hands them to the user-supplied callback ((query, results) => Promise<results>), and trims the callback's output back down to the caller's limit. Provider-agnostic — bring your own Voyage / Cohere / local cross-encoder. multiplier must be a finite number >= 1; invalid configs are rejected at plugin init. Callback errors propagate to the caller.

Patch Changes

• Updated dependencies [19ea3a0]:
  • payloadcms-vectorize@1.0.0

payloadcms-vectorize@0.7.4

Patch Changes

• #55 b7862c7 Thanks @techiejd! - Fix: reorder the vectorize task so existing chunks are deleted only after toKnowledgePool + embeddings succeed. Previously, a transient embedding-provider failure during re-vectorization would wipe a document's existing chunks before the new ones were ready, leaving the document temporarily unsearchable. Also documents a first-class Localization (i18n) pattern in the README.

@payloadcms-vectorize/pg@0.7.4

Patch Changes

• #55 b7862c7 Thanks @techiejd! - Fix: reorder the vectorize task so existing chunks are deleted only after toKnowledgePool + embeddings succeed. Previously, a transient embedding-provider failure during re-vectorization would wipe a document's existing chunks before the new ones were ready, leaving the document temporarily unsearchable. Also documents a first-class Localization (i18n) pattern in the README.

• Updated dependencies [b7862c7]:

  • payloadcms-vectorize@0.7.4