Extract Tags Processor #
Extracts structured keyword tags from a document’s content using a language model.
Documents that have not yet been summarized are skipped. Run
document_summarization before this processor.
Requirements #
A configured language model provider is required. Set model_provider and
model in the processor config, or configure a default language model in the
application settings.
Configuration #
| Parameter | Type | Required | Default | Description |
|---|---|---|---|---|
message_field | string | No | messages | Pipeline context key for the input messages |
output_queue | object | No | null | Queue to push processed documents to |
model_provider | string | No | (app default) | Language model provider ID |
model | string | No | (app default) | Language model name |
model_context_length | int | No | — | Model context window size in tokens (minimum 4000) |
llm_generation_lang | string | No | (app default) | BCP 47 language tag for generated content (e.g. en-US, zh-CN) |
Example #
- extract_tags:
model_provider: openai
model: gpt-4o-mini
output_queue:
name: "documents_tagged"