Retrievers
Document Retriever
A Document retriever, which searches within your knowledge sources.
Category: Retrievers · Plans: Trial, Starter, Pro, Premium, Enterprise · Type: DocumentRetriever
Inputs#
| Field | Type | Required | Default | What it does |
|---|---|---|---|---|
| Input Query | Text | Yes | {input} | The query to search in knowledge sources. |
| Document Categories | Multiple choice | No | — | Select categories of documents to search in. |
| Category IDs | Text | No | (empty) | Category IDs (comma-separated) from upstream flow. |
| Schedules | Multiple choice | No | — | Filters just documents from selected schedules. If empty, all documents from workspace are used. |
| Threshold | Number | No | 0.8 | |
| Documents Count | Number | Yes | 3 | |
| From H1 if exists | Yes/No | No | On | Start from the H1 tag of the document if H1 exists. |
| Skip Last Header | Yes/No | No | On | Skip the last header of the document to optimize output. |
| Include Metadata | Text | No | — | Include metadata in the output if available. Options: Author, Product, BreadcrumbList, VideoObject, BlogPosting, FAQPage, WebSite, opengraph |
| Export Content | Text | No | — | Content types to include in the output. Options: H1, H2, H3, H4, H5, H6, Paragraph |
| Load from pointer | Yes/No | No | On | Load from the pointer best matching the input query in the document, otherwise load all |
| Max Tokens | Number | No | 3000 | The maximum number of tokens in the output text. |
| Strategy | Text | No | Include equal size from each documents | The strategy to use for transforming the knowledge document to text. Options: Concat documents, fill from first up to tokens limit, Include equal size from each documents |
| Title | Text | No | Related resources | Title of the block in the output. |
| URL Filter | Text | No | (empty) | Filter documents by matching only pages with given substring in the URL. |
| Hide Resources | Yes/No | No | Off | Hide resources that were retrieved by the retriever. |
| Enable Document Links | Yes/No | No | Off | When enabled, users can click on cited documents to download/view them. Only affects uploaded files. (Advanced) |
Outputs#
| Output | Type |
|---|---|
| Documents | Any |
| Document Count | Any |
| Hide Resources | Any |
Use as an agent tool#
Tool name: documentretriever
A Document retriever, which searches within your knowledge sources.