AI Generator protokołów spotkań i treści
Przekształć załączniki ze spotkań w praktyczne podsumowania, maile follow-up oraz wpisy na bloga dzięki AI. Ten workflow automatyzuje tworzenie treści z plików ...
3 min czytania
File Retriever Component
The File Retriever is a flow component designed to extract and convert the contents of files into structured documents, making them accessible for further processing in your AI workflows. This component is especially useful when you need to integrate knowledge from user-uploaded files or documents as part of your AI pipeline, such as for document analysis, summarization, or retrieval-augmented generation tasks.
What the Component Does
At its core, the File Retriever takes one or more file attachments and processes them to extract their textual content, transforming them into documents that can be used downstream in your workflow. It offers advanced options such as Optical Character Recognition (OCR) for image-based documents, customizable extraction strategies, and output token control.
Key Inputs
| Input Name | Description | Type | Default / Options | Advanced |
|---|---|---|---|---|
| Files | Files to convert into documents. | FlowSessionAttachmentResponse | — (User uploads or provides files) | No |
| Apply OCR | Apply OCR to extract text from image-based documents. Useful for scanned PDFs, images. | Boolean | false | Yes |
| Max Tokens | Maximum number of tokens in the output text. Controls the size/length of extracted text. | Integer | 3000 | Yes |
| Strategy | Strategy for transforming documents: - Concat documents, fill from first up to tokens limit - Include equal size from each documents | String (Dropdown) | Include equal size from each documents | Yes |
| Tool Name | Optional name to refer to this tool in agent-based workflows. | String | — | Yes |
| Tool Description | Optional description to help agents understand how to use this tool. | String (Multiline) | — | Yes |
| Verbose | Whether to print verbose output (for debugging or detailed logs). | Boolean | false | Yes |
Outputs
The component provides multiple outputs to suit different downstream needs:
| Output Name | Type | Description |
|---|---|---|
| Documents | Message | Processed documents as message objects, ready for workflow use. |
| Raw Documents | Document | The raw extracted documents, giving you direct access to content. |
| Tool | Tool | The documents made available as a tool for agent-based workflows. |
Why Use the File Retriever?
- Seamless Integration of File-Based Knowledge: Effortlessly bring content from uploaded files (PDFs, images, text, etc.) into your AI workflows.
- OCR Support: Automatically extracts text from images or scanned documents, expanding the range of usable file types.
- Flexible Extraction Strategies: Choose between concatenating documents or distributing content equally, with token limits to fit downstream model constraints.
- Agent-Ready: Optional fields for tool naming and description make it easy to reference in agent-driven processes.
- Customizable and Transparent: Advanced options for verbose output and token control help with debugging and optimizing workflow performance.
Typical Use Cases
- Knowledge Ingestion: Transforming user-provided documents into structured data for AI models.
- Document Analysis: Preparing documents for summarization, search, or question-answering tasks.
- Agent Tooling: Enabling AI agents to access documents as tools with clear descriptions and references.
This component provides a robust, flexible foundation for incorporating external file content into your AI-driven processes, enhancing the intelligence and adaptability of your workflows.
