Documents
Organize documents into groups and upload PDFs, Office files, text, images, and audio to turn them into searchable knowledge. Enrich indexing with OCR, image captions, audio transcription, and a knowledge graph.
The Documents screen is the starting point for knowledge in Flux. When you upload a file, its text is automatically extracted, split, and indexed for search so that the chatbot and agents can ground their answers in it. The groups list appears on the left, and the upload area and document list for the selected group appear on the right.
Organized into groups
Every document belongs to a group. Organize by topic, project, or team so agents can target which groups to search.
Many formats
Handle PDF, Word, PowerPoint, Excel, text, Markdown, HTML, image, and audio files in one place.
Automatic enrichment
Image captions, audio transcription, and knowledge graph extraction index details that plain text alone would miss.
Search ready
Once processed, a document becomes searchable by the chatbot and agents and is cited as a source in answers.
Document groups
A group is a folder that holds documents. To upload anything, you first create and select a group. Each group can have its own sharing scope and knowledge graph settings, which makes it easy to keep knowledge separated by team or topic.
Create a group
Use the new group button in the groups list to enter a name and description. A name is required; the description is optional.
Set the sharing scope
Choose whether the group is visible to the whole organization or only to specific teams. With a team-only scope, only members of those teams can search and view its documents.
Select the group
Click a group in the list to reveal the upload area and document list on the right. You can now add files to this group.
When you change a group from public to team-only, a brief undo prompt appears so you can revert. Change the sharing scope carefully — narrowing it means some members can no longer search those documents.
Upload documents
After selecting a group, drag and drop files onto the upload area or click it to choose files. You can upload several files at once, and progress appears in the panel at the bottom of the screen.
Choose or drop files
Drag files onto the upload area or click to pick them. Before uploading, Flux checks the format, extension, size, and file signature (the actual content), so files that only rename their extension are rejected.
Watch processing
Once an upload starts, the file is stored and background processing follows. When a document moves from 'processing' to 'completed', it is ready for search.
Cancel if needed
You can stop a processing document with the cancel button in the list. A cancelled document is marked 'cancelled' and can be uploaded again.
| Type | Supported formats | How it is processed |
|---|---|---|
Documents | PDF, DOCX/DOC, PPTX, XLSX | Structure is detected to extract text and tables, then split into searchable chunks. |
Text | TXT, Markdown, HTML | The body is read and indexed as-is. HTML is safely sanitized before processing. |
Images | PNG, JPG, WEBP, GIF | OCR reads any text, and automatic captions describe the image so it becomes searchable. |
Audio | MP3, WAV | Speech is transcribed to text so spoken content can be searched and cited. |
The maximum size for a single file is 1GB. The number of simultaneous uploads is limited automatically to avoid overloading the server, and remaining files are processed in order.
Processing options
Expand the processing options in the upload area to control how a file is analyzed. The defaults work well, but tuning them to a document's nature can improve results. Set these just before uploading.
OCR
Recognizes text in scans and images. Turn it on for PDFs where text cannot be selected, or for photo-style documents.
Extract images
Pulls out figures and charts embedded in a document. Extracted images can become candidates for automatic captioning.
Table analysis mode
Turn table-structure analysis off, or choose fast or accurate. Documents with many tables benefit from accurate mode.
Language
Set the document language to Korean or English, or leave it on auto-detect. Leaving it empty uses your organization's default language.
Automatic enrichment
Beyond text extraction, Flux automatically indexes image, audio, and relationship information. This enrichment continues in the background after upload, and what gets applied depends on your settings.
Automatic image captions
A vision language model analyzes images and writes captions that describe their content, so even images with no text become searchable and citable. Turn it on or off in organization or group settings.
Audio transcription
Speech in audio files is transcribed to text. Use the built-in transcription engine, or connect an external transcription service in organization settings.
Knowledge graph integration
Enable the knowledge graph on a group to extract entities such as people, organizations, and concepts and the relationships between them. Explore the result visually in the Knowledge Graph Explorer. Enabling it requires an LLM key and a configured GraphRAG model first.
Enrichment is separate from text indexing. For example, even if image captioning fails, the document itself is still indexed and available for search.
Managing documents
Each document shows its processing state as a badge. Open a document row or its menu to preview content, move it to another group, download it, or delete it.
| Status | Meaning |
|---|---|
Pending | The upload finished and the document is waiting its turn to be processed. |
Processing | The file is being analyzed and indexed. It will turn to completed shortly. |
Completed | Ready to be used in search and chat. |
Failed | Something went wrong during processing. Check the file and try uploading again. |
Cancelled | Processing was stopped by a user. Upload again to process it. |
Chunk preview
See how a document was split into chunks for search. These chunks are exactly the units the chatbot cites.
Move to group
Move a document to a different group. Once moved, it follows the new group's sharing scope and settings.
Download
Download the original file. For security, download links expire after a set time.
Delete
Removes the document together with its index. Cancel a processing document first, then delete it.
Deletion cannot be undone. Removing a document deletes both the original file and its search index, so it is no longer cited in answers that relied on it.
Frequently asked questions
No. The file is extracted to text and split into small, search-friendly units called chunks, then indexed. The chatbot retrieves only the chunks relevant to a question and uses them as grounding.
It needs processing time. Once a document's status becomes 'completed', it is used in search and chat. Large files, or files that need OCR or transcription, can take longer.
Yes. Images are turned into text via OCR and automatic captions, and audio via transcription, then indexed. As a result, even text-free pictures or recorded speech can appear in search results.
Uploads currently support document, text, image, and audio formats. Video files are not supported, so extract the audio and upload that if needed.
Both the original file and its search index are removed, and this cannot be undone. After deletion, the document is no longer cited as a source in answers.