Skip to main content
Version: 2025.6

Research Archive (Vector Stores)

The Research Archive is a knowledge base system that uses vector stores to enable AI Chatbots to provide context-aware, accurate responses based on your organization's documents.

Vector stores convert your documents into searchable embeddings, allowing chatbots to quickly retrieve relevant information when grounding.filesearch is enabled.

Viewing Your Archives​

Navigate to Backoffice > Reflectis AI > Research Archives to see all available archives.

bo_vector_store_list

Each archive card displays:

  • Archive name
  • Total files – total number of files uploaded
  • Completed – successfully processed files (green number)
  • Usage – storage size in KB
  • Created – creation date
  • Status badge:
    • Ready – all files processed successfully
    • With Errors – some files failed processing

Creating a New Archive​

  1. Click Create New Research Archive (top right)
  2. Provide a name for your archive
  3. (Optional) Upload initial files or add them later
  4. Click Create

Archive Detail View​

Click any archive card to open its detail page.

bo_vector_store_detail

Information Panel​

Displays:

  • ID – unique vector store identifier (copyable)
  • Name – archive name
  • Status – current status (Ready / With Errors)
  • Usage – total storage used
  • Created – creation timestamp

Processing Status​

A real-time overview of file processing:

  • Total (blue) – total files
  • Completed (green) – successfully vectorized
  • In progress (yellow) – currently processing
  • Failed (red) – failed uploads/processing

File Management​

The table shows all uploaded files with:

  • File Name – file name and format icon
  • Status – Completed (ready) or error indicator
  • ID – file-specific identifier
  • Size – file size
  • Upload Date – upload timestamp
  • Actions – delete button (trash icon)

Uploading Files​

  1. Open the archive detail page
  2. Click Upload File (bottom right)
  3. Select one or more files from your computer
  4. Wait for processing:
    • Files appear immediately in the table with status In progress
    • Once vectorization completes, status changes to Completed
    • Failed files show an error indicator

Supported File Types​

  • PDF documents
  • Microsoft Word (.docx)
  • Plain text (.txt)
  • Markdown (.md)
  • Other formats may be supported depending on configuration

Managing Archives​

Updating the AI System​

After uploading or deleting files, click Update AI System to synchronize changes with active chatbots.

Deleting Files​

  • Click the trash icon in the file row to remove a specific file
  • The file is removed from the vector store immediately

Deleting an Archive​

  • Click Delete (top right) to delete the entire archive
  • ⚠️ This action removes all files and cannot be undone
  • Any chatbots using this archive will lose access to its content

Linking Archives to Chatbots​

Once your archive status is Ready:

  1. Go to Chatbot configuration
  2. Enable grounding.filesearch (or "Enable Research Archive")
  3. Select this archive from the dropdown
  4. Save

The chatbot will now query this archive when answering user questions.

Best Practices​

  • Use readable documents: ensure files contain actual text, not just scanned images. PDFs should be text-based or OCR-processed. Avoid corrupted or image-only files that cannot be parsed.
  • Organize by domain: create separate archives for different topics (e.g., HR Policies, Product Manuals, Training Materials)
  • Monitor processing status: check that files reach Completed before linking to chatbots
  • Handle errors: if files show Failed, check format/size and re-upload
  • Use clear file names: helps you identify content later
  • Keep archives updated: remove outdated documents and upload new versions
  • Test with chatbots: verify that answers reflect your documents accurately

Troubleshooting​

IssueSolution
File stuck in In progressWait a few minutes; large files take longer. If it persists, try re-uploading.
File marked FailedCheck file format and size. Try converting to PDF or splitting large documents.
Chatbot not using archive contentEnsure grounding.filesearch is ON and archive status is Ready. Click Update AI System.
Archive shows With ErrorsOpen detail view, check failed files, fix and re-upload.