Documentación

    Data Administration

    Data Administration allows you to manage which documents are included when querying with the Ask feature. This gives you fine-grained control over the AI's knowledge base for each query.

    Data Administration Interface

    Purpose

    When you have many documents in your library, you may want to:

    • Focus queries on specific project documents
    • Exclude outdated or irrelevant files
    • Create document sets for different use cases
    • Improve response relevance by narrowing scope

    Accessing Data Administration

    1. Click the Data Administration tab in the navigation
    2. A table displays all documents in your vector store

    Understanding the Interface

    Document Table

    The table shows three columns:

    | Column | Description | |--------|-------------| | Checkbox | Select/deselect document for inclusion | | Key | Document identifier or filename | | Content | Preview of document content (truncated) |

    Viewing Full Content

    Document content is truncated in the table. To see more:

    1. Hover over any cell in the Content column
    2. A tooltip appears showing more of the document text
    3. Move your mouse away to close the tooltip

    Managing Document Selection

    Selecting Documents

    1. Click the checkbox next to documents you want to include
    2. Selected documents will be checked
    3. Your selection is automatically saved

    Bulk Actions

    Currently, documents must be selected or deselected individually:

    • To select multiple documents: Check each document checkbox one by one
    • To clear all selections: Uncheck each document individually

    Selection Persistence

    Your document selections are saved in your browser:

    • Selections persist across page refreshes
    • Selections persist across browser sessions
    • Each browser maintains its own selections

    Using Selections in Ask

    After selecting documents in Data Administration:

    1. Go to the Ask tab
    2. In the header, find the view mode selector
    3. Choose Custom documents
    4. Your queries will now only search selected documents

    View Modes Explained

    | Mode | Behavior | |------|----------| | All documents | Searches entire document library (ignores selections) | | Custom documents | Searches only documents selected in Data Administration | | None | Disables document context entirely |

    Use Cases

    Project-Specific Queries

    1. Upload all project documentation
    2. In Data Administration, select only current project files
    3. Use "Custom documents" mode in Ask
    4. Queries focus on relevant project content

    Excluding Outdated Content

    1. Review documents in Data Administration
    2. Uncheck outdated or deprecated files
    3. Use "Custom documents" mode
    4. AI won't reference excluded content

    Creating Document Sets

    You can organize documents into logical sets for different purposes:

    • Technical documentation — API docs, architecture guides, technical specs
    • Policy documents — Compliance, security policies, standards
    • Project specifications — Requirements, project plans, design docs
    • Meeting notes — Minutes, action items, decision records

    Select and activate the appropriate document set based on your current task or query focus.

    Tips

    • Review periodically — Check selections when adding new documents
    • Use content preview — Hover to verify document content before selecting
    • Remember view mode — Selections only apply in "Custom documents" mode
    • Clear for broad searches — Use "All documents" when you need comprehensive results

    Technology: Data Administration fetches document metadata from the /vector/records endpoint, which returns all entries in the PostgreSQL pgvector table. Selections are stored client-side using Zustand with localStorage persistence. When Ask is used with "Custom documents" mode, selected document keys are sent as a filter list to the /ask/doc endpoint, limiting the vector similarity search to those documents.