Skip to main content

The Ingestion Phase

The Documents panel serves as a centralized, localized repository for user-generated assets and external data sources. Instead of managing disjointed conversational files, this module functions as an integrated file management layout that feeds content directly into your local vector space to anchor downstream agent reasoning.

📂 Interface Layout Views

The platform provides two distinct visual layouts to audit and sort your data assets.
The default view of the Documents panel displays your indexed files as organized, MIME-type categorized folders. This layout emphasizes a high-level overview of your data inventory, allowing you to quickly assess the types and quantities of files currently ingested into your local memory vault.
Default empty state view of the documents tab

MIME-Type Asset Filters

Assets are parsed dynamically into distinct, count-badged categorization headers:
  • All: The total aggregation of all files in the repository.
  • PDF: Document Format files.
  • Office: Professional suites including Word documents (.docx), spreadsheets (.xlsx), and presentations (.pptx).
  • Text: Plain text files and Markdown documents.
  • Code & Data: Source code files, raw data arrays, and structured logs (e.g., .json, .csv).
  • Images: Image formats and vision-processed graphic assets.

⚙️ Row-Level Target Actions

Each item in the document list view exposes interactive control icons on the right-hand margin of its row to modify metadata or open source files.

1. Memory Space Assignments

Clicking the secondary icon on the right margin opens an inline context drawer displaying your Memory Spaces. This menu maps out exactly which active sandboxes inherit the file’s knowledge payload.
Inline view of Memory Space selection assignments

2. Contextual Options Menu

Clicking the final vertical-dots menu icon on the right-hand margin exposes immediate contextual utility operations for the file:
Contextual vertical dots dropdown menu options
  • Folder Link Icon (Open Source Location): Opens your operating system’s native file manager (macOS Finder or Windows File Explorer) directly to the specific local directory where the file is physically stored.
  • Rename: Edits the document’s metadata string title inside the index without altering the source file name on disk.
  • Delete: Permanently wipes the file entry and its compiled vector embeddings from your local memory bank.

Local Folder Automation: You can sync and track entire local folders seamlessly by configuring target watch paths using the directory monitoring systems found inside your General App Settings.

Next Step: Phase 2 — Persona Initialization

Now that your core documents are securely ingested into the local repository, proceed to configure behavioral archetypes and execution constraints.