Knowledge Base & AI Autopilot

Autopilot AI Training Subsystem

1. Overview

The Autopilot dashboard allows managers to upload structural training files to optimize your autopilot's response context. The underlying automation parses these training inputs, enabling the AI to generate accurate, natural, human-like answers based on your business data.

2. Training Data Table

Tracks system status across five descriptive column attributes:

• ID: Unique systemic hash indexes mapping individual document sources.
• Name: The raw filename string metadata identifier (e.g., document.pdf).
• Type: Valid mime-type structural identifiers (e.g., application/pdf).
• Language: The designated language profile parameter (e.g., English, Arabic).
• Status: Tracks file processing lifecycle states using distinct visual badges:
  • ✅ : Document processing complete; vector embeddings loaded into model memory.
  • 🔄 : Document undergoing background ingestion or extraction analysis.
• Creation Date: Absolute timestamp logging file registration events (e.g., March 20, 2025).
• Action Buttons: Access controls to run execution scripts for Delete or Regenerate data triggers.

3. Document Ingestion Procedures

1. Click the green Add + action button located in the upper right.
2. In the Upload knowledge base file configuration dialog, select your document's primary language context from the dropdown menu.
3. Drop your files directly onto the dropzone container labeled Drop the file here or click to browse.
4. Review document processing parameters, verify file sizes in bytes, confirm filename string indicators, and press the green Create button to start ingestion.

Knowledge Base Document Catalogs

The Knowledge Base section is divided into two primary tabs:

Tab A: Question Source

Tracks structural external references and scraped data URLs used to populate platform memory:

• Adding Sources: Clicking Add opens an asset dialog to register external content fields or target web domains.
• Data Properties Table: Organizes source material across explicit columns including Id, Name, target web URL, source Type (e.g., URL), data ingestion lifecycle Status, and Action tools to trigger background content recalculations.

Tab B: Generated Q&A

Houses explicit, structured question-and-answer pairs that bypass automated crawling layers.

• Manual Entry Creation: Clicking Create a new knowledge base launches a structured text input form containing clear text areas for a Question and its matching Answer.
• Management Grid: Organizes individual QA items into query tables detailing unique numeric IDs, question contexts, answer descriptions, and specific tools to modify or prune row entries.

Shared FAQs Repositories

The FAQ system features a dual-pane workspace designed to customize user-facing answer trees:

Pane 1: FAQs Source

Renders verification tracking logs for external source documents or scraped web links dedicated to powering client-facing FAQ components. Includes action parameters to manually trigger script recalculations.

Pane 2: Generated FAQs

• Search Filters Layout: Combines textual entry fields with structured language configuration menus (Arabic vs English).
• Adding New FAQ Records: Clicking the Add asset button prompts a specialized multi-tier composition dialog box (Add Question and Answer).
• Multi-Tier Answer Architecture: The interface supports a primary question input field (Question), a main resolution field (Answer 1), and an optional secondary fallback resolution field (Answer 2), enforcing an absolute structural system limit of two sub-answers per index entry.

Clicking the green Create action button registers your FAQ components and publishes them instantly to your client-facing widgets.