Atlas User Manual & Operator Guide

1. Ingesting Files & Assets

• Drop files, documents, scripts, or graphics into Input/ (or drag-and-drop into the Activity Console).
• Atlas will analyze the content, synthesize Wikipedia-grade articles in Wiki/, and immediately archive sources to Output/YYYY-MM-DD/.
• Short-Turn Resilience: Large batches of files are processed in short incremental turns. Finished turns are committed to Drive immediately; if a high-demand error occurs, completed work is never repeated.

2. Activity Log Memory & Loop Protection

• If an upstream model spike occurs (e.g. Gemini 503 high demand), the Log Memory Engine checks the console logs and output archive at the start of every cycle.
• Already-archived images and processed documents are automatically recognized, preventing redundant resizing or duplicate operations.

3. Assigning Priority Tasks

• Add bulleted items to the root TODO file.
• Atlas executes tasks in TODO before ingesting files in Input, erasing completed tasks upon verified execution.

4. Answering Questions & Rule Escalations

• Check Questions.md. When Atlas encounters ambiguous data or potential credentials, it escalates an inquiry.
• Add your answer below A:. That answer becomes a permanent binding operational rule!

5. Google NotebookLM & Notion Sinks

• NotebookLM: Open the Integrations modal to connect the persistent Master Google Doc ([MASTER_MONOLITH] Tesseract City Wiki) or companion markdown file. The Drive pointer remains constant, enabling in-place source syncing.
• Notion: Mirror synthesized knowledge directly into Notion using semantic blocks (headers, callouts, code blocks, bulleted lists).