NotebookLM Research (JSON Direct Ingest)

This is the student workflow for Track 2 (Research Library). You will curate high-quality notebooks and import them directly into Writers Factory to power your Story Bible.

The Process at a Glance

1. CURATE        2. CLEANSE             3. EXPORT              4. IMPORT              5. CONNECT
──────────       ─────────────────      ─────────────────      ──────────             ──────────
Fill Notebook    Run Cleaning           Use Extension          Drag JSON to           App finds
with Sources     Protocol & Delete      to download            Writers Factory        ingredients
                 True Noise             .json file             (Research Tab)         automatically

Step 1: Curate Your Notebooks

Pick your notebook categories (e.g., “Magic System”, “World Politics”). For each notebook:

  1. Add Sources: PDFs, URLs, audio transcripts, etc.
  2. Curate for Quality: Focus on sources that genuinely fascinate you—the Codex is story-agnostic and can power multiple novels.

[!TIP] Use the Templates We provide 7 standardized notebook templates (Arena, Speculation, Voice, etc.) in the Research Notebooks section.

[!TIP] Optional: Context Note You can create a “WHY I CHOSE THIS [NOTEBOOK]” note to document your intent. See individual notebook templates for examples. After writing, convert it to a source so the AI can read it.

Step 2: Cleanup Pass (Critical)

Copy and paste this prompt into your notebook to identify sources to delete:

## Cleaning Protocol

**Role:** Developmental Editor and Data Curator.

**Objective:** Clean this notebook for a Knowledge Graph. Your goal is to remove "True Noise" while protecting "Narrative Assets" (metaphors, jargon, and sensory details), even if they seem off-topic.

**Step 1: Establish Narrative Anchors**
Scan any Context Notes and sources to lock in these three parameters:

1. **Setting:** (Where are we? e.g., UK Estate, Mars Colony).
2. **Central Metaphor:** (The lens for this notebook, e.g., "Immigration is Legacy Software").
3. **Core Conflict:** (The engine of the plot, e.g., "Systemic Integrity vs. Evolution").

**Step 2: The "Keep vs. Kill" Evaluation**
Filter all *other* sources through those anchors:

• **KEEP (The "Hidden Gems"):**
  ◦ **Lexicon:** Technical manuals or reports that provide the specific *jargon* my characters speak (e.g., software terms for a hacker protagonist).
  ◦ **Texture:** Sources that offer visceral details, slurs, or rhetoric that adds realism, even if the jurisdiction is wrong (e.g., US laws used as a model for a fictional regime).
  ◦ **Psychology:** Academic papers that explain the *motive* of the killer or the *trauma* of the victim.
  ◦ **Strategic Analysis:** Any Context Notes or audit documents you created.

• **DELETE (The "True Noise"):**
  ◦ **Conflicting Reality:** Data that contradicts the story's internal logic (e.g., biological bird migration stats if the metaphor is purely digital).
  ◦ **Functional Clutter:** Instructions, prompts, or table generation scripts (e.g., `Table_Generation_Engine.md`).
  ◦ **Dry Data:** Raw statistics or tables (GDP, crop yields) that lack narrative context or emotional weight.

**Step 3: The Output**
Provide a numbered **DELETE LIST**.
• For each file, write a 1-sentence **"Disconnect Reason"** explaining why it fails to serve the Narrative Anchors.

Action: Use the DELETE LIST to remove sources in NotebookLM before exporting. The NotebookLM Tools extension (next step) makes bulk deletion easier.

Step 3: Tool Setup (One-Time)

To manage and export your data cleanly, you need the NotebookLM Tools browser extension.

[!IMPORTANT] Install NotebookLM Tools

  1. Install the NotebookLM Tools by _trungpv extension.
  2. This tool allows for easy bulk deletion of sources (critical for Step 2B cleanup).
  3. It also allows you to export the JSON (full metadata) format required by Writers Factory.

Step 4: Export to JSON

For each cleaned notebook you want to import:

  1. Click on the NotebookLM Tools icon in the browser bar.
  2. Find and select your notebook.
  3. Click the backup/restore button.
  4. Select JSON (full metadata).
  5. Click Export Backup and save the file.

Step 5: Drag & Drop to Codex

  1. Go to Writers Factory → Architect Mode → Research Library.
  2. Drag your JSON files into the dropzone.
  3. Watch the progress bar: “Indexing to Codex”.

What happens: The system embeds every paragraph of your research into a mathematical vector space. This makes it “searchable” by concept, not just keyword.

Step 6: The Connection Point

Once you have:

  1. Imported your JSON research (The Codex)
  2. Completed your Story Development (The Story Bible)

You are ready to connect them.

  1. Go to Architect Mode.
  2. Click “Connect Research”.
  3. The system uses your Story Bible (Protagonist, Theme, World Rules) as a lens.
  4. It searches your entire Codex for relevant ingredients.
  5. It creates a focused Research Graph ~50 ingredients for this story.

Why this is better: You don’t need to manually tag connections. The AI finds that your “Submarine Manual” (Research) is relevant to your “Isolation Theme” (Story) automatically.

Summary: The Complete Workflow

CURATE → CLEANSE → EXPORT → IMPORT → CONNECT
   ↓         ↓         ↓        ↓        ↓
Build    Run Cleaning  JSON    Codex   Research
Notebook  Protocol &  Export  Indexing  Graph
          Delete Noise

Key insight: The Cleaning Protocol (Step 2) is what transforms a raw collection into a curated knowledge base. It removes data that contradicts your story while keeping “hidden gems” that add lexicon, texture, and psychology—even if they seem off-topic at first glance.