Moving Froward Approach

This is a complex transition from a "Passive Viewer" to an "Active State Machine." To create the most accurate prompt for Google Antigravity, I need to clarify two specific UI/UX preferences.

My Questions for You

1. The "Map" Visualization: When the user sees their progress, do you want:
   • Option A: A simple List/Card View (e.g., "✅ Gravity", "✅ Orbit", "🔒 Relativity").
   • Option B: An interactive Network Graph (using streamlit-agraph or pyvis) where nodes light up gold as they are mastered? (Option B looks cooler but is harder to implement perfectly in Streamlit.)
2. The "Stitching" Phase: Since your current ingestion just has a list of concepts, we need to create the connections (Prerequisites). Do you want the app to:
   • Auto-Stitch: Have the LLM read all your nodes right now and guess the order? (Easier, takes ~5 mins).
   • Manual Mode: Give you a drag-and-drop UI to link them yourself? (Harder to build, more control).

---

The Updated "Mega-Prompt" (Adaptive Edition)

Copy and paste this into Google Antigravity:

codeText

Act as a Senior Python Full-Stack Developer. We are building "The Professor," a local-first adaptive AI tutor.

### CONTEXT
I have `db_manager.py` (FalkorDBLite), `app.py` (Streamlit), and `prompts.py` (LLM instructions).

### THE "KILLER FEATURE": ADAPTIVE CURRICULUM
The app must track student performance. If a student fails a concept twice, the system must **physically alter the Knowledge Graph** to make the learning curve shallower.

### STEP 1: DATABASE UPDATES (`db_manager.py`)
Update the schema and functions:

1. **Schema Change:** The `[:MASTERED]` and `[:ATTEMPTED]` relationships need a property `failure_count`.
2. `record_attempt(user_id, concept_name, passed: bool)`:
   - If `passed`: Create `[:MASTERED]` edge. Reset `failure_count`.
   - If `failed`: Increment `failure_count` on the `[:ATTEMPTED]` edge.
   - **Return:** The new `failure_count`.
3. `refactor_graph(concept_name, new_sub_concepts_list)`:
   - **Critical Logic:** Insert new nodes *between* the current prerequisites and the target concept.
   - Logic:
     1. Find target concept `(Target)`.
     2. Find immediate prerequisites `(Pre)`.
     3. Create new nodes `(Sub1)`, `(Sub2)` from the list.
     4. Delete edge `(Pre)-[:PREREQUISITE]->(Target)`.
     5. Create edges `(Pre)->(Sub1)->(Sub2)->(Target)`.

### STEP 2: PROMPTS (`prompts.py`)
Add `get_decomposition_prompt(concept, user_mistakes)`:
- Task: "The user is failing to understand '{concept}'. Based on their mistakes, break this concept down into 2-3 simpler, foundational sub-concepts that bridge the gap."
- Output: JSON list of new concepts with definitions.

### STEP 3: UI LOGIC (`app.py`)
Update the "Feedback" state logic:

- When User clicks "Submit Test":
  1. Call Grader LLM.
  2. If **PASS**: Show fireworks, unlock next node.
  3. If **FAIL**:
     - Call `db.record_attempt()`.
     - Check: Is `failure_count >= 2`?
     - **IF YES (Trigger Adaptation):**
       - Show UI Message: "I see this is tricky. Let me break this down for you..."
       - Call LLM (`get_decomposition_prompt`).
       - Call `db.refactor_graph()` with the new nodes.
       - Rerun `st.experimental_rerun()` to reload the syllabus. The "Next Lesson" will now be one of the new sub-concepts.
     - **IF NO:**
       - Show UI Message: "Let's try explaining it again."

### EXECUTION
Implement the database logic first, then the prompts, then the UI state machine. Ensure the Cypher queries for refactoring the graph handle edge deletions safely.

---

How this works in practice

1. Scenario: You are learning "Quantum Entanglement."

2. Action: You fail the test twice.

3. The Adaptation:

   • The App pauses and says: "I see this is tricky. Let me break this down..."

   • Behind the scenes: It asks Qwen-72B, "Break down Entanglement."

   • Qwen says: "They need to understand 'Spin' and 'Superposition' first."

   • FalkorDB Update: It inserts "Spin" and "Superposition" into your graph before "Entanglement."

4. The Result: The screen refreshes. Your "Next Lesson" is no longer "Entanglement." It is now "Electron Spin."

This is the feature that makes the app feel magical. Go for it.