🎙️ Voice Activation
Click the microphone to record a drawing command sequence. Try stacking multiple shapes or drawing complex paths!
Ready to listen...
📊 Execution Telemetry
Raw Transcription:
None
Interpretation Status:
IDLE
Parsed Parameters (JSON):
🔄 Policy Feedback Loop
Reward good outcomes or correct bad ones. Corrections now support edit → preview/sample execution on the canvas → explicit +1.0 reinforcement to MemRL only after you approve the result.
Need to correct the interpretation? Edit the JSON in the box above, then use the controls below for sample execution preview before MemRL update.
💡 Inspiration Sandbox
Click any example command to run it in the telemetry box, or speak it to train the agent:
💾 SQLite Episodic Memory
Live viewer of the `episodic_memory` SQLite database table. Shows Q-values for learned transcription-to-action paths.
| Query (Voice Transcription) | Resolved Actions | Q-Value |
|---|---|---|
| No learned memories yet. Use voice activation to begin. | ||