Planner Model Test not loaded

Tests Akshayram1/omnibrowser-planner-1p5b-q4f16_1-MLC — your fine-tuned Qwen2.5-1.5B quantized model — by loading it via WebLLM and checking that it outputs valid AgentAction JSON. Requires a browser with WebGPU (Chrome 113+ / Edge 113+).

Step 0 — Check model files

Verifies that the required MLC files exist and are publicly accessible on HuggingFace before attempting the full download.

—

Step 1 — Load model

Akshayram1/omnibrowser-planner-1p5b-q4f16_1-MLC

Idle — click Load to begin

Step 2 — Test generation

User goal (sent to the planner)

Quick goal examples

Raw model output

— load the model first —

Parsed result

—

Step 3 — Self-test (5 prompts)

Runs five built-in planner prompts and checks each produces a valid AgentAction JSON.

—