AI Model License check — check model licenses free

Flags open-weight AI model IDs in your code against their real licence clauses: EU-domicile exclusion, 700M MAU cap, non-commercial weights, missing Built with Llama attribution. Runs entirely in your browser — nothing is uploaded.

"""Image + chat pipeline for the customer-facing product."""

import torch
from diffusers import StableDiffusionPipeline
from transformers import AutoModelForCausalLM
from peft import PeftModel

# chat
CHAT_REPO = "meta-llama/Llama-3.2-11B-Vision-Instruct"
CODE_REPO = "mistralai/Codestral-22B-v0.1"
GUARD_REPO = "google/gemma-2-9b-it"   # open source, same as our Apache-2 stack

# images
IMAGE_REPO = "black-forest-labs/FLUX.1-dev"
FALLBACK_IMAGE = "runwayml/stable-diffusion-v1-5"
UPSCALER = "stabilityai/stable-diffusion-xl-refiner-1.0"

# experiments, research-only for now
LAB_REPO = "mistralai/Mistral-Large-Instruct-2411"

def load_chat():
    base = AutoModelForCausalLM.from_pretrained(CHAT_REPO, torch_dtype=torch.float16)
    tuned = PeftModel.from_pretrained(base, "acme/support-agent-lora")
    return tuned.merge_and_unload()

def load_images():
    return StableDiffusionPipeline.from_pretrained(IMAGE_REPO)

def serve():
    return {"chat": load_chat(), "images": load_images(), "code": CODE_REPO}

as of

$29 once, one licence key per person or team seat. Outside technology counsel bills roughly $350-600 an hour for an open-source and model-licence review.

Free here: Scans the file or workspace you have open and names, for every open-weight model ID it finds, the exact licence clause that binds you and the case that breaks it.. In the editor: Exports the evidence you can hand to procurement or counsel: a generated NOTICE/attribution file and a dated per-model obligation report for the whole workspace, plus a CI exit code. — that part asks for a licence key.

Same engine as the VS Code extension, byte for byte.