src/models/provider.py

from __future__ import annotations

import os


def generate_with_hosted_model(prompt: str) -> tuple[str | None, str | None]:
    token = os.getenv("HF_TOKEN", "").strip()
    model_id = os.getenv("HF_MODEL_ID", "Qwen/Qwen3-Coder-30B-A3B-Instruct").strip()
    if not token:
        return None, "HF_TOKEN is not configured."
    try:
        from huggingface_hub import InferenceClient

        client = InferenceClient(model=model_id, token=token)
        response = client.chat.completions.create(
            messages=[
                {
                    "role": "system",
                    "content": "You are MM1's external under-32B hosted model. Use only the sanitized task. Do not ask for private profile context.",
                },
                {"role": "user", "content": prompt},
            ],
            max_tokens=700,
        )
        content = response.choices[0].message.content
        return content, None
    except Exception as exc:
        return None, f"Hosted model call failed: {exc.__class__.__name__}: {exc}"


def provider_status() -> dict[str, str]:
    return {
        "MM1_MODE": os.getenv("MM1_MODE", "local"),
        "MODEL_PROVIDER": os.getenv("MODEL_PROVIDER", "hf"),
        "HF_MODEL_ID": os.getenv("HF_MODEL_ID", "Qwen/Qwen3-Coder-30B-A3B-Instruct"),
        "MOCK_MODE": os.getenv("MOCK_MODE", "false"),
        "CODEX_RUNTIME_MODE": os.getenv("CODEX_RUNTIME_MODE", "true"),
        "HACKATHON_COMPLIANT_MODE": os.getenv("HACKATHON_COMPLIANT_MODE", "false"),
    }