app.py

import gradio as gr
from huggingface_hub import InferenceClient
from typing import List, Dict, Optional

# Your endpoint root (no trailing /v1 here; the client adds it for chat)
ENDPOINT_URL = "https://x6leavj4hgm2fdyx.us-east-2.aws.endpoints.huggingface.cloud/v1/"
def respond(
    user_msg: str,
    history: List[Dict[str, str]],
    system_message: str,
    max_tokens: int,
    temperature: float,
    top_p: float,
    hf_token: Optional[gr.OAuthToken],   # from LoginButton (kept)
    pat_override: str,                   # NEW: user-pasted PAT (password field)
):
    """
    Use PAT override if provided; otherwise fall back to LoginButton token.
    NOTE: OAuth token from LoginButton usually lacks `inference.endpoints.infer.write`,
    so for Inference Endpoints you almost always need to paste a PAT here.
    """

    # Choose a token: prefer user-supplied PAT with endpoints write scope
    token = pat_override.strip() or (getattr(hf_token, "token", None) if hf_token else None)
    if not token:
        yield "🔒 Please click **Login** OR paste a **Hugging Face PAT** with `inference.endpoints.infer.write`."
        return

    client = InferenceClient(base_url=ENDPOINT_URL, token=token)

    # Build messages (OpenAI-style)
    messages = []
    if system_message:
        messages.append({"role": "system", "content": system_message})
    messages.extend(history or [])
    messages.append({"role": "user", "content": user_msg})

    # Try OpenAI-compatible chat route first: /v1/chat/completions
    try:
        out = ""
        for chunk in client.chat_completion(
            messages=messages,
            max_tokens=max_tokens,
            temperature=temperature,
            top_p=top_p,
            stream=True,
        ):
            tok = ""
            if getattr(chunk, "choices", None) and getattr(chunk.choices[0], "delta", None):
                tok = chunk.choices[0].delta.content or ""
            out += tok
            yield out
        return
    except Exception as e_chat:
        chat_err = str(e_chat)

    # Fallback to plain generation (for non-OpenAI runtimes)
    try:
        def to_prompt(msgs: List[Dict[str, str]]) -> str:
            lines = []
            for m in msgs:
                role = m.get("role", "user")
                content = m.get("content", "")
                tag = {"system": "SYSTEM", "user": "USER"}.get(role, "ASSISTANT")
                lines.append(f"[{tag}] {content}")
            lines.append("[ASSISTANT]")
            return "\n".join(lines)

        prompt = to_prompt(messages)
        out = ""
        for tok in client.text_generation(
            prompt,
            max_new_tokens=max_tokens,
            temperature=temperature,
            top_p=top_p,
            stream=True,
            return_full_text=False,
        ):
            piece = getattr(tok, "token", tok)
            if isinstance(piece, dict) and "text" in piece:
                piece = piece["text"]
            out += str(piece)
            yield out
    except Exception as e_gen:
        yield (
            "❗ Endpoint call failed.\n\n"
            f"• Chat API error: {chat_err}\n"
            f"• Text-generation fallback error: {e_gen}\n\n"
            "Most likely cause: the token used does NOT have `inference.endpoints.infer.write`.\n"
            "Paste a PAT with that scope in the sidebar."
        )

# --- UI ---
chat = gr.ChatInterface(
    respond,
    type="messages",
    additional_inputs=[
        gr.Textbox(value="You are a friendly Chatbot.", label="System message"),
        gr.Slider(1, 4096, value=512, step=1, label="Max new tokens"),
        gr.Slider(0.0, 4.0, value=0.7, step=0.1, label="Temperature"),
        gr.Slider(0.0, 1.0, value=0.95, step=0.05, label="Top-p"),
        # NEW: secure PAT override
        gr.Textbox(value="", label="HF PAT (with `inference.endpoints.infer.write`)", type="password"),
    ],
)

with gr.Blocks() as demo:
    with gr.Sidebar():
        gr.Markdown("### Hugging Face Login (optional)")
        gr.LoginButton()
        gr.Markdown(
            "**Important:** Inference Endpoints require a PAT with\n"
            "`inference.endpoints.infer.write`. The Login token usually does **not** have this.\n"
            "Paste a PAT in the password field if you see 403 errors."
        )
        gr.Markdown(f"**Endpoint**: `{ENDPOINT_URL}`")
    chat.render()

if __name__ == "__main__":
    demo.launch()