app.py

import os
import json
import re
from pathlib import Path
from fastapi.responses import HTMLResponse
from fastapi.staticfiles import StaticFiles
import gradio as gr
from openai import OpenAI

# Default system prompt for DeepSeek V4 Flash
DEFAULT_SYSTEM_PROMPT = "You are DeepSeek, a helpful AI assistant powered by DeepSeek V4 Flash. You provide accurate, detailed, and thoughtful responses."

# ── Initialize Gradio Server (FastAPI subclass) ─────────────────────────
app = gr.Server()

# Create static directory if it doesn't exist
STATIC_DIR = Path(__file__).parent / "static"
os.makedirs(STATIC_DIR, exist_ok=True)


# ── API Endpoint: Chat ──────────────────────────────────────────────────
@app.api(name="chat_with_deepseek")
def chat_with_deepseek(
    messages_json: str,
    reasoning_effort: str = "medium",
    max_tokens: str = "2048",
    temperature: str = "0.7",
    system_prompt: str = "",
) -> str:
    try:
        messages = json.loads(messages_json)
        max_tokens = int(max_tokens)
        temperature = float(temperature)

        key = os.environ.get("SILICONFLOW_API_KEY", "").strip()
        if not key:
            return json.dumps({
                "status": "error",
                "message": "SILICONFLOW_API_KEY environment variable is not configured on the server."
            })

        client = OpenAI(
            api_key=key,
            base_url="https://api.siliconflow.com/v1",
        )

        sys_prompt = system_prompt.strip() if system_prompt and system_prompt.strip() else DEFAULT_SYSTEM_PROMPT
        messages_no_system = [m for m in messages if m.get("role") != "system"]
        final_messages = [{"role": "system", "content": sys_prompt}] + messages_no_system

        params = {
            "model": "deepseek-ai/DeepSeek-V4-Flash",
            "messages": final_messages,
            "max_tokens": max_tokens,
            "temperature": temperature
        }

        if reasoning_effort in ["low", "medium", "high"]:
            params["reasoning_effort"] = reasoning_effort

        response = client.chat.completions.create(**params)

        content = response.choices[0].message.content
        reasoning_content = getattr(response.choices[0].message, "reasoning_content", "")

        if not reasoning_content and content and "<think" in content and "</think" in content:
            think_match = re.search(r'<think[^>]*>(.*?)</think', content, re.DOTALL)
            if think_match:
                reasoning_content = think_match.group(1).strip()
                content = re.sub(r'<think[^>]*>.*?</think\s*>', '', content, flags=re.DOTALL).strip()

        return json.dumps({
            "status": "success",
            "content": content,
            "reasoning_content": reasoning_content or ""
        })

    except Exception as e:
        return json.dumps({
            "status": "error",
            "message": str(e)
        })


# ── Serve the main HTML page ────────────────────────────────────────────
@app.get("/")
async def homepage():
    html_path = STATIC_DIR / "index.html"
    if html_path.exists():
        with open(html_path, "r", encoding="utf-8") as f:
            return HTMLResponse(content=f.read(), status_code=200)
    return HTMLResponse(
        content="<h1>Frontend is building. Please refresh in a few seconds...</h1>",
        status_code=200
    )


# ── Mount static folder for CSS, JS, and image assets ──────────────────
app.mount("/static", StaticFiles(directory=str(STATIC_DIR)), name="static")


# ── Launch ──────────────────────────────────────────────────────────────
if __name__ == "__main__":
    app.launch(show_error=True)