Update app.py

app.py

@@ -229,8 +229,7 @@ def get_inference_client(model_id):
    if provider == "deepseek":
        if not DEEPSEEK_API_KEY:
            raise ValueError("DEEPSEEK_API_KEY environment variable not set.")
-       from deepseek.client import DeepSeekClient
-       return DeepSeekClient(api_key=DEEPSEEK_API_KEY)
+       return "deepseek_client" # Return a string placeholder, logic is handled in generation_code
 
    if provider == "groq":
        return InferenceClient(
@@ -1087,25 +1086,38 @@ This will help me create a better design for you."""
                        sandbox: send_to_sandbox(clean_code) if language == "html" else "<div style='padding:1em;color:#888;text-align:center;'>Preview is only available for HTML. Please download your code using the download button above.</div>",
                    }
         elif _current_model["provider"] == "deepseek":
-            # Use Deepseek client
-            stream = client.chat.completions.create(
-                model=_current_model["id"],
-                messages=[{"role": m["role"], "content": m["content"]} for m in messages],
-                stream=True,
-                max_tokens=5000,
-            )
-
+            # Use requests to call Deepseek API directly
+            url = "https://api.deepseek.com/v1/chat/completions"
+            headers = {
+                "Authorization": f"Bearer {DEEPSEEK_API_KEY}",
+                "Content-Type": "application/json"
+            }
+            payload = {
+                "model": _current_model["id"].split('/')[-1], # e.g., "DeepSeek-V3-0324" -> "DeepSeek-V3-0324"
+                "messages": [{"role": m["role"], "content": m["content"]} for m in messages],
+                "stream": True,
+                "max_tokens": 5000
+            }
+            
+            response = requests.post(url, headers=headers, json=payload, stream=True)
+            response.raise_for_status()
+            
             content = ""
-            for chunk in stream:
-                if chunk.choices[0].delta.content:
-                    content += chunk.choices[0].delta.content
-                    clean_code = remove_code_block(content)
-                    # The logic here is the same as for OpenAI, can be refactored
-                    yield {
-                        code_output: gr.update(value=clean_code, language=get_gradio_language(language)),
-                        history_output: history_to_chatbot_messages(_history),
-                        sandbox: send_to_sandbox(clean_code) if language == "html" else "<div style='padding:1em;color:#888;text-align:center;'>Preview is only available for HTML. Please download your code using the download button above.</div>",
-                    }
+            for line in response.iter_lines():
+                if line:
+                    line_str = line.decode('utf-8')
+                    if line_str.startswith('data: '):
+                        json_data = line_str[6:]
+                        if json_data.strip() != '[DONE]':
+                            chunk = json.loads(json_data)
+                            if chunk['choices'][0]['delta'].get('content'):
+                                content += chunk['choices'][0]['delta']['content']
+                                clean_code = remove_code_block(content)
+                                yield {
+                                    code_output: gr.update(value=clean_code, language=get_gradio_language(language)),
+                                    history_output: history_to_chatbot_messages(_history),
+                                    sandbox: send_to_sandbox(clean_code) if language == "html" else "<div style='padding:1em;color:#888;text-align:center;'>Preview is only available for HTML. Please download your code using the download button above.</div>",
+                                }
 
         else:
             # Use Hugging Face Inference Client (or Groq) for other models