Update src/streamlit_app.py

src/streamlit_app.py

@@ -1,19 +1,14 @@
 import os, pathlib
 
 
-home = "/tmp"
-os.environ["HOME"] = home    
-cfg_dir = pathlib.Path(home) / ".streamlit"
-cfg_dir.mkdir(parents=True, exist_ok=True)
 
+os.environ["HOME"] = "/tmp"       
+cfg_dir = pathlib.Path("/tmp/.streamlit")
+cfg_dir.mkdir(parents=True, exist_ok=True)
 
 (cfg_dir / "config.toml").write_text(
-    "[server]\n"
-    "headless = true\n"
-    "port = 7860\n"
-    "address = \"0.0.0.0\"\n\n"
-    "[browser]\n"
-    "gatherUsageStats = false\n",
+    "[server]\nheadless = true\nport = 7860\naddress = \"0.0.0.0\"\n\n"
+    "[browser]\ngatherUsageStats = false\n",
     encoding="utf-8",
 )
 import streamlit as st
@@ -62,6 +57,7 @@ class ConfigManager:
         self.hf_token = os.getenv("HF_TOKEN")
         self.google_creds_json = os.getenv("GOOGLE_SHEETS_CREDENTIALS")
         self.google_sheets_id = os.getenv("GOOGLE_SHEETS_ID")
+        
 
 
         missing_vars = []
@@ -153,7 +149,7 @@ class AIAssistant:
 
         self.client = InferenceClient(
             model=self.model,
-            token=None,
+            token=token,
             timeout=60.0,
         )
 
@@ -186,59 +182,39 @@ Please follow these guidelines:
     
     def generate_response(self, button_name: str, question: str, retry_count: int = 0) -> str:
         try:
-            # Build prompts
             system_text = self.base_prompt + self.prompt_templates.get(button_name, "")
             if retry_count > 0:
                 system_text += f"\nPlease provide a different explanation. This is attempt {retry_count + 1}."
             user_text = f"Question:\n{question}"
 
-            full_prompt = f"{system_text}\n\n{user_text}"  # still used for text_generation
-
             try:
-                # Try classic text-generation first
+                messages = [
+                    {"role": "system", "content": system_text},
+                    {"role": "user", "content": user_text},
+                ]
+                chat = self.client.chat_completion(
+                    messages=messages,
+                    max_tokens=350,
+                    temperature=0.7,
+                )
+                text = ""
+                try:
+                    choices = getattr(chat, "choices", None) or chat.get("choices", [])
+                    if choices:
+                        msg0 = choices[0].get("message") or {}
+                        text = msg0.get("content") or ""
+                    if not text:
+                        text = getattr(chat, "generated_text", None) or chat.get("generated_text", "") or ""
+                except Exception:
+                    text = str(chat)
+            except (HfHubHTTPError, ValueError, AttributeError):
+                full_prompt = f"{system_text}\n\n{user_text}"
                 text = self.client.text_generation(
                     prompt=full_prompt,
                     max_new_tokens=300,
                     temperature=0.7,
                     repetition_penalty=1.1,
-                    model=self.model,  # explicit even though client is bound
-                )
-            except (HfHubHTTPError, ValueError) as e:
-                # If the provider/model doesn't support text-generation, fall back to chat
-                msg = str(e)
-                unsupported = (
-                    "Task 'text-generation' not supported" in msg
-                    or "doesn't support task 'text-generation'" in msg
-                    or "Available tasks: ['conversational']" in msg
                 )
-                if unsupported:
-                    # OpenAI-style chat interface
-                    # Build messages: system + user
-                    messages = [
-                        {"role": "system", "content": system_text},
-                        {"role": "user", "content": user_text},
-                    ]
-                    chat = self.client.chat_completion(
-                        messages=messages,
-                        max_tokens=350,
-                        temperature=0.7,
-                        model=self.model,
-                    )
-                    # Robust extraction
-                    text = ""
-                    try:
-                        # chat.choices[0].message.content (OpenAI-like)
-                        choices = getattr(chat, "choices", None) or chat.get("choices", [])
-                        if choices:
-                            msg0 = choices[0].get("message") or {}
-                            text = msg0.get("content") or ""
-                        if not text:
-                            # Some providers return 'generated_text'
-                            text = getattr(chat, "generated_text", None) or chat.get("generated_text", "") or ""
-                    except Exception:
-                        text = str(chat)
-                else:
-                    raise
             except (httpx.ReadTimeout, httpx.ConnectTimeout):
                 return "The model request timed out. Please try again."