sunholo 0.140.7__py3-none-any.whl → 0.140.9__py3-none-any.whl

sunholo/agents/flask/vac_routes.py

@@ -36,7 +36,7 @@ except ImportError:
 # Cache dictionary to store validated API keys
 api_key_cache = {}
 cache_duration = timedelta(minutes=5)  # Cache duration
-# Global caches and thread pool (add these after your existing globals)
+# Global caches and thread pool 
 _config_cache = {}
 _config_lock = threading.Lock()
 _thread_pool = ThreadPoolExecutor(max_workers=4)
@@ -94,17 +94,24 @@ if __name__ == "__main__":
             except Exception as e:
                 log.warning(f"Failed to pre-load config for {vector_name}: {e}")
 
-
-    @lru_cache(maxsize=100)
     def _get_cached_config(self, vector_name: str):
-        """Cached config loader with thread safety"""
+        """Cached config loader with thread safety - CORRECTED VERSION"""
+        # Check cache first (without lock for read)
+        if vector_name in _config_cache:
+            log.debug(f"Using cached config for {vector_name}")
+            return _config_cache[vector_name]
+        
+        # Need to load config
         with _config_lock:
+            # Double-check inside lock (another thread might have loaded it)
             if vector_name in _config_cache:
                 return _config_cache[vector_name]
             
             try:
+                log.info(f"Loading fresh config for {vector_name}")
                 config = ConfigManager(vector_name)
                 _config_cache[vector_name] = config
+                log.info(f"Cached config for {vector_name}")
                 return config
             except Exception as e:
                 log.error(f"Error loading config for {vector_name}: {e}")
@@ -286,13 +293,13 @@ if __name__ == "__main__":
         if is_async:
             log.info(f"Stream interpreter is async: {observed_stream_interpreter}")
 
-        # Fast prep
-        prep = self.prep_vac(request, vector_name)
-        
-        # Check for prep errors
-        if isinstance(prep, tuple) and len(prep) == 2:
-            error_response, status_code = prep
-            return jsonify(error_response), status_code
+        # Call prep_vac and handle errors properly
+        try:
+            prep = self.prep_vac(request, vector_name)
+        except Exception as e:
+            log.error(f"prep_vac failed: {e}")
+            error_response = {'error': f'Prep error: {str(e)}'}
+            return jsonify(error_response), 500
         
         log.info(f"Processing prep completed in {time.time() - request_start:.3f}s")
         
@@ -343,7 +350,9 @@ if __name__ == "__main__":
                                     else:
                                         result_queue.put(chunk)
                             except Exception as e:
-                                result_queue.put(f"Streaming Error: {str(e)} {traceback.format_exc()}")
+                                error_msg = f"Streaming Error: {str(e)} {traceback.format_exc()}"
+                                log.error(error_msg)
+                                result_queue.put(error_msg)
                             finally:
                                 result_queue.put(None)  # Sentinel
                         
@@ -385,16 +394,18 @@ if __name__ == "__main__":
                             yield chunk
 
             except Exception as e:
-                yield f"Streaming Error: {str(e)} {traceback.format_exc()}"
+                error_msg = f"Streaming Error: {str(e)} {traceback.format_exc()}"
+                log.error(error_msg)
+                yield error_msg
 
         # Create streaming response
         response = Response(generate_response_content(), content_type='text/plain; charset=utf-8')
         response.headers['Transfer-Encoding'] = 'chunked'
 
-        log.debug(f"Streaming response created in {time.time() - request_start:.3f}s")
+        log.info(f"Streaming response created in {time.time() - request_start:.3f}s")
         
+        # Do final trace operations in background (don't block the response)
         if trace:
-            # Do final trace operations in background
             _thread_pool.submit(self._finalize_trace_background, trace, span, response, all_input)
 
         return response
@@ -744,59 +755,42 @@ if __name__ == "__main__":
     def prep_vac(self, request, vector_name):
         start_time = time.time()
         
-        # Fast request parsing
-        try:
-            if request.content_type.startswith('application/json'):
-                data = request.get_json()
-            elif request.content_type.startswith('multipart/form-data'):
-                data = request.form.to_dict()
-                # Handle file upload in background if present
-                if 'file' in request.files:
-                    file = request.files['file']
-                    if file.filename != '':
-                        log.info(f"Found file: {file.filename} - uploading in background")
-                        # Start file upload in background, don't block
-                        upload_future = _thread_pool.submit(self._handle_file_upload_background, file, vector_name)
-                        data["_upload_future"] = upload_future
-            else:
-                return {"error": "Unsupported content type"}, 400
-        except Exception as e:
-            return {"error": f"Request parsing error: {str(e)}"}, 400
+        # Fast request parsing - KEEP ORIGINAL ERROR HANDLING STYLE
+        if request.content_type.startswith('application/json'):
+            data = request.get_json()
+        elif request.content_type.startswith('multipart/form-data'):
+            data = request.form.to_dict()
+            # Handle file upload in background if present
+            if 'file' in request.files:
+                file = request.files['file']
+                if file.filename != '':
+                    log.info(f"Found file: {file.filename} - uploading in background")
+                    # Start file upload in background, don't block
+                    upload_future = _thread_pool.submit(self._handle_file_upload_background, file, vector_name)
+                    data["_upload_future"] = upload_future
+        else:
+            # KEEP ORIGINAL STYLE - return the error response directly
+            raise ValueError("Unsupported content type")
 
         log.info(f"vac/{vector_name} got data keys: {list(data.keys())}")
 
-        # Extract essential data first
-        try:
-            user_input = data.pop('user_input').strip()
-            stream_wait_time = data.pop('stream_wait_time', 7)
-            stream_timeout = data.pop('stream_timeout', 120)
-            chat_history = data.pop('chat_history', None)
-            eval_percent = data.pop('eval_percent', 0.01)
-            vector_name_param = data.pop('vector_name', vector_name)
-            trace_id = data.pop('trace_id', None)
-            
-            if not user_input:
-                return {"error": "No user input provided"}, 400
-                
-        except Exception as e:
-            return {"error": f"Required field missing: {str(e)}"}, 400
-
-        # Get config from cache (should be very fast)
+        # Get config from cache first (before processing other data)
         try:
             vac_config = self._get_cached_config(vector_name)
         except Exception as e:
-            log.error(f"Config error: {e}")
-            return {"error": f"Unable to find vac_config for {vector_name} - {str(e)}"}, 500
+            raise ValueError(f"Unable to find vac_config for {vector_name} - {str(e)}")
 
-        # Process chat history with caching (should be fast)
-        paired_messages = extract_chat_history_with_cache(chat_history)
+        # Extract data (keep original logic)
+        user_input = data.pop('user_input').strip()
+        stream_wait_time = data.pop('stream_wait_time', 7)
+        stream_timeout = data.pop('stream_timeout', 120)
+        chat_history = data.pop('chat_history', None)
+        eval_percent = data.pop('eval_percent', 0.01)
+        vector_name_param = data.pop('vector_name', vector_name)
+        data.pop('trace_id', None)  # to ensure not in kwargs
 
-        # Start tracing in background (don't block)
-        trace = None
-        span = None
-        if self.add_langfuse_eval:
-            trace_future = _thread_pool.submit(self._create_langfuse_trace_background, request, vector_name, trace_id)
-            # We'll get the trace result later if needed
+        # Process chat history with caching
+        paired_messages = extract_chat_history_with_cache(chat_history)
 
         # Wait for file upload if it was started (with timeout)
         if "_upload_future" in data:
@@ -809,7 +803,7 @@ if __name__ == "__main__":
             finally:
                 data.pop("_upload_future", None)
 
-        # Build final input
+        # BUILD all_input BEFORE trace creation (this was moved inside try/catch by mistake)
         all_input = {
             'user_input': user_input, 
             'vector_name': vector_name_param, 
@@ -820,10 +814,17 @@ if __name__ == "__main__":
             'kwargs': data
         }
 
-        # Try to get trace result if available (don't block long)
+        # Initialize trace variables
+        trace = None
+        span = None
         if self.add_langfuse_eval:
+            trace_id = data.get('trace_id')
+            # Create trace in background - don't block
+            trace_future = _thread_pool.submit(self._create_langfuse_trace_background, request, vector_name, trace_id)
+
+            # Try to get trace result if available (don't block long)
             try:
-                trace = trace_future.result(timeout=0.5)  # 500ms max wait
+                trace = trace_future.result(timeout=0.1)  # Very short timeout
                 if trace:
                     this_vac_config = vac_config.configs_by_kind.get("vacConfig")
                     metadata_config = None
@@ -851,67 +852,6 @@ if __name__ == "__main__":
             "vac_config": vac_config
         }
 
-    async def prep_vac_async(self, request, vector_name):
-        """Async version of prep_vac."""
-        # Parse request data
-        if request.content_type.startswith('application/json'):
-            data = request.get_json()
-        elif request.content_type.startswith('multipart/form-data'):
-            data = request.form.to_dict()
-            if 'file' in request.files:
-                file = request.files['file']
-                if file.filename != '':
-                    log.info(f"Found file: {file.filename} to upload to GCS")
-                    try:
-                        # Make file upload async if possible
-                        image_uri, mime_type = await self.handle_file_upload_async(file, vector_name)
-                        data["image_uri"] = image_uri
-                        data["mime"] = mime_type
-                    except Exception as e:
-                        log.error(traceback.format_exc())
-                        return jsonify({'error': str(e), 'traceback': traceback.format_exc()}), 500
-                else:
-                    log.error("No file selected")
-                    return jsonify({"error": "No file selected"}), 400
-        else:
-            return jsonify({"error": "Unsupported content type"}), 400
-
-        log.info(f"vac/{vector_name} got data: {data}")
-
-        # Run these operations concurrently
-        tasks = []
-        
-        # Extract other data while configs load
-        user_input = data.pop('user_input').strip()
-        stream_wait_time = data.pop('stream_wait_time', 7)
-        stream_timeout = data.pop('stream_timeout', 120)
-        chat_history = data.pop('chat_history', None)
-        vector_name_param = data.pop('vector_name', vector_name)
-        data.pop('trace_id', None)  # to ensure not in kwargs
-        
-        # Task 3: Process chat history
-        chat_history_task = asyncio.create_task(extract_chat_history_async_cached(chat_history))
-        tasks.append(chat_history_task)
-        
-        # Await all tasks concurrently
-        results = await asyncio.gather(*tasks, return_exceptions=True)
-
-        paired_messages = results[0] if not isinstance(results[0], Exception) else []
-        
-        # Only create span after we have trace
-        all_input = {
-            'user_input': user_input, 
-            'vector_name': vector_name_param, 
-            'chat_history': paired_messages, 
-            'stream_wait_time': stream_wait_time,
-            'stream_timeout': stream_timeout,
-            'kwargs': data
-        }
-        
-        return {
-            "all_input": all_input
-        }
-
     def handle_file_upload(self, file, vector_name):
         try:
             file.save(file.filename)

{sunholo-0.140.7.dist-info → sunholo-0.140.9.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: sunholo
-Version: 0.140.7
+Version: 0.140.9
 Summary: AI DevOps - a package to help deploy GenAI to the Cloud.
 Author-email: Holosun ApS <multivac@sunholo.com>
 License: Apache License, Version 2.0

{sunholo-0.140.7.dist-info → sunholo-0.140.9.dist-info}/RECORD

@@ -14,7 +14,7 @@ sunholo/agents/fastapi/base.py,sha256=W-cyF8ZDUH40rc-c-Apw3-_8IIi2e4Y9qRtnoVnsc1
 sunholo/agents/fastapi/qna_routes.py,sha256=lKHkXPmwltu9EH3RMwmD153-J6pE7kWQ4BhBlV3to-s,3864
 sunholo/agents/flask/__init__.py,sha256=dEoByI3gDNUOjpX1uVKP7uPjhfFHJubbiaAv3xLopnk,63
 sunholo/agents/flask/base.py,sha256=vnpxFEOnCmt9humqj-jYPLfJcdwzsop9NorgkJ-tSaU,1756
-sunholo/agents/flask/vac_routes.py,sha256=obdnMF6x1m5YJ4GcrHDpmsOy6kjw7FyZZMcswkargQk,37354
+sunholo/agents/flask/vac_routes.py,sha256=eafqIudPKAtsOC73bnIXCpreL8AhMz_LQ212HuXqGhc,35101
 sunholo/archive/__init__.py,sha256=qNHWm5rGPVOlxZBZCpA1wTYPbalizRT7f8X4rs2t290,31
 sunholo/archive/archive.py,sha256=PxVfDtO2_2ZEEbnhXSCbXLdeoHoQVImo4y3Jr2XkCFY,1204
 sunholo/auth/__init__.py,sha256=TeP-OY0XGxYV_8AQcVGoh35bvyWhNUcMRfhuD5l44Sk,91
@@ -168,9 +168,9 @@ sunholo/vertex/init.py,sha256=1OQwcPBKZYBTDPdyU7IM4X4OmiXLdsNV30C-fee2scQ,2875
 sunholo/vertex/memory_tools.py,sha256=tBZxqVZ4InTmdBvLlOYwoSEWu4-kGquc-gxDwZCC4FA,7667
 sunholo/vertex/safety.py,sha256=S9PgQT1O_BQAkcqauWncRJaydiP8Q_Jzmu9gxYfy1VA,2482
 sunholo/vertex/type_dict_to_json.py,sha256=uTzL4o9tJRao4u-gJOFcACgWGkBOtqACmb6ihvCErL8,4694
-sunholo-0.140.7.dist-info/licenses/LICENSE.txt,sha256=SdE3QjnD3GEmqqg9EX3TM9f7WmtOzqS1KJve8rhbYmU,11345
-sunholo-0.140.7.dist-info/METADATA,sha256=OdauM1Q7Yq7QdH01pOWb5ULjgVebsZMOuIpglmSZaoQ,10067
-sunholo-0.140.7.dist-info/WHEEL,sha256=Nw36Djuh_5VDukK0H78QzOX-_FQEo6V37m3nkm96gtU,91
-sunholo-0.140.7.dist-info/entry_points.txt,sha256=bZuN5AIHingMPt4Ro1b_T-FnQvZ3teBes-3OyO0asl4,49
-sunholo-0.140.7.dist-info/top_level.txt,sha256=wt5tadn5--5JrZsjJz2LceoUvcrIvxjHJe-RxuudxAk,8
-sunholo-0.140.7.dist-info/RECORD,,
+sunholo-0.140.9.dist-info/licenses/LICENSE.txt,sha256=SdE3QjnD3GEmqqg9EX3TM9f7WmtOzqS1KJve8rhbYmU,11345
+sunholo-0.140.9.dist-info/METADATA,sha256=PrTJywV40Kp5vGdiwPaLEzu2pujsY0DuoJSKuhis0MA,10067
+sunholo-0.140.9.dist-info/WHEEL,sha256=Nw36Djuh_5VDukK0H78QzOX-_FQEo6V37m3nkm96gtU,91
+sunholo-0.140.9.dist-info/entry_points.txt,sha256=bZuN5AIHingMPt4Ro1b_T-FnQvZ3teBes-3OyO0asl4,49
+sunholo-0.140.9.dist-info/top_level.txt,sha256=wt5tadn5--5JrZsjJz2LceoUvcrIvxjHJe-RxuudxAk,8
+sunholo-0.140.9.dist-info/RECORD,,