""" title: Prompt Enhancer author: Haervwe author_url: https://github.com/Haervwe funding_url: https://github.com/Haervwe/open-webui-tools version: 0.5.2 important note: if you are going to sue this filter with custom pipes, do not use the show enhanced prompt valve setting """ import logging from pydantic import BaseModel, Field from typing import Callable, Awaitable, Any, Optional import json from dataclasses import dataclass from fastapi import Request from open_webui.utils.chat import generate_chat_completion from open_webui.utils.misc import get_last_user_message from open_webui.models.models import Models from open_webui.models.users import User from open_webui.routers.models import get_models name = "enhancer" def setup_logger(): logger = logging.getLogger(name) if not logger.handlers: logger.setLevel(logging.DEBUG) handler = logging.StreamHandler() handler.set_name(name) formatter = logging.Formatter( "%(asctime)s - %(name)s - %(levelname)s - %(message)s" ) handler.setFormatter(formatter) logger.addHandler(handler) logger.propagate = False return logger logger = setup_logger() def setup_logger(): logger = logging.getLogger(name) if not logger.handlers: logger.setLevel(logging.DEBUG) handler = logging.StreamHandler() handler.set_name(name) formatter = logging.Formatter( "%(asctime)s - %(name)s - %(levelname)s - %(message)s" ) handler.setFormatter(formatter) logger.addHandler(handler) logger.propagate = False return logger logger = setup_logger() class Filter: class Valves(BaseModel): user_customizable_template: str = Field( default="""\ You are an expert prompt engineer. Your task is to enhance the given prompt by making it more detailed, specific, and effective. Consider the context and the user's intent. Response Format: - Provide only the enhanced prompt. - No additional text, markdown, or titles. - The enhanced prompt should start immediately without any introductory phrases. Example: Given Prompt: Write a poem about flowers. Enhanced Prompt: Craft a vivid and imaginative poem that explores the beauty and diversity of flowers, using rich imagery and metaphors to bring each bloom to life. Now, enhance the following prompt: """, description="Prompt to use in the Prompt enhancer System Message", ) show_status: bool = Field( default=False, description="Show status indicators", ) show_enhanced_prompt: bool = Field( default=False, description="Show Enahcend Prompt in chat", ) model_id: Optional[str] = Field( default=None, description="Model to use for the prompt enhancement, leave empty to use the same as selected for the main response.", ) def __init__(self): self.valves = self.Valves() self.__current_event_emitter__ = None self.__user__ = None self.__model__ = None self.__request__ = None async def inlet( self, body: dict, __event_emitter__: Callable[[Any], Awaitable[None]], __user__: Optional[dict] = None, __model__: Optional[dict] = None, __request__: Optional[Request] = None, ) -> dict: self.__current_event_emitter__ = __event_emitter__ self.__request__ = __request__ self.__model__ = __model__ self.__user__ = User(**__user__) if isinstance(__user__, dict) else __user__ # Fetch available models and log their relevant details available_models = await get_models(self.__request__, self.__user__) logger.debug("Available Models (truncated image data):") for model in available_models: truncated_model = model.model_dump() # Convert to dict for modification if "meta" in truncated_model: if isinstance(truncated_model["meta"], dict): if "profile_image_url" in truncated_model["meta"]: truncated_model["meta"]["profile_image_url"] = ( truncated_model["meta"]["profile_image_url"][:50] + "..." if isinstance( truncated_model["meta"]["profile_image_url"], str ) else None ) if "profile_image_url" in truncated_model["user"]: truncated_model["user"]["profile_image_url"] = ( truncated_model["user"]["profile_image_url"][:50] + "..." if isinstance( truncated_model["user"]["profile_image_url"], str ) else None ) else: logger.warning( f"Unexpected type for model.meta: {type(truncated_model['meta'])}" ) else: logger.warning("Model missing 'meta' key: %s", model) # Truncate files information if "knowledge" in truncated_model and isinstance( truncated_model["knowledge"], list ): for knowledge_item in truncated_model["knowledge"]: if isinstance(knowledge_item, dict) and "files" in knowledge_item: knowledge_item["files"] = "List of files (truncated)" logger.debug(json.dumps(truncated_model, indent=2)) messages = body["messages"] user_message = get_last_user_message(messages) if self.valves.show_status: await __event_emitter__( { "type": "status", "data": { "description": "Enhancing the prompt...", "done": False, }, } ) # Prepare context from chat history, excluding the last user message context_messages = [ msg for msg in messages if msg["role"] != "user" or msg["content"] != user_message ] context = "\n".join( [f"{msg['role'].upper()}: {msg['content']}" for msg in context_messages] ) # Build context block context_str = f'\n\nContext:\n"""{context}"""\n\n' if context else "" # Construct the system prompt with clear delimiters system_prompt = self.valves.user_customizable_template user_prompt = ( f"Context: {context_str}" f'Prompt to enhance:\n"""{user_message}"""\n\n' ) # Log the system prompt before sending to LLM logger.debug("System Prompt: %s", system_prompt) # Fixed string formatting # Determine the model to use # model_to_use = self.valves.model_id if self.valves.model_id else (body["model"]) model_to_use = None if self.valves.model_id: model_to_use = self.valves.model_id else: model_to_use = body["model"] # Check if the selected model has "-pipe" or "pipe" in its name. is_pipeline_model = False if "-pipe" in model_to_use.lower() or "pipe" in model_to_use.lower(): is_pipeline_model = True logger.warning( f"Selected model '{model_to_use}' appears to be a pipeline model. Consider using the base model." ) # If a pipeline model is *explicitly* chosen, use it. Otherwise, fall back to the main model. if not self.valves.model_id and is_pipeline_model: logger.warning( f"Pipeline model '{model_to_use}' selected without explicit model_id. Using main model instead." ) model_to_use = body["model"] # Fallback to main model is_pipeline_model = False # Construct payload for LLM request payload = { "model": model_to_use, "messages": [ {"role": "system", "content": system_prompt}, { "role": "user", "content": f"Enhance the given user prompt based on context: {user_prompt}", }, ], "stream": False, } try: # Use the User object directly, as done in other scripts logger.debug( "API CALL:\n Request: %s\n Form_data: %s\n User: %s", str(self.__request__), json.dumps(payload), self.__user__, ) response = await generate_chat_completion( self.__request__, payload, user=self.__user__, bypass_filter=True ) enhanced_prompt = response["choices"][0]["message"]["content"] logger.debug("Enhanced prompt: %s", enhanced_prompt) # Update the messages with the enhanced prompt messages[-1]["content"] = enhanced_prompt body["messages"] = messages if self.valves.show_status: await __event_emitter__( { "type": "status", "data": { "description": "Prompt successfully enhanced.", "done": True, }, } ) if self.valves.show_enhanced_prompt: enhanced_prompt_message = f"<details>\n<summary>Enhanced Prompt</summary>\n{enhanced_prompt}\n\n---\n\n</details>" await __event_emitter__( { "type": "message", "data": { "content": enhanced_prompt_message, }, } ) except ValueError as ve: logger.error("Value Error: %s", str(ve)) if self.valves.show_status: await __event_emitter__( { "type": "status", "data": { "description": f"Error: {str(ve)}", "done": True, }, } ) except Exception as e: logger.error("Unexpected error: %s", str(e)) if self.valves.show_status: await __event_emitter__( { "type": "status", "data": { "description": "An unexpected error occurred.", "done": True, }, } ) return body async def outlet( self, body: dict, __event_emitter__: Callable[[Any], Awaitable[None]], __user__: Optional[dict] = None, __model__: Optional[dict] = None, __request__: Optional[Request] = None, ) -> dict: self.__current_event_emitter__ = __event_emitter__ self.__request__ = __request__ self.__model__ = __model__ self.__user__ = User(**__user__) if isinstance(__user__, dict) else __user__ print(body) return body