ws-bom-robot-app 0.0.77__tar.gz → 0.0.79__tar.gz

{ws_bom_robot_app-0.0.77/ws_bom_robot_app.egg-info → ws_bom_robot_app-0.0.79}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: ws_bom_robot_app
-Version: 0.0.77
+Version: 0.0.79
 Summary: A FastAPI application serving ws bom/robot/llm platform ai.
 Home-page: https://github.com/websolutespa/bom
 Author: Websolute Spa
@@ -120,15 +120,51 @@ robot_cms_files_folder=llmKbFile
   #gunicorn -w 4 -k uvicorn.workers.UvicornWorker main:app --bind
   ```
 
-### 🔖 Windows requirements  
+## 📖 API documentation
 
-  #### libmagic (mandatory)
+- [swagger](http://localhost:6001/docs)
+- [redoc](http://localhost:6001/redoc)
+
+### 💬 multimodal chat
+
+The multimodal message allows users to interact with the application using both text and media files. 
+`robot` accept multimodal input in a uniform way, regarding the llm provider used. Can also be used the llm/model specific input format.
+
+- simple message
+
+```json
+{
+  "role": "user",
+  "content": "What is the capital of France?"
+}
+```
+
+- multimodal message
+
+```json
+{
+  "role": "user",
+  "content": [
+    {"type": "text", "text": "Read carefully all the attachments, analize the content and provide a summary for each one:"},        
+    {"type": "image", "url": "https://www.example.com/image/foo.jpg"},        
+    {"type": "file", "url": "https://www.example.com/pdf/bar.pdf"},   
+    {"type": "file", "url": "data:plain/text;base64,CiAgICAgIF9fX19fCiAgICAgLyAgIC..."}, # base64 encoded file     
+    {"type": "media", "mime_type": "plain/text", "data": "CiAgICAgIF9fX19fCiAgICAgLyAgIC..."} # google/gemini specific input format    
+  ]          
+}
+```
+
+---
+
+## 🔖 Windows requirements  
+
+### libmagic (mandatory)
 
   ```bash
   py -m pip install --upgrade python-magic-bin
   ```
   
-  #### tesseract-ocr (mandatory)
+### tesseract-ocr (mandatory)
 
   [Install tesseract](https://github.com/UB-Mannheim/tesseract/wiki)
   [Last win-64 release](https://github.com/tesseract-ocr/tesseract/releases/download/5.5.0/tesseract-ocr-w64-setup-5.5.0.20241111.exe)
@@ -143,15 +179,15 @@ robot_cms_files_folder=llmKbFile
   }
   ```
 
-  #### docling
-  
+### docling
+
   Set the following environment variables
 
   ```pwsh
   KMP_DUPLICATE_LIB_OK=TRUE
   ```    
 
-  #### libreoffice (optional: for robot_env set to development/production)
+### libreoffice (optional: for robot_env set to development/production)
 
   [Install libreoffice](https://www.libreoffice.org/download/download-libreoffice/)
   [Last win-64 release](https://download.documentfoundation.org/libreoffice/stable/24.8.2/win/x86_64/LibreOffice_24.8.2_Win_x86-64.msi)
@@ -166,7 +202,7 @@ robot_cms_files_folder=llmKbFile
   }
   ```
 
-  #### poppler (optional: for robot_env set to development/production)
+### poppler (optional: for robot_env set to development/production)
 
   [Download win poppler release](https://github.com/oschwartz10612/poppler-windows/releases)
   Extract the zip, copy the nested folder "poppler-x.x.x." to a program folder (e.g. C:\Program Files\poppler-24.08.0)
@@ -210,7 +246,7 @@ py -m build && twine check dist/*
 Install the package in editable project location
 
 ```pwsh
-py -m pip install --upgrade -e .
+py -m pip install -U -e .
 py -m pip show ws-bom-robot-app
 ```
 
@@ -278,12 +314,12 @@ docker run --rm --name ws-bom-robot-app-src -d -v "$(pwd)/ws_bom_robot_app:/app/
 
   ```pwsh
   twine upload --verbose -r testpypi dist/*
-  #py -m pip install -i https://test.pypi.org/simple/ --upgrade ws-bom-robot-app 
+  #pip install -i https://test.pypi.org/simple/ -U ws-bom-robot-app 
   ```
 
 - [pypi](https://pypi.org/project/ws-bom-robot-app/)
 
   ```pwsh
   twine upload --verbose dist/* 
-  #py -m pip install --upgrade ws-bom-robot-app
+
   ```

{ws_bom_robot_app-0.0.77 → ws_bom_robot_app-0.0.79}/README.md

@@ -54,15 +54,51 @@ robot_cms_files_folder=llmKbFile
   #gunicorn -w 4 -k uvicorn.workers.UvicornWorker main:app --bind
   ```
 
-### 🔖 Windows requirements  
+## 📖 API documentation
 
-  #### libmagic (mandatory)
+- [swagger](http://localhost:6001/docs)
+- [redoc](http://localhost:6001/redoc)
+
+### 💬 multimodal chat
+
+The multimodal message allows users to interact with the application using both text and media files. 
+`robot` accept multimodal input in a uniform way, regarding the llm provider used. Can also be used the llm/model specific input format.
+
+- simple message
+
+```json
+{
+  "role": "user",
+  "content": "What is the capital of France?"
+}
+```
+
+- multimodal message
+
+```json
+{
+  "role": "user",
+  "content": [
+    {"type": "text", "text": "Read carefully all the attachments, analize the content and provide a summary for each one:"},        
+    {"type": "image", "url": "https://www.example.com/image/foo.jpg"},        
+    {"type": "file", "url": "https://www.example.com/pdf/bar.pdf"},   
+    {"type": "file", "url": "data:plain/text;base64,CiAgICAgIF9fX19fCiAgICAgLyAgIC..."}, # base64 encoded file     
+    {"type": "media", "mime_type": "plain/text", "data": "CiAgICAgIF9fX19fCiAgICAgLyAgIC..."} # google/gemini specific input format    
+  ]          
+}
+```
+
+---
+
+## 🔖 Windows requirements  
+
+### libmagic (mandatory)
 
   ```bash
   py -m pip install --upgrade python-magic-bin
   ```
   
-  #### tesseract-ocr (mandatory)
+### tesseract-ocr (mandatory)
 
   [Install tesseract](https://github.com/UB-Mannheim/tesseract/wiki)
   [Last win-64 release](https://github.com/tesseract-ocr/tesseract/releases/download/5.5.0/tesseract-ocr-w64-setup-5.5.0.20241111.exe)
@@ -77,15 +113,15 @@ robot_cms_files_folder=llmKbFile
   }
   ```
 
-  #### docling
-  
+### docling
+
   Set the following environment variables
 
   ```pwsh
   KMP_DUPLICATE_LIB_OK=TRUE
   ```    
 
-  #### libreoffice (optional: for robot_env set to development/production)
+### libreoffice (optional: for robot_env set to development/production)
 
   [Install libreoffice](https://www.libreoffice.org/download/download-libreoffice/)
   [Last win-64 release](https://download.documentfoundation.org/libreoffice/stable/24.8.2/win/x86_64/LibreOffice_24.8.2_Win_x86-64.msi)
@@ -100,7 +136,7 @@ robot_cms_files_folder=llmKbFile
   }
   ```
 
-  #### poppler (optional: for robot_env set to development/production)
+### poppler (optional: for robot_env set to development/production)
 
   [Download win poppler release](https://github.com/oschwartz10612/poppler-windows/releases)
   Extract the zip, copy the nested folder "poppler-x.x.x." to a program folder (e.g. C:\Program Files\poppler-24.08.0)
@@ -144,7 +180,7 @@ py -m build && twine check dist/*
 Install the package in editable project location
 
 ```pwsh
-py -m pip install --upgrade -e .
+py -m pip install -U -e .
 py -m pip show ws-bom-robot-app
 ```
 
@@ -212,12 +248,12 @@ docker run --rm --name ws-bom-robot-app-src -d -v "$(pwd)/ws_bom_robot_app:/app/
 
   ```pwsh
   twine upload --verbose -r testpypi dist/*
-  #py -m pip install -i https://test.pypi.org/simple/ --upgrade ws-bom-robot-app 
+  #pip install -i https://test.pypi.org/simple/ -U ws-bom-robot-app 
   ```
 
 - [pypi](https://pypi.org/project/ws-bom-robot-app/)
 
   ```pwsh
   twine upload --verbose dist/* 
-  #py -m pip install --upgrade ws-bom-robot-app
+
   ```

{ws_bom_robot_app-0.0.77 → ws_bom_robot_app-0.0.79}/setup.py

@@ -4,7 +4,7 @@ _requirements = [line.split('#')[0].strip() for line in open("requirements.txt")
 
 setup(
     name="ws_bom_robot_app",
-    version="0.0.77",
+    version="0.0.79",
     description="A FastAPI application serving ws bom/robot/llm platform ai.",
     long_description=open("README.md", encoding='utf-8').read(),
     long_description_content_type="text/markdown",

{ws_bom_robot_app-0.0.77 → ws_bom_robot_app-0.0.79}/ws_bom_robot_app/config.py

@@ -14,6 +14,8 @@ class Settings(BaseSettings):
     robot_data_db_folder_out: str = 'out'
     robot_data_db_folder_store: str = 'store'
     robot_data_db_retention_days: float = 60
+    robot_data_attachment_folder: str = 'attachment'
+    robot_data_attachment_retention_days: float = 1
     robot_loader_max_threads: int = 1
     robot_task_max_total_parallelism: int = 2 * (os.cpu_count() or 1)
     robot_task_retention_days: float = 1

{ws_bom_robot_app-0.0.77 → ws_bom_robot_app-0.0.79}/ws_bom_robot_app/cron_manager.py

@@ -8,7 +8,7 @@ from apscheduler.triggers.date import DateTrigger
 from fastapi import APIRouter
 from datetime import datetime
 from ws_bom_robot_app.task_manager import task_manager
-from ws_bom_robot_app.llm.utils.kb import kb_cleanup_data_file
+from ws_bom_robot_app.llm.utils.cleanup import kb_cleanup_data_file, chat_cleanup_attachment
 from ws_bom_robot_app.util import _log
 from ws_bom_robot_app.config import config
 
@@ -57,7 +57,8 @@ class Job:
 class CronManager:
     _list_default = [
             Job('cleanup-task',task_manager.cleanup_task, interval=5 * 60),
-            Job('cleanup-data',kb_cleanup_data_file, interval=180 * 60),
+            Job('cleanup-kb-data',kb_cleanup_data_file, interval=180 * 60),
+            Job('cleanup-chat-attachment',chat_cleanup_attachment, interval=120 * 60),
         ]
     def __get_jobstore_strategy(self) -> JobstoreStrategy:
         if True or config.runtime_options().is_multi_process:
@@ -139,22 +140,22 @@ class CronManager:
 
     def execute_recurring_jobs(self):
         for job in self.scheduler.get_jobs():
-            if job.interval:
-                job.job_func()
+            if job.trigger.interval:
+                job.func()
 
     def pause_recurring_jobs(self):
         for job in self.scheduler.get_jobs():
-            if job.interval:
+            if job.trigger.interval:
                 self.pause_job(job.id)
 
     def resume_recurring_jobs(self):
         for job in self.scheduler.get_jobs():
-            if job.interval:
+            if job.trigger.interval:
                 self.resume_job(job.id)
 
     def remove_recurring_jobs(self):
         for job in self.scheduler.get_jobs():
-            if job.interval:
+            if job.trigger.interval:
                 self.remove_job(job.id)
 
     def clear(self):

{ws_bom_robot_app-0.0.77 → ws_bom_robot_app-0.0.79}/ws_bom_robot_app/llm/main.py

@@ -39,6 +39,7 @@ def _parse_formatted_message(message: str) -> str:
   except:
     result = message
   return result
+
 async def __stream(rq: StreamRequest, ctx: Request, queue: Queue, formatted: bool = True) -> None:
   #os.environ["KMP_DUPLICATE_LIB_OK"] = "TRUE"
 
@@ -47,11 +48,21 @@ async def __stream(rq: StreamRequest, ctx: Request, queue: Queue, formatted: boo
   for tool in rq.app_tools:
     tool.thread_id = rq.thread_id
 
+  #llm
+  __llm: LlmInterface = rq.get_llm()
+
   #chat history
   chat_history: list[BaseMessage] = []
   for message in rq.messages:
     if message.role in ["human","user"]:
-      chat_history.append(HumanMessage(content=message.content))
+      _content = message.content
+      # multimodal content parsing
+      if isinstance(_content, list):
+        try:
+          _content = await __llm.format_multimodal_content(_content)
+        except Exception as e:
+          logging.warning(f"Error parsing multimodal content {_content[:100]}: {e}")
+      chat_history.append(HumanMessage(content=_content))
     elif message.role in ["ai","assistant"]:
       message_content = ""
       if formatted:
@@ -78,8 +89,6 @@ async def __stream(rq: StreamRequest, ctx: Request, queue: Queue, formatted: boo
       if message_content:
         chat_history.append(AIMessage(content=message_content))
 
-  #llm
-  __llm: LlmInterface = rq.get_llm()
 
   #agent handler
   if formatted:

{ws_bom_robot_app-0.0.77 → ws_bom_robot_app-0.0.79}/ws_bom_robot_app/llm/nebuly_handler.py

@@ -6,7 +6,7 @@ from langchain_core.messages import BaseMessage, HumanMessage, AIMessage
 from langchain_core.outputs import ChatGenerationChunk, GenerationChunk
 
 class NebulyHandler(AsyncCallbackHandler):
-    def __init__(self, llm_model: str | None, threadId: str = None, chat_history: list = [],  url: str = None, api_key: str = None, user_id: str | None = None):
+    def __init__(self, llm_model: str | None, threadId: str = None, chat_history: list[BaseMessage] = [],  url: str = None, api_key: str = None, user_id: str | None = None):
         super().__init__()
         self.__started: bool = False
         self.__url: str = url
@@ -76,7 +76,7 @@ class NebulyHandler(AsyncCallbackHandler):
         self.interaction.output = finish.return_values["output"]
         # Trace
         self.llm_trace.output = finish.return_values["output"]
-        message_history =  self._convert_to_json_format(self.chat_history)[:-1]
+        message_history =  self._convert_to_json_format(self.chat_history)
         self.llm_trace.messages = self.__parse_multimodal_history(message_history)
         await self.__send_interaction()
 
@@ -145,16 +145,20 @@ class NebulyHandler(AsyncCallbackHandler):
         return payload
 
     def __parse_multimodal_input(self, input: list[dict]) -> str:
-        # Parse the multimodal input and return a string representation
-        # This is a placeholder implementation, you can customize it as needed
-        parsed_input = ""
-        for item in input:
-            if item.get("type") == "text":
-                parsed_input += item.get("text", "")
-            elif item.get("type") == "image_url":
-                parsed_input += " <image>"
-        # print(parsed_input)
-        return parsed_input
+      """Parse multimodal input and return a string representation."""
+      type_mapping = {
+        "text": lambda item: item.get("text", ""),
+        "image": lambda _: " <image>",
+        "image_url": lambda _: " <image>",
+        "file": lambda _: " <file>",
+        "media": lambda _: " <file>",
+        "document": lambda _: " <file>",
+      }
+
+      return "".join(
+        type_mapping.get(item.get("type", ""), lambda item: f" <{item.get('type', '')}>")
+        (item) for item in input
+      )
 
     def __parse_multimodal_history(self, messages: list[dict]) -> list[dict]:
         # Parse the multimodal history and return a list of dictionaries

{ws_bom_robot_app-0.0.77 → ws_bom_robot_app-0.0.79}/ws_bom_robot_app/llm/providers/llm_manager.py

@@ -3,6 +3,7 @@ from langchain_core.embeddings import Embeddings
 from langchain_core.language_models import BaseChatModel
 from pydantic import BaseModel, ConfigDict, Field
 import os
+from ws_bom_robot_app.llm.utils.download import Base64File
 
 class LlmConfig(BaseModel):
     api_url: Optional[str] = None
@@ -35,6 +36,30 @@ class LlmInterface:
     def get_parser(self):
         from langchain.agents.output_parsers.openai_tools import OpenAIToolsAgentOutputParser
         return OpenAIToolsAgentOutputParser()
+    async def _format_multimodal_image_message(self, message: dict) -> dict:
+        return {
+            "type": "image_url",
+            "image_url": {
+                "url": message.get("url")
+            }
+        }
+    async def _format_multimodal_file_message(self, message: dict, file: Base64File = None) -> dict:
+        _file = file or await Base64File.from_url(message.get("url"))
+        return {"type": "text", "text": f"Here's a file attachment named `{_file.name}` of type `{_file.mime_type}` in base64: `{_file.base64_content}`"}
+    async def format_multimodal_content(self, content: list) -> list:
+        _content = []
+        for message in content:
+            if isinstance(message, dict):
+                if message.get("type") == "image" and "url" in message:
+                    _content.append(await self._format_multimodal_image_message(message))
+                elif message.get("type") == "file" and "url" in message:
+                    _content.append(await self._format_multimodal_file_message(message))
+                else:
+                    # pass through text or other formats unchanged
+                    _content.append(message)
+            else:
+                _content.append(message)
+        return _content
 
 class Anthropic(LlmInterface):
     def get_llm(self):
@@ -62,6 +87,16 @@ class Anthropic(LlmInterface):
         response = client.models.list()
         return response.data
 
+    async def _format_multimodal_image_message(self, message: dict) -> dict:
+        file = await Base64File.from_url(message.get("url"))
+        return { "type": "image_url", "image_url": { "url": file.base64_url }}
+    async def _format_multimodal_file_message(self, message: dict, file: Base64File = None) -> dict:
+        _file = file or await Base64File.from_url(message.get("url"))
+        if _file.extension in ["pdf"]:
+            return {"type": "document", "source": {"type": "base64", "media_type": _file.mime_type, "data": _file.base64_content}}
+        else:
+          return await super()._format_multimodal_file_message(message, _file)
+
 class OpenAI(LlmInterface):
     def __init__(self, config: LlmConfig):
         super().__init__(config)
@@ -84,6 +119,13 @@ class OpenAI(LlmInterface):
         response = openai.models.list()
         return response.data
 
+    async def _format_multimodal_file_message(self, message: dict, file: Base64File = None) -> dict:
+        _file = file or await Base64File.from_url(message.get("url"))
+        if _file.extension in ["pdf"]:
+            return {"type": "file", "file": { "source_type": "base64", "file_data": _file.base64_url, "mime_type": _file.mime_type, "filename": _file.name}}
+        else:
+          return await super()._format_multimodal_file_message(message, _file)
+
 class DeepSeek(LlmInterface):
     def get_llm(self):
         from langchain_openai import ChatOpenAI
@@ -104,33 +146,48 @@ class DeepSeek(LlmInterface):
         response = openai.models.list()
         return response.data
 
+    async def _format_multimodal_image_message(self, message: dict) -> dict:
+        print(f"{DeepSeek.__name__} does not support image messages")
+        return None
+
+    async def _format_multimodal_file_message(self, message: dict, file: Base64File = None) -> dict:
+        print(f"{DeepSeek.__name__} does not support file messages")
+        return None
+
 class Google(LlmInterface):
-  def get_llm(self):
-    from langchain_google_genai.chat_models import ChatGoogleGenerativeAI
-    return ChatGoogleGenerativeAI(
-      model=self.config.model,
-      google_api_key=self.config.api_key or os.getenv("GOOGLE_API_KEY"),
-      temperature=self.config.temperature,
-      disable_streaming=False,
-    )
-
-  def get_embeddings(self):
-    from langchain_google_genai.embeddings import GoogleGenerativeAIEmbeddings
-    return GoogleGenerativeAIEmbeddings(
-      google_api_key=self.config.api_key or os.getenv("GOOGLE_API_KEY"),
-      model="models/text-embedding-004")
-
-  def get_models(self):
-    import google.generativeai as genai
-    genai.configure(api_key=self.config.api_key or os.getenv("GOOGLE_API_KEY"))
-    response = genai.list_models()
-    return [{
-      "id": model.name,
-      "name": model.display_name,
-      "description": model.description,
-      "input_token_limit": model.input_token_limit,
-      "output_token_limit": model.output_token_limit
-    } for model in response if "gemini" in model.name.lower()]
+    def get_llm(self):
+      from langchain_google_genai.chat_models import ChatGoogleGenerativeAI
+      return ChatGoogleGenerativeAI(
+        model=self.config.model,
+        google_api_key=self.config.api_key or os.getenv("GOOGLE_API_KEY"),
+        temperature=self.config.temperature,
+        disable_streaming=False,
+      )
+
+    def get_embeddings(self):
+      from langchain_google_genai.embeddings import GoogleGenerativeAIEmbeddings
+      return GoogleGenerativeAIEmbeddings(
+        google_api_key=self.config.api_key or os.getenv("GOOGLE_API_KEY"),
+        model="models/text-embedding-004")
+
+    def get_models(self):
+      import google.generativeai as genai
+      genai.configure(api_key=self.config.api_key or os.getenv("GOOGLE_API_KEY"))
+      response = genai.list_models()
+      return [{
+        "id": model.name,
+        "name": model.display_name,
+        "description": model.description,
+        "input_token_limit": model.input_token_limit,
+        "output_token_limit": model.output_token_limit
+      } for model in response if "gemini" in model.name.lower()]
+
+    async def _format_multimodal_file_message(self, message: dict, file: Base64File = None) -> dict:
+        _file = file or await Base64File.from_url(message.get("url"))
+        if _file.extension in ["pdf", "csv"]:
+            return {"type": "media", "mime_type": _file.mime_type, "data": _file.base64_content }
+        else:
+          return await super()._format_multimodal_file_message(message, _file)
 
 class Gvertex(LlmInterface):
     def get_llm(self):
@@ -168,6 +225,13 @@ class Gvertex(LlmInterface):
         finally:
           return _models
 
+    async def _format_multimodal_file_message(self, message: dict, file: Base64File = None) -> dict:
+        _file = file or await Base64File.from_url(message.get("url"))
+        if _file.extension in ["pdf", "csv"]:
+            return {"type": "media", "mime_type": _file.mime_type, "data": _file.base64_content }
+        else:
+          return await super()._format_multimodal_file_message(message, _file)
+
 class Groq(LlmInterface):
     def get_llm(self):
         from langchain_groq import ChatGroq
@@ -286,6 +350,10 @@ class Ollama(LlmInterface):
               "details": model['details']
             } for model in models]
 
+    async def _format_multimodal_image_message(self, message: dict) -> dict:
+        file = await Base64File.from_url(message.get("url"))
+        return { "type": "image_url", "image_url": { "url": file.base64_url }}
+
 class LlmManager:
 
     #class variables (static)

ws_bom_robot_app-0.0.79/ws_bom_robot_app/llm/utils/cleanup.py

@@ -0,0 +1,74 @@
+import os, logging
+from ws_bom_robot_app.config import config
+from datetime import datetime, timedelta
+
+def _cleanup_data_file(folders: list[str], retention: float) -> dict:
+    """
+    clean up old data files in the specified folder
+
+    Returns:
+    - Dictionary with cleanup statistics
+    """
+    _deleted_files = []
+    _deleted_dirs = []
+    _freed_space = 0
+
+    for folder in folders:
+        if not os.path.exists(folder):
+            logging.warning(f"Folder does not exist: {folder}")
+            continue
+
+        # delete old files
+        for root, dirs, files in os.walk(folder, topdown=False):
+            for file in files:
+                file_path = os.path.join(root, file)
+                try:
+                    file_stat = os.stat(file_path)
+                    file_creation_time = datetime.fromtimestamp(file_stat.st_mtime)
+                    if file_creation_time < datetime.now() - timedelta(days=retention):
+                        _freed_space += file_stat.st_size
+                        os.remove(file_path)
+                        _deleted_files.append(file_path)
+                except (OSError, IOError) as e:
+                    logging.error(f"Error deleting file {file_path}: {e}")
+
+        # clean up empty directories (bottom-up)
+        for root, dirs, files in os.walk(folder, topdown=False):
+            # skip the root folder itself
+            if root == folder:
+                continue
+            try:
+                # check if directory is empty
+                if not os.listdir(root):
+                    os.rmdir(root)
+                    _deleted_dirs.append(root)
+            except OSError as e:
+                logging.debug(f"Could not remove directory {root}: {e}")
+    logging.info(f"Deleted {len(_deleted_files)} files; Freed space: {_freed_space / (1024 * 1024):.2f} MB")
+
+    return {
+        "deleted_files_count": len(_deleted_files),
+        "deleted_dirs_count": len(_deleted_dirs),
+        "freed_space_mb": _freed_space / (1024 * 1024)
+    }
+
+def kb_cleanup_data_file() -> dict:
+    """
+    clean up vector db data files
+    """
+
+    folders = [
+        os.path.join(config.robot_data_folder, config.robot_data_db_folder, config.robot_data_db_folder_out),
+        os.path.join(config.robot_data_folder, config.robot_data_db_folder, config.robot_data_db_folder_store),
+        os.path.join(config.robot_data_folder, config.robot_data_db_folder, config.robot_data_db_folder_src)
+    ]
+    return _cleanup_data_file(folders, config.robot_data_db_retention_days)
+
+def chat_cleanup_attachment() -> dict:
+    """
+    clean up chat attachment files
+    """
+    folders = [
+        os.path.join(config.robot_data_folder, config.robot_data_attachment_folder)
+        ]
+    return _cleanup_data_file(folders, config.robot_data_attachment_retention_days)