ws-bom-robot-app 0.0.33__py3-none-any.whl → 0.0.35__py3-none-any.whl

ws_bom_robot_app/llm/utils/download.py

@@ -1,79 +1,79 @@
-from typing import List,Optional
-import os, logging, aiohttp, asyncio
-from tqdm.asyncio import tqdm
-
-async def download_files(urls: List[str], destination_folder: str, authorization: str = None):
-    tasks = [download_file(file, os.path.join(destination_folder, os.path.basename(file)), authorization=authorization) for file in urls]
-    results = await asyncio.gather(*tasks, return_exceptions=False)
-    for i, result in enumerate(results):
-        if not result:
-            raise Exception(f"Download failed for file: {urls[i]}")
-
-async def download_file(url: str, destination: str, chunk_size: int = 8192, authorization: str = None) -> Optional[str]:
-  """
-  Downloads a file from a given URL to a destination path asynchronously.
-
-  Args:
-      url: The URL of the file to download
-      destination: The local path where the file should be saved
-      chunk_size: Size of chunks to download (default: 8192 bytes)
-
-  Returns:
-      str: Path to the downloaded file if successful, None otherwise
-
-  Raises:
-      Various exceptions are caught and logged
-  """
-  try:
-      # Ensure the destination directory exists
-      os.makedirs(os.path.dirname(os.path.abspath(destination)), exist_ok=True)
-
-      async with aiohttp.ClientSession() as session:
-          if authorization:
-            headers = {'Authorization': authorization}
-            session.headers.update(headers)
-          async with session.get(url) as response:
-              # Check if the request was successful
-              if response.status != 200:
-                  logging.error(f"Failed to download file. Status code: {response.status}")
-                  return None
-
-              # Get the total file size if available
-              total_size = int(response.headers.get('content-length', 0))
-              # Open the destination file and write chunks
-              with open(destination, 'wb') as f:
-                  with tqdm(
-                      total=total_size,
-                      desc="Downloading",
-                      unit='B',
-                      unit_scale=True,
-                      unit_divisor=1024
-                  ) as pbar:
-                      async for chunk in response.content.iter_chunked(chunk_size):
-                          if chunk:
-                              f.write(chunk)
-                              pbar.update(len(chunk))
-
-              logging.info(f"File downloaded successfully to {destination}")
-              return destination
-
-  except aiohttp.ClientError as e:
-      logging.error(f"Network error occurred: {str(e)}")
-      return None
-  except asyncio.TimeoutError:
-      logging.error("Download timed out")
-      return None
-  except IOError as e:
-      logging.error(f"IO error occurred: {str(e)}")
-      return None
-  except Exception as e:
-      logging.error(f"Unexpected error occurred: {str(e)}")
-      return None
-  finally:
-      # If download failed and file was partially created, clean it up
-      if os.path.exists(destination) and os.path.getsize(destination) == 0:
-          try:
-              os.remove(destination)
-              logging.info(f"Cleaned up incomplete download: {destination}")
-          except OSError:
-              pass
+from typing import List,Optional
+import os, logging, aiohttp, asyncio
+from tqdm.asyncio import tqdm
+
+async def download_files(urls: List[str], destination_folder: str, authorization: str = None):
+    tasks = [download_file(file, os.path.join(destination_folder, os.path.basename(file)), authorization=authorization) for file in urls]
+    results = await asyncio.gather(*tasks, return_exceptions=False)
+    for i, result in enumerate(results):
+        if not result:
+            raise Exception(f"Download failed for file: {urls[i]}")
+
+async def download_file(url: str, destination: str, chunk_size: int = 8192, authorization: str = None) -> Optional[str]:
+  """
+  Downloads a file from a given URL to a destination path asynchronously.
+
+  Args:
+      url: The URL of the file to download
+      destination: The local path where the file should be saved
+      chunk_size: Size of chunks to download (default: 8192 bytes)
+
+  Returns:
+      str: Path to the downloaded file if successful, None otherwise
+
+  Raises:
+      Various exceptions are caught and logged
+  """
+  try:
+      # Ensure the destination directory exists
+      os.makedirs(os.path.dirname(os.path.abspath(destination)), exist_ok=True)
+
+      async with aiohttp.ClientSession() as session:
+          if authorization:
+            headers = {'Authorization': authorization}
+            session.headers.update(headers)
+          async with session.get(url) as response:
+              # Check if the request was successful
+              if response.status != 200:
+                  logging.error(f"Failed to download file. Status code: {response.status}")
+                  return None
+
+              # Get the total file size if available
+              total_size = int(response.headers.get('content-length', 0))
+              # Open the destination file and write chunks
+              with open(destination, 'wb') as f:
+                  with tqdm(
+                      total=total_size,
+                      desc="Downloading",
+                      unit='B',
+                      unit_scale=True,
+                      unit_divisor=1024
+                  ) as pbar:
+                      async for chunk in response.content.iter_chunked(chunk_size):
+                          if chunk:
+                              f.write(chunk)
+                              pbar.update(len(chunk))
+
+              logging.info(f"File downloaded successfully to {destination}")
+              return destination
+
+  except aiohttp.ClientError as e:
+      logging.error(f"Network error occurred: {str(e)}")
+      return None
+  except asyncio.TimeoutError:
+      logging.error("Download timed out")
+      return None
+  except IOError as e:
+      logging.error(f"IO error occurred: {str(e)}")
+      return None
+  except Exception as e:
+      logging.error(f"Unexpected error occurred: {str(e)}")
+      return None
+  finally:
+      # If download failed and file was partially created, clean it up
+      if os.path.exists(destination) and os.path.getsize(destination) == 0:
+          try:
+              os.remove(destination)
+              logging.info(f"Cleaned up incomplete download: {destination}")
+          except OSError:
+              pass

ws_bom_robot_app/llm/utils/print.py

@@ -1,29 +1,29 @@
-import os, sys, json
-
-class HiddenPrints:
-    def __enter__(self):
-        self._original_stdout = sys.stdout
-        self._original_stderr = sys.stderr
-
-        sys.stdout = open(os.devnull, 'w')
-        sys.stderr = open(os.devnull, 'w')
-
-    def __exit__(self, exc_type, exc_val, exc_tb):
-        sys.stdout.close()
-        sys.stderr.close()
-        sys.stdout = self._original_stdout
-        sys.stderr = self._original_stderr
-
-def printJson(data) -> str:
-    return f"{json.dumps(data, indent=2, sort_keys=True)},"
-
-def printSingleJson(data) -> str:
-    return f"{json.dumps(data, indent=2, sort_keys=True)}"
-
-def printString(data: str) -> str:
-    if data != "":
-        return printJson(data)
-
-def printSingleString(data: str) -> str:
-    if data != "":
-        return printSingleJson(data)
+import os, sys, json
+
+class HiddenPrints:
+    def __enter__(self):
+        self._original_stdout = sys.stdout
+        self._original_stderr = sys.stderr
+
+        sys.stdout = open(os.devnull, 'w')
+        sys.stderr = open(os.devnull, 'w')
+
+    def __exit__(self, exc_type, exc_val, exc_tb):
+        sys.stdout.close()
+        sys.stderr.close()
+        sys.stdout = self._original_stdout
+        sys.stderr = self._original_stderr
+
+def printJson(data) -> str:
+    return f"{json.dumps(data, indent=2, sort_keys=True)},"
+
+def printSingleJson(data) -> str:
+    return f"{json.dumps(data, indent=2, sort_keys=True)}"
+
+def printString(data: str) -> str:
+    if data != "":
+        return printJson(data)
+
+def printSingleString(data: str) -> str:
+    if data != "":
+        return printSingleJson(data)

ws_bom_robot_app/llm/utils/secrets.py

@@ -0,0 +1,26 @@
+import json, os, logging
+from typing import Union
+class Secrets:
+  @staticmethod
+  def from_file(path: str) -> dict[str, Union[str, int, list]] | None:
+    if os.path.exists(path):
+      with open(path, 'r') as file:
+        _content = file.read()
+        try:
+          return json.loads(_content)
+        except json.JSONDecodeError:
+          logging.error(f"Failed to parse secret file: {path}")
+    else:
+      logging.error(f"Secret file not found: {path}")
+    return None
+  @staticmethod
+  def from_env(key: str) -> dict[str, Union[str, int, list]] | None:
+    _content = os.getenv(key)
+    if _content:
+      try:
+        return json.loads(_content)
+      except json.JSONDecodeError:
+        logging.error(f"Failed to parse environment variable: {key}")
+    else:
+      logging.error(f"Environment variable not found: {key}")
+    return None

ws_bom_robot_app/llm/vector_store/generator.py

@@ -1,137 +1,137 @@
-import os, gc, shutil, logging, traceback
-import asyncio, aiofiles, aiofiles.os
-from fastapi import HTTPException
-from fastapi.responses import StreamingResponse
-from langchain_core.documents import Document
-from ws_bom_robot_app.llm.vector_store.loader.base import Loader
-from ws_bom_robot_app.llm.models.api import RulesRequest, KbRequest, VectorDbResponse
-from ws_bom_robot_app.llm.vector_store.integration.manager import IntegrationManager
-from ws_bom_robot_app.llm.utils.faiss_helper import FaissHelper
-from ws_bom_robot_app.config import config
-from ws_bom_robot_app.llm.models.kb import load_endpoints
-from ws_bom_robot_app.llm.utils.download import download_files
-
-async def _cleanup_directory(directory_path: str):
-  if os.path.exists(directory_path):
-    await asyncio.to_thread(shutil.rmtree, directory_path)
-
-#@timer
-async def rules(rq: RulesRequest) -> VectorDbResponse:
-  api_key = rq.api_key()
-  _config = rq.config()
-  db_name = rq.out_name()
-  store_path = os.path.join(_config.robot_data_folder, _config.robot_data_db_folder, _config.robot_data_db_folder_store, db_name)
-  try:
-    await FaissHelper.create([Document(page_content=rule, metadata={"source": "rules"}) for rule in rq.rules], store_path, api_key) #type: ignore
-    db_file_path = shutil.make_archive(os.path.join(_config.robot_data_folder, _config.robot_data_db_folder, _config.robot_data_db_folder_out, db_name), "zip", store_path)
-    return VectorDbResponse(file = os.path.basename(db_file_path))
-  except Exception as e:
-    await _cleanup_directory(store_path)
-    return VectorDbResponse(success = False, error = str(e))
-  finally:
-    gc.collect()
-
-#@atimer
-async def kb(rq: KbRequest) -> VectorDbResponse:
-  os.environ['MPLCONFIGDIR'] = './tmp/.matplotlib'
-  api_key = rq.api_key()
-  _config = rq.config()
-  db_name = rq.out_name()
-  src_path = os.path.join(_config.robot_data_folder, _config.robot_data_db_folder, _config.robot_data_db_folder_src)
-  working_path = os.path.join(src_path, db_name)
-
-  if all([not rq.files,not rq.endpoints,not rq.integrations]):
-    return VectorDbResponse(success = False, error = "No files, endpoints or integrations provided")
-  else:
-    await aiofiles.os.makedirs(src_path, exist_ok=True)
-    await aiofiles.os.makedirs(working_path, exist_ok=True)
-
-  documents: list[Document] = []
-  # Download/copy all files
-  if rq.files:
-    try:
-      loaders = Loader(working_path)
-      filter_file_extensions = loaders.managed_file_extensions()
-      files_to_download = [file for file in rq.files if not os.path.exists(os.path.join(src_path, os.path.basename(file)))]
-      if files_to_download:
-        await download_files(
-          [f"{_config.robot_cms_host}/{_config.robot_cms_kb_folder}/{os.path.basename(file)}" for file in files_to_download if any([file.endswith(ext) for ext in filter_file_extensions])],
-          src_path, authorization=_config.robot_cms_auth)
-      # copy files to working tmp folder
-      for file in rq.files:
-        async with aiofiles.open(os.path.join(src_path, os.path.basename(file)), 'rb') as src_file:
-          async with aiofiles.open(os.path.join(working_path, os.path.basename(file)), 'wb') as dest_file:
-            await dest_file.write(await src_file.read())
-      #load files
-      try:
-        documents.extend(await loaders.load())
-      except Exception as e:
-        tb = traceback.format_exc()
-        _error = f"File loader failure: {e} | {tb}"
-        logging.warning(_error)
-        return VectorDbResponse(success = False, error = _error)
-    except Exception as e:
-      await _cleanup_directory(working_path)
-      return VectorDbResponse(success = False, error = f"Failed to download file {e}")
-
-  if rq.endpoints:
-    try:
-      documents.extend(await load_endpoints(rq.endpoints, working_path))
-    except Exception as e:
-      await _cleanup_directory(working_path)
-      tb = traceback.format_exc()
-      _error = f"Endpoint failure: {e} | {tb}"
-      logging.warning(_error)
-      return VectorDbResponse(success = False, error = _error)
-
-  if rq.integrations:
-    tasks = []
-    for integration in rq.integrations:
-      tasks.append(
-        IntegrationManager
-        .get_strategy(integration.type.lower(), working_path, integration.__pydantic_extra__) #type: ignore
-        .load()
-      )
-    try:
-      integration_documents = await asyncio.gather(*tasks)
-      for docs in integration_documents:
-        documents.extend(docs)
-    except Exception as e:
-      await _cleanup_directory(working_path)
-      tb = traceback.format_exc()
-      _error = f"Integration failure: {e} | {tb}"
-      logging.warning(_error)
-      return VectorDbResponse(success=False, error=_error)
-
-  #cleanup
-  await _cleanup_directory(working_path)
-
-  if documents and len(documents) > 0:
-    try:
-      store_path = os.path.join(_config.robot_data_folder, _config.robot_data_db_folder, _config.robot_data_db_folder_store, db_name)
-      db_file_path = await aiofiles.os.wrap(shutil.make_archive)(
-          os.path.join(_config.robot_data_folder, _config.robot_data_db_folder, _config.robot_data_db_folder_out, db_name),
-          "zip",
-          await FaissHelper.create(documents, store_path, api_key, return_folder_path=True)
-      )
-      return VectorDbResponse(file = os.path.basename(db_file_path))
-    except Exception as e:
-      await _cleanup_directory(store_path)
-      return VectorDbResponse(success = False, error = str(e))
-    finally:
-      del documents
-      gc.collect()
-  else:
-    _error = "No documents found in the knowledgebase folder"
-    logging.warning(_error)
-    return VectorDbResponse(success = False, error = _error)
-
-async def kb_stream_file(filename: str):
-    file_path = os.path.join(config.robot_data_folder, config.robot_data_db_folder, config.robot_data_db_folder_out, filename)
-    if not os.path.isfile(file_path):
-        raise HTTPException(status_code=404, detail="File not found")
-    def iter_file():
-        with open(file_path, mode="rb") as file:
-            while chunk := file.read(1024*8):
-                yield chunk
-    return StreamingResponse(iter_file(), media_type="application/octet-stream", headers={"Content-Disposition": f"attachment; filename={filename}"})
+import os, gc, shutil, logging, traceback
+import asyncio, aiofiles, aiofiles.os
+from fastapi import HTTPException
+from fastapi.responses import StreamingResponse
+from langchain_core.documents import Document
+from ws_bom_robot_app.llm.vector_store.loader.base import Loader
+from ws_bom_robot_app.llm.models.api import RulesRequest, KbRequest, VectorDbResponse
+from ws_bom_robot_app.llm.vector_store.integration.manager import IntegrationManager
+from ws_bom_robot_app.llm.vector_store.db.manager import VectorDbManager
+from ws_bom_robot_app.config import config
+from ws_bom_robot_app.llm.models.kb import load_endpoints
+from ws_bom_robot_app.llm.utils.download import download_files
+
+async def _cleanup_directory(directory_path: str):
+  if os.path.exists(directory_path):
+    await asyncio.to_thread(shutil.rmtree, directory_path)
+
+#@timer
+async def rules(rq: RulesRequest) -> VectorDbResponse:
+  _config = rq.config()
+  db_name = rq.out_name()
+  store_path = os.path.join(_config.robot_data_folder, _config.robot_data_db_folder, _config.robot_data_db_folder_store, db_name)
+  try:
+    await VectorDbManager.get_strategy(rq.vector_type).create(rq.embeddings(),[Document(page_content=rule, metadata={"source": "rules"}) for rule in rq.rules], store_path) #type: ignore
+    db_file_path = shutil.make_archive(os.path.join(_config.robot_data_folder, _config.robot_data_db_folder, _config.robot_data_db_folder_out, db_name), "zip", store_path)
+    return VectorDbResponse(file = os.path.basename(db_file_path), vector_type=rq.vector_type)
+  except Exception as e:
+    try:
+      await _cleanup_directory(store_path)
+    finally:
+      return VectorDbResponse(success = False, error = str(e))
+  finally:
+    gc.collect()
+
+#@atimer
+async def kb(rq: KbRequest) -> VectorDbResponse:
+  os.environ['MPLCONFIGDIR'] = './tmp/.matplotlib'
+  _config = rq.config()
+  db_name = rq.out_name()
+  src_path = os.path.join(_config.robot_data_folder, _config.robot_data_db_folder, _config.robot_data_db_folder_src)
+  working_path = os.path.join(src_path, db_name)
+
+  if all([not rq.files,not rq.endpoints,not rq.integrations]):
+    return VectorDbResponse(success = False, error = "No files, endpoints or integrations provided")
+  else:
+    await aiofiles.os.makedirs(src_path, exist_ok=True)
+    await aiofiles.os.makedirs(working_path, exist_ok=True)
+
+  documents: list[Document] = []
+  # Download/copy all files
+  if rq.files:
+    try:
+      loaders = Loader(working_path)
+      filter_file_extensions = loaders.managed_file_extensions()
+      files_to_download = [file for file in rq.files if not os.path.exists(os.path.join(src_path, os.path.basename(file)))]
+      if files_to_download:
+        await download_files(
+          [f"{_config.robot_cms_host}/{_config.robot_cms_kb_folder}/{os.path.basename(file)}" for file in files_to_download if any([file.endswith(ext) for ext in filter_file_extensions])],
+          src_path, authorization=_config.robot_cms_auth)
+      # copy files to working tmp folder
+      for file in rq.files:
+        async with aiofiles.open(os.path.join(src_path, os.path.basename(file)), 'rb') as src_file:
+          async with aiofiles.open(os.path.join(working_path, os.path.basename(file)), 'wb') as dest_file:
+            await dest_file.write(await src_file.read())
+      #load files
+      try:
+        documents.extend(await loaders.load())
+      except Exception as e:
+        tb = traceback.format_exc()
+        _error = f"File loader failure: {e} | {tb}"
+        logging.warning(_error)
+        return VectorDbResponse(success = False, error = _error)
+    except Exception as e:
+      await _cleanup_directory(working_path)
+      return VectorDbResponse(success = False, error = f"Failed to download file {e}")
+
+  if rq.endpoints:
+    try:
+      documents.extend(await load_endpoints(rq.endpoints, working_path))
+    except Exception as e:
+      await _cleanup_directory(working_path)
+      tb = traceback.format_exc()
+      _error = f"Endpoint failure: {e} | {tb}"
+      logging.warning(_error)
+      return VectorDbResponse(success = False, error = _error)
+
+  if rq.integrations:
+    tasks = []
+    for integration in rq.integrations:
+      tasks.append(
+        IntegrationManager
+        .get_strategy(integration.type.lower(), working_path, integration.__pydantic_extra__) #type: ignore
+        .load()
+      )
+    try:
+      integration_documents = await asyncio.gather(*tasks)
+      for docs in integration_documents:
+        documents.extend(docs)
+    except Exception as e:
+      await _cleanup_directory(working_path)
+      tb = traceback.format_exc()
+      _error = f"Integration failure: {e} | {tb}"
+      logging.warning(_error)
+      return VectorDbResponse(success=False, error=_error)
+
+  #cleanup
+  await _cleanup_directory(working_path)
+
+  if documents and len(documents) > 0:
+    try:
+      store_path = os.path.join(_config.robot_data_folder, _config.robot_data_db_folder, _config.robot_data_db_folder_store, db_name)
+      db_file_path = await aiofiles.os.wrap(shutil.make_archive)(
+          os.path.join(_config.robot_data_folder, _config.robot_data_db_folder, _config.robot_data_db_folder_out, db_name),
+          "zip",
+          await VectorDbManager.get_strategy(rq.vector_type).create(rq.embeddings(), documents, store_path, return_folder_path=True)
+      )
+      return VectorDbResponse(file = os.path.basename(db_file_path), vector_type=rq.vector_type)
+    except Exception as e:
+      await _cleanup_directory(store_path)
+      return VectorDbResponse(success = False, error = str(e))
+    finally:
+      del documents
+      gc.collect()
+  else:
+    _error = "No documents found in the knowledgebase folder"
+    logging.warning(_error)
+    return VectorDbResponse(success = False, error = _error)
+
+async def kb_stream_file(filename: str):
+    file_path = os.path.join(config.robot_data_folder, config.robot_data_db_folder, config.robot_data_db_folder_out, filename)
+    if not os.path.isfile(file_path):
+        raise HTTPException(status_code=404, detail="File not found")
+    def iter_file():
+        with open(file_path, mode="rb") as file:
+            while chunk := file.read(1024*8):
+                yield chunk
+    return StreamingResponse(iter_file(), media_type="application/octet-stream", headers={"Content-Disposition": f"attachment; filename={filename}"})

ws_bom_robot_app/llm/vector_store/integration/base.py

@@ -4,11 +4,22 @@ from abc import ABC, abstractmethod
 from unstructured_ingest.v2.interfaces import ProcessorConfig
 from unstructured_ingest.v2.pipeline.pipeline import Pipeline, PartitionerConfig, FiltererConfig
 from typing import Union
+from ws_bom_robot_app.llm.utils.secrets import Secrets
 
 class IntegrationStrategy(ABC):
+  @classmethod
+  def _parse_data(cls, data: dict[str, Union[str, int, list]]) -> dict[str, Union[str, int, list]]:
+    for key, fn in (
+      ("__from_env", Secrets.from_env),
+      ("__from_file", Secrets.from_file),
+    ):
+      if key in data:
+        if secret := fn(data[key]):
+          return secret
+    return data
   def __init__(self, knowledgebase_path: str, data: dict[str, Union[str,int,list]]):
     self.knowledgebase_path = knowledgebase_path
-    self.data = data
+    self.data = self._parse_data(data)
     self.working_directory = os.path.join(self.knowledgebase_path,self.working_subdirectory())
     os.makedirs(self.working_directory, exist_ok=True)
   @property

ws_bom_robot_app/llm/vector_store/loader/base.py

@@ -13,6 +13,7 @@ from langchain_community.document_loaders import (
     CSVLoader,
     UnstructuredEmailLoader,
     UnstructuredImageLoader,
+    UnstructuredWordDocumentLoader,
     UnstructuredXMLLoader,
     UnstructuredPowerPointLoader,
     TextLoader
@@ -39,7 +40,7 @@ class Loader():
     '.org': None,
     '.odt': None,
     '.ppt': None,
-    '.pptx': LoaderConfig(loader=UnstructuredPowerPointLoader,kwargs={"strategy":"auto"}), #docling issue with WMF https://github.com/DS4SD/docling/issues/594
+    '.pptx': LoaderConfig(loader=DoclingLoader, kwargs={"fallback":UnstructuredPowerPointLoader, "strategy":"auto"}),
     '.txt': LoaderConfig(loader=TextLoader, kwargs={"autodetect_encoding": True}),
     '.rst': None,
     '.rtf': None,
@@ -49,15 +50,15 @@ class Loader():
     '.htm': LoaderConfig(loader=BSHTMLLoader),
     '.html': LoaderConfig(loader=BSHTMLLoader),
     ".pdf": LoaderConfig(loader=DoclingLoader),
-    '.png': LoaderConfig(loader=UnstructuredImageLoader,kwargs={"strategy":"auto","mode":"single"}),
-    '.jpg': LoaderConfig(loader=UnstructuredImageLoader,kwargs={"strategy":"auto","mode":"single"}),
-    '.jpeg': LoaderConfig(loader=UnstructuredImageLoader,kwargs={"strategy":"auto","mode":"single"}),
+    '.png': LoaderConfig(loader=DoclingLoader, kwargs={"fallback":UnstructuredImageLoader, "strategy":"auto"}),
+    '.jpg': LoaderConfig(loader=DoclingLoader, kwargs={"fallback":UnstructuredImageLoader, "strategy":"auto"}),
+    '.jpeg': LoaderConfig(loader=DoclingLoader, kwargs={"fallback":UnstructuredImageLoader, "strategy":"auto"}),
     '.gif': None,
     ".emf": None,
     ".wmf": None,
     '.tiff': None,
     '.doc': None, #see liberoffice dependency
-    '.docx': LoaderConfig(loader=DoclingLoader),
+    '.docx': LoaderConfig(loader=DoclingLoader, kwargs={"fallback":UnstructuredWordDocumentLoader, "strategy":"auto"}),
     '.xml': LoaderConfig(loader=UnstructuredXMLLoader,kwargs={"strategy":"auto"}),
     '.js': None,
     '.py': None,

ws_bom_robot_app/llm/vector_store/loader/docling.py

@@ -1,20 +1,27 @@
 import os, logging, traceback
-from typing import Iterator, AsyncIterator, Optional
+from typing import Any, Iterator, AsyncIterator, Optional, Union
 from langchain_core.document_loaders import BaseLoader
 from langchain_core.documents import Document
 from langchain_core.runnables import run_in_executor
-from docling.document_converter import DocumentConverter, InputFormat, PdfFormatOption
-from docling.datamodel.pipeline_options import PdfPipelineOptions, TableStructureOptions, TableFormerMode
+from docling.document_converter import DocumentConverter, InputFormat, PdfFormatOption, ImageFormatOption
+from docling.datamodel.pipeline_options import PdfPipelineOptions, TableStructureOptions, TableFormerMode, TesseractCliOcrOptions
+from langchain_community.document_loaders import UnstructuredFileLoader
 
 class DoclingLoader(BaseLoader):
-  def __init__(self, file_path: str | list[str]) -> None:
+  def __init__(self, file_path: str | list[str], **kwargs: Any) -> None:
       self._file_paths = file_path if isinstance(file_path, list) else [file_path]
       self._converter = DocumentConverter(format_options={
             InputFormat.PDF: PdfFormatOption(
               pipeline_options=PdfPipelineOptions(
                 table_structure_options=TableStructureOptions(mode=TableFormerMode.ACCURATE)
-              ))
+            )),
+            InputFormat.IMAGE: ImageFormatOption(
+              pipeline_options=PdfPipelineOptions(
+                ocr_options=TesseractCliOcrOptions(lang=["auto"]),
+                table_structure_options=TableStructureOptions(mode=TableFormerMode.ACCURATE)
+            ))
         })
+      self._kwargs = kwargs
   def load(self) -> list[Document]:
       """Load data into Document objects."""
       return list(self.lazy_load())
@@ -40,4 +47,18 @@ class DoclingLoader(BaseLoader):
             text = doc.export_to_markdown(image_placeholder="")
             yield Document(page_content=text, metadata={"source": source})
           except Exception as e:
-            logging.warning(f"Failed to load document from {source}: {e} | {traceback.format_exc()}")
+            if 'fallback' in self._kwargs:
+                if issubclass(self._kwargs['fallback'], (BaseLoader, UnstructuredFileLoader)):
+                  logging.info(f"Using fallback loader {self._kwargs['fallback']} for {source}")
+                  try:
+                    loader: Union[BaseLoader, UnstructuredFileLoader] = self._kwargs['fallback'](
+                       source,
+                       **{k: v for k, v in self._kwargs.items() if k != 'fallback'}
+                       )
+                    yield from loader.lazy_load()
+                  except Exception as e:
+                    logging.warning(f"Failed to load document from {source}: {e} | {traceback.format_exc()}")
+                else:
+                   logging.warning(f"Invalid fallback loader {self._kwargs['fallback']}[{type(self._kwargs['fallback'])}] for {source}")
+            else:
+              logging.warning(f"Failed to load document from {source}: {e} | {traceback.format_exc()}")