If pass multi files, e.g. from gradio, then show those files that hav…

…e no handler for instead of just filtering

gpt_langchain.py

@@ -512,7 +512,7 @@ def file_to_doc(file, base_path=None, verbose=False, fail_any_exception=False, c
             # recurse
             doc1 = path_to_docs(base_path, verbose=verbose, fail_any_exception=fail_any_exception)
     else:
-        raise RuntimeError("No file handler for %s" % file)
+        raise RuntimeError("No file handler for %s" % os.path.basename(file))
 
     # allow doc1 to be list or not.  If not list, did not chunk yet, so chunk now
     if not isinstance(doc1, list):
@@ -580,26 +580,27 @@ def path_to_docs(path_or_paths, verbose=False, fail_any_exception=False, n_jobs=
     globs_image_types = []
     globs_non_image_types = []
     if url:
-        globs = [url]
+        globs_non_image_types = [url]
     elif text:
-        globs = [text]
+        globs_non_image_types = [text]
     elif isinstance(path_or_paths, str):
-        # single file
+        # single path, only consume allowed files
         path = path_or_paths
         # Below globs should match patterns in file_to_doc()
-        [globs_image_types.extend(glob.glob(os.path.join(path, "./**/*.%s" % ftype), recursive=True)) for ftype in
-         image_types]
-        [globs_non_image_types.extend(glob.glob(os.path.join(path, "./**/*.%s" % ftype), recursive=True)) for ftype in
-         non_image_types]
-        globs = globs_non_image_types + globs_image_types
+        [globs_image_types.extend(glob.glob(os.path.join(path, "./**/*.%s" % ftype), recursive=True))
+         for ftype in image_types]
+        [globs_non_image_types.extend(glob.glob(os.path.join(path, "./**/*.%s" % ftype), recursive=True))
+         for ftype in non_image_types]
     else:
-        # list/tuple of files
+        # list/tuple of files (consume what can, and exception those that selected but cannot consume so user knows)
         assert isinstance(path_or_paths, (list, tuple)), "Wrong type for path_or_paths: %s" % type(path_or_paths)
-        globs = path_or_paths
         # reform out of allowed types
-        globs_image_types = flatten_list([[x for x in globs if x.endswith(y)] for y in image_types])
-        globs_non_image_types = flatten_list([[x for x in globs if x.endswith(y)] for y in non_image_types])
-        globs = globs_non_image_types + globs_image_types
+        globs_image_types = flatten_list([[x for x in path_or_paths if x.endswith(y)] for y in image_types])
+        # could do below:
+        # globs_non_image_types = flatten_list([[x for x in path_or_paths if x.endswith(y)] for y in non_image_types])
+        # But instead, allow fail so can collect unsupported too
+        set_globs_image_types = set(globs_image_types)
+        globs_non_image_types = [x for x in path_or_paths if x not in set_globs_image_types]
     # could use generator, but messes up metadata handling in recursive case
     if caption_loader and not isinstance(caption_loader, (bool, str)) and \
             caption_loader.device != 'cpu' or \

gradio_runner.py

@@ -1234,13 +1234,13 @@ def update_user_db(file, db1, x, y, *args, dbs=None, langchain_mode='UserData',
 
 
 def _update_user_db(file, db1, x, y, dbs=None, db_type=None, langchain_mode='UserData', use_openai_embedding=False,
-                   hf_embedding_model="sentence-transformers/all-MiniLM-L6-v2",
-                   caption_loader=None,
-                   enable_captions=True,
-                   captions_model="Salesforce/blip-image-captioning-base",
-                   enable_ocr=False,
-                   verbose=False,
-                   chunk=True, chunk_size=512, is_url=False, is_txt=False):
+                    hf_embedding_model="sentence-transformers/all-MiniLM-L6-v2",
+                    caption_loader=None,
+                    enable_captions=True,
+                    captions_model="Salesforce/blip-image-captioning-base",
+                    enable_ocr=False,
+                    verbose=False,
+                    chunk=True, chunk_size=512, is_url=False, is_txt=False):
     assert isinstance(dbs, dict), "Wrong type for dbs: %s" % str(type(dbs))
     assert db_type in ['faiss', 'chroma'], "db_type %s not supported" % db_type
     from gpt_langchain import add_to_db, get_db, path_to_docs