#7656 Bug: Empty JSON chat history files cause JSONDecodeError (very rare)

Describe the bug

When using text-generation-webui with a remote llama.cpp server (http://192.168.1.24:12546/), 0-byte JSON files are created in user_data/logs/chat/<character>/ and user_data/logs/instruct/. Attempting to list chat histories or load a session throws: json.JSONDecodeError: Expecting value: line 1 column 1 (char 0)

Summary of Code Changes (How to fixed it) Here are the three fixes applied to G:\textgen-main\modules\chat.py to handle empty/0-byte JSON chat history files:

find_all_histories_with_first_prompts (lines 1903-1904) Purpose: Skip empty JSON files when listing chat histories for the UI sidebar.

if not file_content:
    continue

Added after reading file content, before searching. Prevents empty files from appearing in history list.

load_history (lines 2043-2045) Purpose: Gracefully handle empty JSON files when loading a specific chat session.

with open(p, 'rb') as fh:
    content = fh.read()
if not content:
    return {'internal': [], 'visible': [], 'metadata': {}}
f = json.loads(content)

Returns empty history structure instead of raising json.JSONDecodeError on 0-byte files.

load_history_json (lines 2071-2072) Purpose: Handle empty file uploads/imports in the chat history import feature.

if not file.strip():
    return history

Returns original history unchanged if uploaded file is empty/whitespace.

Is there an existing issue for this?

I have searched the existing issues

Reproduction

None

Specific Circumstances

This issue occurs specifically when:

Using a llama.cpp server via the --llama-cpp-server flag or API endpoint configuration (?) ,or loading/create a new char/chat on a remote device.
Starting a new chat session with a character - the server creates a placeholder JSON file before the first.message exchange
Network interruption or timeout during the first model response - leaves a 0-byte file that never gets written to.
Switching characters/models mid-session - triggers new history file creation that may remain empty if the switch fails.
High latency connections to the llama.cpp server (>5s response time) where the UI creates the history file but the request times out before content is written.

The issue does not occur with:

Local model loading (transformers, GGUF via llama.cpp Python bindings, ExLlamaV2, etc.)
Completed chat sessions that have at least one message exchange
Fresh installations with no prior chat history

Fix Applied

Three defensive checks added to modules/chat.py:

find_all_histories_with_first_prompts (line 1903-1904): Skip empty files when building history list
load_history (lines 2043-2045): Return empty history struct instead of crashing on empty file
load_history_json (lines 2071-2072): Ignore empty uploaded history files

Files Changed

modules/chat.py: Added 3 guard clauses for empty file handling

Screenshot

⚠️ AI-Assisted Fix Disclaimer This fix was written with AI assistance. I do not have programming expertise to fully verify the correctness and completeness of these changes. Please review and test thoroughly before merging. Specifically:

The fix handles the symptom (empty files) but may not address why llama.cpp creates 0-byte files in the first place
Edge cases around concurrent file access or race conditions may exist
The background cleanup script is a workaround, not a root cause fix
Community testing with various llama.cpp server versions and configurations is needed
Due to that, I won't push that codes changed to PR

Your verification and feedback are greatly appreciated.

Logs

Traceback (most recent call last):
  File "G:\textgen-main\installer_files\env\Lib\site-packages\gradio\queueing.py", line 549, in process_events
    response = await route_utils.call_process_api(
               ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
    ...<5 lines>...
    )
    ^
  File "G:\textgen-main\installer_files\env\Lib\site-packages\gradio\route_utils.py", line 276, in call_process_api
    output = await app.get_blocks().process_api(
             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
    ...<11 lines>...
    )
    ^
  File "G:\textgen-main\installer_files\env\Lib\site-packages\gradio\blocks.py", line 1904, in process_api
    result = await self.call_function(
             ^^^^^^^^^^^^^^^^^^^^^^^^^
    ...<8 lines>...
    )
    ^
  File "G:\textgen-main\installer_files\env\Lib\site-packages\gradio\blocks.py", line 1490, in call_function
    prediction = await anyio.to_thread.run_sync(
                 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
        fn, *processed_input, limiter=self.limiter
        ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
    )
    ^
  File "G:\textgen-main\installer_files\env\Lib\site-packages\anyio\to_thread.py", line 63, in run_sync
    return await get_async_backend().run_sync_in_worker_thread(
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
        func, args, abandon_on_cancel=abandon_on_cancel, limiter=limiter
        ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
    )
    ^
  File "G:\textgen-main\installer_files\env\Lib\site-packages\anyio\_backends\_asyncio.py", line 2518, in run_sync_in_worker_thread
    return await future
           ^^^^^^^^^^^^
  File "G:\textgen-main\installer_files\env\Lib\site-packages\anyio\_backends\_asyncio.py", line 1002, in run
    result = context.run(func, *args)
  File "G:\textgen-main\installer_files\env\Lib\site-packages\gradio\\\utils.py", line 812, in wrapper
    response = f(*args, **kwargs)
  File "G:\textgen-main\modules\chat.py", line 2546, in handle_start_new_chat_click
    histories = find_all_histories_with_first_prompts(state)
  File "G:\textgen-main\modules\chat.py", line 1906, in find_all_histories_with_first_prompts
    data = json.loads(file_content)
  File "G:\textgen-main\installer_files\env\Lib\json\__init__.py", line 352, in loads
    return _default_decoder.decode(s)
           ~~~~~~~~~~~~~~~~~~~~~~~^^^
  File "G:\textgen-main\installer_files\env\Lib\json\decoder.py", line 345, in decode
    obj, end = self.raw_decode(s, idx=_w(s, 0).end())
               ~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^
  File "G:\textgen-main\installer_files\env\Lib\json\decoder.py", line 363, in raw_decode
    raise JSONDecodeError("Expecting value", s, err.value) from None
json.decoder.JSONDecodeError: Expecting value: line 1 column 1 (char 0)

System Info

Windows 11 Pro Workstation
Gunnir Intel Arc B580 Graphics
Intel i5-14600k