BrokenResponseError: Can not build a coherent char history after a broken streaming response (See the previous Exception fro details). To inspect the last response object, use `chat.last`.To remove the last request/response `Content` objects from the chat call `last_send, last_received = chat.rewind()` and continue without it

reloginn commented 5 months ago

Traceback (most recent call last): File "", line 1, in File "/home/asahi/Проекты/Python/gemini/gemini/init.py", line 5, in main run() File "/home/asahi/Проекты/Python/gemini/gemini/main.py", line 81, in run process_directory(SRC, DST) File "/home/asahi/Проекты/Python/gemini/gemini/main.py", line 73, in process_directory translate_file(src_path=src_path, dst_path=dst_path) File "/home/asahi/Проекты/Python/gemini/gemini/main.py", line 48, in translate_file translated = translate(src_path) ^^^^^^^^^^^^^^^^^^^ File "/home/asahi/Проекты/Python/gemini/gemini/main.py", line 36, in translate second_response = chat.send_message(content=MESSAGE_FOR_SECOND_PART + second_part, stream=True) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/home/asahi/.cache/pypoetry/virtualenvs/gemini-UM_oiH3g-py3.12/lib/python3.12/site-packages/google/generativeai/generative_models.py", line 467, in send_message history = self.history[:] ^^^^^^^^^^^^ File "/home/asahi/.cache/pypoetry/virtualenvs/gemini-UM_oiH3g-py3.12/lib/python3.12/site-packages/google/generativeai/generative_models.py", line 686, in history raise generation_types.BrokenResponseError( google.generativeai.types.generation_types.BrokenResponseError: Can not build a coherent char history after a broken streaming response (See the previous Exception fro details). To inspect the last response object, use chat.last.To remove the last request/response Content objects from the chat call last_send, last_received = chat.rewind() and continue without it.

code:

# --snip--
chat = model.start_chat()
PART_LEN = len(content) // 2
first_part = content[:PART_LEN]
second_part = content[PART_LEN:]
first_response = chat.send_message(content=MESSAGE_FOR_FIRST_PART + first_part, stream=True)
for chunk in first_response:
        chunks.append(chunk)
second_response = chat.send_message(content=MESSAGE_FOR_SECOND_PART + second_part, stream=True)
for chunk in second_response:
        chunks.append(chunk)

MarkDaoust commented 5 months ago

Typo fix: https://github.com/google-gemini/generative-ai-python/pull/313

It's trying to tell you that the first stream broke, and it's not sure what to do. Is the first part not raising an error?

reloginn commented 4 months ago

The first part doesn't give any errors. The second part gives an error, previously everything worked, with identical code (I automate translation, and I was able to translate about 15 .md files), then just started to pop up this error and that's it, nothing helps.

reloginn commented 4 months ago

def translate(path: str) -> str:
    file = io.FileIO(file=path)
    content = str(file.read())
    model = genai.GenerativeModel(model_name=MODEL, generation_config=GENERATION_CONFIG)
    chunks = []
    chat = model.start_chat()
    PART_LEN = len(content) // 2
    first_part = content[:PART_LEN]
    second_part = content[PART_LEN:]
    first_response = chat.send_message(content=MESSAGE_FOR_FIRST_PART + first_part, stream=True)
    for chunk in first_response:
        chunks.append(chunk)
    second_response = chat.send_message(content=MESSAGE_FOR_SECOND_PART + second_part, stream=True)
    for chunk in second_response:
        chunks.append(chunk)
    result = "".join(part.text for chunk in chunks for candidate in chunk.candidates for part in candidate.content.parts)
    return result

The error occurs here:

second_response = chat.send_message(content=MESSAGE_FOR_SECOND_PART + second_part, stream=True)

google-gemini / generative-ai-python