新版本容易出现报错

[assassinliujie]

Exception: Still failed after 3 attempts: JSON parsing still failed after 3 attempts: ❎ API response error: Missing required key: split Please check your network connection or API key or output/gpt_log/error.json to debug. Traceback: File "/root/anaconda3/envs/videoaaa/lib/python3.10/site-packages/streamlit/runtime/scriptrunner/exec_code.py", line 88, in exec_func_with_error_handling result = func() File "/root/anaconda3/envs/videoaaa/lib/python3.10/site-packages/streamlit/runtime/scriptrunner/script_runner.py", line 590, in code_to_exec exec(code, module.dict) File "/root/VideoLingo/st.py", line 123, in main() File "/root/VideoLingo/st.py", line 119, in main text_processing_section() File "/root/VideoLingo/st.py", line 33, in text_processing_section process_text() File "/root/VideoLingo/st.py", line 50, in process_text step3_2_splitbymeaning.split_sentences_by_meaning() File "/root/VideoLingo/core/step3_2_splitbymeaning.py", line 119, in split_sentences_by_meaning sentences = parallel_split_sentences(sentences, max_length=load_key("max_split_length"), max_workers=load_key("max_workers"), nlp=nlp, retry_attempt=retry_attempt) File "/root/VideoLingo/core/step3_2_splitbymeaning.py", line 101, in parallel_split_sentences split_result = future.result() File "/root/anaconda3/envs/videoaaa/lib/python3.10/concurrent/futures/_base.py", line 458, in result return self.get_result() File "/root/anaconda3/envs/videoaaa/lib/python3.10/concurrent/futures/_base.py", line 403, in get_result raise self._exception File "/root/anaconda3/envs/videoaaa/lib/python3.10/concurrent/futures/thread.py", line 58, in run result = self.fn(*self.args, *self.kwargs) File "/root/VideoLingo/core/step3_2_splitbymeaning.py", line 60, in split_sentence response_data = ask_gpt(split_prompt + ' ' retry_attempt, response_json=True, valid_def=valid_split, log_title='sentence_splitbymeaning') File "/root/VideoLingo/core/ask_gpt.py", line 103, in ask_gpt raise Exception(f"Still failed after {max_retries} attempts: {e}") 多个视频出现这种情况，之前从来没有遇到过，是否是切分算法有调整。现在很难完整切完一个视频

[assassinliujie]

用的claude-3-5-sonnet

[assassinliujie]

nAI staff working on the browser[br]aren't the same ones who accidentally erased critical evidence\"\n}\n\n\nThis split adheres to the Netflix subtitle standards by:\n1. Maintaining sentence meaning coherence\n2. Keeping both parts roughly equal in length (9 words and 10 words)\n3. Splitting at a natural point before the word \"aren't\"\n4. Ensuring each part is at least 3 words long\n5. Keeping each part under 20 words\n\nThe [br] tag is placed at the split position as requested.", "message": "json_repair parsing failed." }, { "model": "claude-3-5-sonnet-all", "prompt": "### Role\nYou are a professional Netflix subtitle splitter in en.\n\n### Task\nSplit the given subtitle text into 2 parts, each less than 20 words.\n\n### Instructions\n1. Maintain sentence meaning coherence according to Netflix subtitle standards\n2. Keep parts roughly equal in length (minimum 3 words each)\n3. Split at natural points like punctuation marks or conjunctions\n\n### Output Format in JSON\n{\n \"analysis\": \"Brief analysis of the text structure\",\n \"split\": \"Complete sentence with [br] tags at split positions\"\n}\n\n### Given Text\n<split_this_sentence>\nI can only hope the OpenAI staff working on the browser aren't the same ones who accidentally erased critical evidence\n</split_this_sentence>", "response": [ { "analysis": "The sentence is 19 words long, making it suitable for a single split. The natural break point is after 'browser', which creates two balanced parts.", "split": "I can only hope the OpenAI staff working on the browser[br]aren't the same ones who accidentally erased critical evidence" }, [ "br] tag is placed at the split position as requested." ] ], "message": "Missing required key: `split`" }, { "model": "claude-3-5-sonnet-all", "prompt": "### Role\nYou are a professional Netflix subtitle splitter in en.\n\n### Task\nSplit the given subtitle text into 2 parts, each less than 20 words.\n\n### Instructions\n1. Maintain sentence meaning coherence according to Netflix subtitle standards\n2. Keep parts roughly equal in length (minimum 3 words each)\n3. Split at natural points like punctuation marks or conjunctions\n\n### Output Format in JSON\n{\n \"analysis\": \"Brief analysis of the text structure\",\n \"split\": \"Complete sentence with [br] tags at split positions\"\n}\n\n### Given Text\n<split_this_sentence>\nI can only hope the OpenAI staff working on the browser aren't the same ones who accidentally erased critical evidence\n</split_this_sentence>", "response": "Here's the subtitle split for the given text, formatted as requested:\n\njson\n{\n \"analysis\": \"The sentence is 19 words long, making it suitable for a single split. The natural break point is after 'browser', which creates two balanced parts.\",\n \"split\": \"I can only hope the OpenAI staff working on the browser[br]aren't the same ones who accidentally erased critical evidence\"\n}\n```\n\nThis split adheres to the Netflix subtitle standards by:\n1. Maintaining sentence meaning coherence\n2. Keeping both parts roughly equal in length (9 words and 10 words)\n3. Splitting at a natural point before the word \"aren't\"\n4. Ensuring each part is at least 3 words long\n5. Keeping each part under 20 words\n\nThe [br] tag is placed at the split position as requested.", "message": "json_repair parsing failed." }

[Huanshere]

“Missing required key: split Please check your network connection or API key or output/gpt_log/error.json to debug.”

请问使用的什么来源的 claude 呢？看起来不是 claude 不够聪明，而是响应不稳定。