chore: convert prompts to JSON format

Author: Huanshere

config.yaml

@@ -1,14 +1,15 @@
 # * Settings marked with * are advanced settings that won't appear in the Streamlit page and can only be modified manually in config.py
 # recommend to set in streamlit page
-version: "2.2.3"
+version: "3.0.0"
 ## ======================== Basic Settings ======================== ##
 display_language: "zh-CN"
 
 # API settings
 api:
-  key: 'your_api_key'
-  base_url: 'https://api.302.ai'
-  model: 'gemini-2.0-flash'
+  key: 'your-api-key'
+  base_url: 'https://openrouter.ai/api'
+  model: 'deepseek/deepseek-chat-v3-0324'
+  llm_support_json: true
 
 # Language settings, written into the prompt, can be described in natural language
 target_language: '简体中文'
@@ -145,13 +146,6 @@ allowed_audio_formats:
 - 'flac'
 - 'm4a'
 
-# LLMs that support returning JSON format
-llm_support_json:
-- 'gpt-4o'
-- 'gpt-4o-mini'
-- 'gemini-2.0-flash'
-- 'deepseek-chat'
-
 # Spacy models
 spacy_model_map:
   en: 'en_core_web_md'

core/_4_2_translate.py

@@ -13,7 +13,7 @@
 console = Console()
 
 # Function to split text into chunks
-def split_chunks_by_chars(chunk_size=400, max_i=8): 
+def split_chunks_by_chars(chunk_size, max_i): 
     """Split text into chunks based on character count, return a list of multi-line text chunks"""
     with open(_3_2_SPLIT_BY_MEANING, "r", encoding="utf-8") as file:
         sentences = file.read().strip().split('\n')

core/prompts.py

@@ -223,55 +223,47 @@ def get_prompt_expressiveness(faithfulness_result, lines, shared_prompt):
 ## ================================================================
 # @ step6_splitforsub.py
 def get_align_prompt(src_sub, tr_sub, src_part):
-    TARGET_LANGUAGE = load_key("target_language")
-    src_language = load_key("whisper.detected_language")
+    targ_lang = load_key("target_language")
+    src_lang = load_key("whisper.detected_language")
     src_splits = src_part.split('\n')
     num_parts = len(src_splits)
     src_part = src_part.replace('\n', ' [br] ')
-    align_prompt = '''
+    align_parts_json = ','.join(
+        f'''
+        {{
+            "src_part_{i+1}": "{src_splits[i]}",
+            "target_part_{i+1}": "Corresponding aligned {targ_lang} subtitle part"
+        }}''' for i in range(num_parts)
+    )
+
+    align_prompt = f'''
 ## Role
-You are a Netflix subtitle alignment expert fluent in both {src_language} and {target_language}.
+You are a Netflix subtitle alignment expert fluent in both {src_lang} and {targ_lang}.
 
 ## Task
-We have {src_language} and {target_language} original subtitles for a Netflix program, as well as a pre-processed split version of {src_language} subtitles. Your task is to create the best splitting scheme for the {target_language} subtitles based on this information.
+We have {src_lang} and {targ_lang} original subtitles for a Netflix program, as well as a pre-processed split version of {src_language} subtitles. Your task is to create the best splitting scheme for the {target_language} subtitles based on this information.
 
-1. Analyze the word order and structural correspondence between {src_language} and {target_language} subtitles
-2. Split the {target_language} subtitles according to the pre-processed {src_language} split version
+1. Analyze the word order and structural correspondence between {src_lang} and {targ_lang} subtitles
+2. Split the {targ_lang} subtitles according to the pre-processed {src_lang} split version
 3. Never leave empty lines. If it's difficult to split based on meaning, you may appropriately rewrite the sentences that need to be aligned
 4. Do not add comments or explanations in the translation, as the subtitles are for the audience to read
 
 ## INPUT
 <subtitles>
-{src_language} Original: "{src_sub}"
-{target_language} Original: "{tr_sub}"
-Pre-processed {src_language} Subtitles ([br] indicates split points): {src_part}
+{src_lang} Original: "{src_sub}"
+{targ_lang} Original: "{tr_sub}"
+Pre-processed {src_lang} Subtitles ([br] indicates split points): {src_part}
 </subtitles>
 
 ## Output in only JSON format
 {{
-    "analysis": "Brief analysis of word order, structure, and semantic correspondence between {src_language} and {target_language} subtitles",
+    "analysis": "Brief analysis of word order, structure, and semantic correspondence between two subtitles",
     "align": [
         {align_parts_json}
     ]
 }}
 '''
-
-    align_parts_json = ','.join(
-        f'''
-        {{
-            "src_part_{i+1}": "{src_splits[i]}",
-            "target_part_{i+1}": "Corresponding aligned {TARGET_LANGUAGE} subtitle part"
-        }}''' for i in range(num_parts)
-    )
-
-    return align_prompt.format(
-        src_language=src_language,
-        target_language=TARGET_LANGUAGE,
-        src_sub=src_sub,
-        tr_sub=tr_sub,
-        src_part=src_part,
-        align_parts_json=align_parts_json,
-    )
+    return align_prompt
 
 ## ================================================================
 # @ step8_gen_audio_task.py @ step10_gen_audio.py

core/utils/ask_gpt.py

@@ -58,7 +58,7 @@ def ask_gpt(prompt, resp_type=None, valid_def=None, log_title="default"):
     elif 'v1' not in base_url:
         base_url = base_url.strip('/') + '/v1'
     client = OpenAI(api_key=load_key("api.key"), base_url=base_url)
-    response_format = {"type": "json_object"} if resp_type == "json" and model in load_key("llm_support_json") else None
+    response_format = {"type": "json_object"} if resp_type == "json" and load_key("api.llm_support_json") else None
 
     messages = [{"role": "user", "content": prompt}]