feat(llm): increase Gemini Flash token limit to 400k

Author: lzwjava

scripts/llm/openrouter_client.py

@@ -31,7 +31,7 @@
 DEFAULT_TOKENS = {
     "claude-opus": 8192,
     "claude-sonnet": 8192,
-    "gemini-flash": 32768,
+    "gemini-flash": 400000,
     "gemini-pro": 8192,
     "kimi-k2": 32768,
     "deepseek-v3": 32768,

scripts/translation/translate_client.py

@@ -120,7 +120,7 @@ def run_translate(text, target, kind, model, front_matter, orig_lang, need_en, s
     # Use language-specific model mapping, ignore the model parameter
     actual_model = LANGUAGE_MODEL_MAP.get(target, "mistral-medium")
     prompt = build_prompt_template(target, kind, front_matter) + "\n\n" + text
-    translated = clean_response(call_openrouter_api(prompt, "gemini-flash", max_tokens=400000))
+    translated = clean_response(call_openrouter_api(prompt, model))
     check_commentary(translated)
     if kind == "title":
         check_title_strict(translated, target)