Aug 12

1. 02:58

   fix: disable GLM thinking (reasoning model burns max_tokens on reasoning_content -> empty answers); bump max_tokens to 500

   balade
   08aba80
2. 02:50

   Cokro LLM comparison: GLM-4.7-Flash (free) vs DeepSeek v4-flash

   balade
   aca0ec1
3. 02:50

   initial commit

   balade
   Verified signature

   36b1bf4