Release 0.12.2

Released 2018-05-15.

• Fixed token counting for long prompts
• Added per-model rate limits
• Added support for custom upstream timeouts

Release 39735 · Release 39736 · Release 7282

Back to the changelog