openai-chat-tokens

What's Changed

• feat: support anyOf in function parameters by @b0o in https://github.com/hmarr/openai-chat-tokens/pull/13
• Update openai types by @hmarr in https://github.com/hmarr/openai-chat-tokens/pull/15

Full Changelog: https://github.com/hmarr/openai-chat-tokens/compare/v0.2.7...v0.2.8

What's Changed

• feat: add support for function_call option by @b0o in https://github.com/hmarr/openai-chat-tokens/pull/10
• Add debug step to log OIDC ID token by @feelepxyz in https://github.com/hmarr/openai-chat-tokens/pull/11

New Contributors

• @b0o made their first contribution in https://github.com/hmarr/openai-chat-tokens/pull/10
• @feelepxyz made their first contribution in https://github.com/hmarr/openai-chat-tokens/pull/11

Full Changelog: https://github.com/hmarr/openai-chat-tokens/compare/v0.2.6...v0.2.7

What's Changed

• Fix estimate for nested descriptions by @piotrkuczko in https://github.com/hmarr/openai-chat-tokens/pull/7
• Update OpenAI library & types

New Contributors

• @piotrkuczko made their first contribution in https://github.com/hmarr/openai-chat-tokens/pull/7

Full Changelog: https://github.com/hmarr/openai-chat-tokens/compare/v0.2.5...v0.2.6

What's Changed

• [fix] edge cases for name & function role by @iwasrobbed in https://github.com/hmarr/openai-chat-tokens/pull/6

New Contributors

• @iwasrobbed made their first contribution in https://github.com/hmarr/openai-chat-tokens/pull/6

Full Changelog: https://github.com/hmarr/openai-chat-tokens/compare/v0.2.4...v0.2.5

What's Changed

• Add Support for Array Types by @petrbrzek in https://github.com/hmarr/openai-chat-tokens/pull/4

New Contributors

• @petrbrzek made their first contribution in https://github.com/hmarr/openai-chat-tokens/pull/4

Full Changelog: https://github.com/hmarr/openai-chat-tokens/compare/v0.2.3...v0.2.4

• Pad first system message with newline, fixing off-by-one discrepancy

• Adjust for the combination of a system message and functions (reduces token usage by 3)
• Adjust for functions without descriptions (removes the description line altogether)

• Include function_call property in estimate

• Switch to pure JavaScript tokenizer (js-tiktoken)
• Stop unnecessarily reloading the tiktoken encoding

Initial release