[Feature Request] Claude / Anthropic API Support

[kacperpaluch]

Do you have any plans to add support for Claude / Anthropic API?

[piazzatron]

Definitely. Plan to add Claude/Gemini in the future

[wadid]

Very interesting are also the Openrouter versions of anthropic models which are somehow called "moderated" :D. But the funny thing is that these "moderated" versions are very easy to jailbreak. That is probably the purpose why they are offered by Openrouter.

[anthropic/claude-3-sonnet:beta](https://openrouter.ai/models/anthropic/claude-3-haiku)

just add a prefill assistant message to them: https://docs.anthropic.com/en/docs/build-with-claude/prompt-engineering/prefill-claudes-response

[piazzatron]

Claude coming in the next version in a couple weeks!

[piazzatron]

Hey @kacperpaluch, just released version 2.0 (if you refresh your addons in Anki), which supports Claude Sonnet 3.5 and Haiku 3

[wadid]

Great!!!! :) Regarding the prefilling... something like the following is not possible I guess? https://docs.anthropic.com/en/docs/build-with-claude/prompt-engineering/prefill-claudes-response

[piazzatron]

It would be possible (as well as inputting system messages to influence general behavior) - can you give me an example of what you're trying to accomplish with this?

[wadid]

haha... just if you want to include banned stuff in your cards :D you could use it as advertisement for your plugin ;)