Consider providing `chat_anthropic()` and `ProvideAnthropic` aliases for Claude

[gadenbuie]

All of Anthropic's models are provided under the Claude moniker, but it feels odd to switch between e.g. chat_openai() and chat_claude() when the company providing the claude model is Anthropic. To me chat_claude() feels aligned with chat_chatgpt(), but I may be missing a differentiating detail.

I think it's worth considering if an alias could be used, either moving the implementation to Anthropic-forward names or aliasing the existing functions/classes to Anthopic names.

Happy to provide a PR depending on thoughts and feedback.

[hadley]

If you look at the multi-name providers listed on the homepage:

Anthropic’s Claude: chat_claude().
AWS Bedrock: chat_bedrock().
Azure OpenAI: chat_azure().
Google Gemini: chat_gemini().
Snowflake Cortex: chat_snowflake() and chat_cortex_analyst().

There's not a lot of consistency, but it looks like we generally lean towards using the model name rather than the company name. So maybe it should be chat_chatgpt()? But as soon as I write that I realise why that looks so weird.

Or maybe we should be more consistent and do both? chat_anthropic_claude(), chat_aws_bedrock(), chat_azure_openai(), ...?

[hadley]

Argument in favour of chat_anthropic(): https://www.databricks.com/blog/anthropic-claude-37-sonnet-now-natively-available-databricks

[gadenbuie]

Personally, because the chat_*() functions link Chat to a specific Provider, I'd love to see alignment around the provider, i.e. the service name or company name, over the primary model name. So from your list, I'd lean toward the following

Anthropic’s Claude: chat_claude() → chat_anthropic()
AWS Bedrock: chat_bedrock() → chat_bedrock() (or maybe chat_aws_bedrock() to be specific)
Azure OpenAI: chat_azure() → chat_azure_openai()
Google Gemini: chat_gemini() → chat_google_gemini() or chat_gemini()
Snowflake Cortex: chat_snowflake() and chat_cortex_analyst() → chat_snowflake_cortex() and chat_snowflake_cortex_analyst().

The shorter forms like chat_bedrock(), chat_azure(), chat_snowflake() etc. seem fine to me if there's a clear primary service. For example, I think it's clear chat_bedrock() means AWS Bedrock, but chat_azure() could be Azure OpenAI or maybe Azure AI. chat_gemini() is interesting because even though the primary model is gemini, the name of the service is the Gemini API and chat_google() could confuse Gemini with Vertex AI (which could be chat_vertex()).

[hadley]

Yeah, I think for azure, AWS, and google it's particularly important to include the company name in the provider since it seems likely that these bigger orgs will implement additional brands (which we've already seen with gemini vs vertex AI).

[hadley]

I'd love a PR that provided these new functions and deprecated the old ones. (Feels like this should be a good task for cursor code or similar).