[FEATURE] Integrate with Antropic API

[badmonster0]

CocoIndex 🥥 currently supports OpenAI and Ollama ✨ , to add LLM as part of the data pipeline. https://cocoindex.io/docs/ai/llm

Here is an example of how Cocoindex uses Ollama to extract structured information from PDF.
https://cocoindex.io/blogs/cocoindex-ollama-structured-extraction-from-pdf/

We would like to add support for Anthropic API - https://docs.anthropic.com/en/api/getting-started

Related code to support OpenAI

Steps:

1. Update Rust code:
   • Add Enum to LlmApiType in Rust:

     cocoindex/src/llm/mod.rs

     Lines 9 to 12 in 801ae8f

     	pub enum LlmApiType {
     	Ollama,
     	OpenAi,
     	}

   • Create Anthropic client similar to OpenAI client: https://github.com/cocoindex-io/cocoindex/blob/main/src/llm/openai.rs
   • Update the router to connect to the Anthropic client:

     cocoindex/src/llm/mod.rs

     Lines 53 to 60 in 801ae8f

     	let client = match spec.api_type {
     	LlmApiType::Ollama => {
     	Box::new(ollama::Client::new(spec).await?) as Box<dyn LlmGenerationClient>
     	}
     	LlmApiType::OpenAi => {
     	Box::new(openai::Client::new(spec).await?) as Box<dyn LlmGenerationClient>
     	}
     	};

2. Add Enum to LlmApiType in Python]:

   cocoindex/python/cocoindex/llm.py

   Lines 4 to 7 in 0fbfd3f

   	class LlmApiType(Enum):
   	"""The type of LLM API to use."""
   	OPENAI = "OpenAi"
   	OLLAMA = "Ollama"

3. Test with the existing manuals_llm_extraction example. You can add a few lines similar to OpenAI:

   cocoindex/examples/manuals_llm_extraction/main.py

   Lines 93 to 95 in 555d328

   	# Replace by this spec below, to use OpenAI API model instead of ollama
   	# llm_spec=cocoindex.LlmSpec(
   	# api_type=cocoindex.LlmApiType.OPENAI, model="gpt-4o"),

4. Update documentation: https://cocoindex.io/docs/ai/llm

[par4m]

I can work on this, please assign!

[badmonster0]

SG! thanks a lot @par4m

[iamsb97]

Hi @par4m , have you started working on this?

[par4m]

Hi @par4m , have you started working on this?

Yep @iamsb97 , just doing some final testing before the PR.

[badmonster0]

Thanks a lot @par4m for getting this feature in!