[Feature Request] Add support to Anthropic LLM

[ocss884]

Required prerequisites

• [X] I have searched the Issue Tracker and Discussions that this hasn't already been reported. (+1 or comment there if it has.)
• [X] Consider asking first in a Discussion.

Motivation

The Claude series from Anthropic company is one of the most popular LLMs and they are serious competitors of GPTs. They support a 100,000 tokens context window while still free for personal API usage.

Solution

Add Claude-2 and Claude-instant-1 as backend models

Alternatives

No response

Additional context

No response

[lightaime]

Hi @ocss884. Sounds great. Please feel free to open a PR. Although I do not have access to Claude yet. Here is also a related discussion: https://github.com/camel-ai/camel/pull/271.

[ocss884]

@lightaime Hi, I opened a PR for Anthropic LLM backend. However, it looks like some tests fail for not being able to read the secret variable OPENAI_API_KEY. I check the action log and it is empty:

 env:
   pythonLocation: /opt/hostedtoolcache/Python/3.8.18/x64
  LD_LIBRARY_PATH: /opt/hostedtoolcache/Python/3.8.18/x64/lib
  OPENAI_API_KEY: 

I think it is because PR does not have access to secrets. See link. Could you help to add a "dump api_key" job in actions to fix it? Since we actually don't need a valid api key.

[krrishdholakia]

Hey @lightaime,

If you're integrating via litellm, here's an easy way to test if the anthropic integration is working: https://docs.litellm.ai/docs/proxy_api#step-2-test-a-new-llm

[ishaan-jaff]

Hi @lightaime @ocss884 I believe we can help with this issue. I’m the maintainer of LiteLLM https://github.com/BerriAI/litellm - we allow you to use any LLM as a drop in replacement for gpt-3.5-turbo.

You can use LiteLLM in the following ways:

With your own API KEY:

This calls the provider API directly

from litellm import completion
import os
## set ENV variables 
os.environ["OPENAI_API_KEY"] = "your-key" # 
os.environ["COHERE_API_KEY"] = "your-key" # 

messages = [{ "content": "Hello, how are you?","role": "user"}]

# openai call
response = completion(model="gpt-3.5-turbo", messages=messages)

# cohere call
response = completion(model="command-nightly", messages=messages)

Using the LiteLLM Proxy with a LiteLLM Key

this is great if you don’t have access to claude but want to use the open source LiteLLM proxy to access claude

from litellm import completion
import os

## set ENV variables 
os.environ["OPENAI_API_KEY"] = "sk-litellm-5b46387675a944d2" # [OPTIONAL] replace with your openai key
os.environ["COHERE_API_KEY"] = "sk-litellm-5b46387675a944d2" # [OPTIONAL] replace with your cohere key

messages = [{ "content": "Hello, how are you?","role": "user"}]

# openai call
response = completion(model="gpt-3.5-turbo", messages=messages)

# cohere call
response = completion(model="command-nightly", messages=messages)

[Obs01ete]

Please explain the way you want to add this Antropic/Claude. Will it be another model backend? Pleas also explain the difference between Anthropic and Claude. Pls do it right in the description of the ticket.

[ocss884]

Please explain the way you want to add this Antropic/Claude. Will it be another model backend? Pleas also explain the difference between Anthropic and Claude. Pls do it right in the description of the ticket.

Hi @Obs01ete , Claude series are LLM from a company called Anthropic, which support 100,000 token context windows. They could be another great model backend choice for role-playing agents. I have added more details in this issue

@lightaime Hi, I opened a PR for Anthropic LLM backend. However, it looks like some tests fail for not being able to read the secret variable OPENAI_API_KEY. I check the action log and it is empty:

 env:
   pythonLocation: /opt/hostedtoolcache/Python/3.8.18/x64
  LD_LIBRARY_PATH: /opt/hostedtoolcache/Python/3.8.18/x64/lib
  OPENAI_API_KEY: 

I think it is because PR does not have access to secrets. See link. Could you help to add a "dump api_key" job in actions to fix it? Since we actually don't need a valid api key.

Could you help to check the OPENAI_API_KEY setup? Due to the dangers inherent to automatic processing of PRs, GitHub’s standard pull_request workflow trigger by default prevents write permissions and secrets access to the target repository. I think all PRs cannot pass some of the tests due to the lack of OPENAI_API_KEY in the environment.