Project: ai_engineering/agent_toolkit/
BaseLLM is the only contract the agent depends on; OpenAILLM and ScriptedLLM are the current implementations in agent/llm.py. Adding a third provider proves the dependency-injection seam is real and gives users a production alternative.
Scope: implement AnthropicLLM(BaseLLM) in agent/llm.py using Anthropic's tool-use API, mapping the OpenAI-style tool specs to Anthropic's format and the response back to LLMResponse. Keep CI network-free — cover it with ScriptedLLM-style tests, don't call the live API.
Good first issue for someone who's used the Anthropic API and wants a clean, well-bounded task.
Project:
ai_engineering/agent_toolkit/BaseLLMis the only contract the agent depends on;OpenAILLMandScriptedLLMare the current implementations inagent/llm.py. Adding a third provider proves the dependency-injection seam is real and gives users a production alternative.Scope: implement
AnthropicLLM(BaseLLM)inagent/llm.pyusing Anthropic's tool-use API, mapping the OpenAI-style tool specs to Anthropic's format and the response back toLLMResponse. Keep CI network-free — cover it withScriptedLLM-style tests, don't call the live API.Good first issue for someone who's used the Anthropic API and wants a clean, well-bounded task.