Python: service_thread_id missing in case of Anthropic

[sakhan-rio]

I created an agent using AnthropicClient class with anthropic client as AsyncAnthropicFoundry client and for thread creation I am using get_new_thread().

In case of Anthropic Foundry Agent, after first the thread object does not get assigned any thread_service_id and I see chat message store populated while in case of AzureAIAgentClient agent, after first run the thread_service_id gets assigned to the thread.

For resuming the conversation from the same thread, thread_service_id plays an important role and helps easily persist the conversation in external db.

Can you please suggest how can we resume thread in case of Anthropic Agents as no service_thread_id exist?

Attaching sample snippets for reference:

Agent using AzureAIAgentClient and thread.

Agent using AnthropicClient and thread:

[eavanvalkenburg]

Hi @sakhan-rio I understand the confusion, threads are a somewhat complex topic and we are working on redesigning them. The basic idea is that a thread can be one of two things: a reference to a thread that is stored on some service, or a ChatMessageStore that holds the messages. Some services, like OpenAI Responses and AzureAIClient, can do both, while others, like Anthropic and Ollama can only do local chat message stores. So in this case the thread ID is indeed not set for the anthropic runs, since we do not have a service there that can store it (AnthropicFoundry is the same API as regular anthropic just hosted through Foundry, it does not use Foundry Agents atm). So if you want to use a single thread with both, you would need to call AzureAIClient with store=False so that that is also a local thread and then you can pass that along, beware because some things might not work correctly then, for instance if you have a ContextProvider on one agent, but not the other, or a different chat message store, or some other scenario's, but if you use default chat message store then it should work.