Add chat_llm implementation for GPT model

[arthurscchan]

This PR follows #800 by adding chat_llm implementation for GPT models.

[arthurscchan]

/gcbrun exp -n ascc-pr-807-1 -m vertex_ai_gemini-1-5 -i -b comparison

[maoyixie]

It seems that the implementation did not maintain a chat session. Its function is the same as ask_llm(). The original intention of the chat_llm() is to query the LLM in the given chat session and returns the response.

[maoyixie]

I will update this part at #814

[DonggeLiu]

Thank you both for working on this.
@arthurscchan could you please help us double-check if LLM maintained a chat session in your location experiments?
LLM agents can perform much better in this way.

@maoyixie Do you happen to have the link to the corresponding OpenAI doc in hand? We will appreciate it if you could share it here : )

[maoyixie]

Thank you both for working on this. @arthurscchan could you please help us double-check if LLM maintained a chat session in your location experiments? LLM agents can perform much better in this way.

@maoyixie Do you happen to have the link to the corresponding OpenAI doc in hand? We will appreciate it if you could share it here : )

According to OpenAI Developer Community, to maintain a chat session with OpenAI GPT, you need to pass the entire conversation history, including previous prompts and responses, in each subsequent API call. This allows the model to retain context and remember the flow of the conversation.
Relevant links:
1.https://community.openai.com/t/how-to-preserve-the-context-session-of-a-conversation-with-the-api/324986
2.https://community.openai.com/t/how-can-i-maintain-conversation-continuity-with-chat-gpt-api/574819

[DavidKorczynski]

@arthurscchan could you take a look at the above comments?