fix: prompt variable + langchain agent steps

[desaxce]

To test:

Start by creating a prompt template with name "LC Agent" from within Literal with the following messages:

• system You are a helpful assistant
• assistant {{chat_history}}
• user {{input}}
• assistant {{agent_scratchpad}}

Run the following:

# Set env vars:
# OPENAI_API_KEY
# TAVILY_API_KEY
# LITERAL_API_KEY
# LITERAL_API_URL

import os
import getpass

from literalai import LiteralClient

from langchain_openai import ChatOpenAI
from langchain_community.tools.tavily_search import TavilySearchResults

model = ChatOpenAI(model="gpt-4o")
search = TavilySearchResults(max_results=2)
tools = [search]

lai_client = LiteralClient(api_key=os.environ["LITERAL_API_KEY"], url=os.environ["LITERAL_API_URL"])
lai_prompt = lai_client.api.get_prompt(name="LC Agent")
prompt = lai_prompt.to_langchain_chat_prompt_template()


from langchain.agents import create_tool_calling_agent
from langchain.agents import AgentExecutor

agent = create_tool_calling_agent(model, tools, prompt)
agent_executor = AgentExecutor(agent=agent, tools=tools)

lai_client.reset_context()
cb = lai_client.langchain_callback()

from langchain_core.messages import AIMessage, HumanMessage
from langchain_core.runnables.config import RunnableConfig

agent_executor.invoke({
    "chat_history": [
            # You can specify the intermediary messages as tuples too.
            # ("human", "hi! my name is bob"),
            # ("ai", "Hello Bob! How can I assist you today?")
            HumanMessage(content="hi! my name is bob"),
            AIMessage(content="Hello Bob! How can I assist you today?"),
        ],
    "input": "whats the weather in sf?"
}, config=RunnableConfig(callbacks=[cb], run_name="Weather SF"))

[Dam-Buty]

LGTM