Automating Tweet Interactions with Langchain

​

Objective

1. Retrieve the current user’s Twitter ID.
2. Extract a tweet ID from a given tweet URL or ID.
3. Like the specified tweet as the current user.
4. Confirm the action or report errors.

​

Steps

1. Set up the Twitter Application: Load the twitter application using app_from_slug and configure its integration (e.g., AgentRIntegration).
   Copy

   from universal_mcp.applications import app_from_slug
   from universal_mcp.integrations import AgentRIntegration
   
   def get_application(name: str):
       AppClass = app_from_slug(name)
       integration = AgentRIntegration(name) # Assumes 'twitter' integration on AgentR
       instance = AppClass(integration=integration)
       return instance
   
   twitter_app = get_application("twitter")
   
2. Initialize LLM and ToolManager:
   Copy

   import os
   from langchain_openai import ChatOpenAI
   from universal_mcp.tools import ToolManager
   from universal_mcp.tools.adapters import ToolFormat
   
   model_name = os.environ.get("OPEN_AI_MODEL", "gpt-4o-mini")
   llm = ChatOpenAI(model=model_name)
   tool_manager = ToolManager()
   
3. Register Specific Twitter Tools: Instead of registering all tools from the Twitter app, we’ll explicitly add the ones needed for this task. This demonstrates finer-grained tool selection.
   Copy

    Assuming twitter_app has methods like:
    users_id_like(self, id, tweet_id)
    users_id_unlike(self, id, tweet_id)
    find_my_user(self) -> returns {"data": {"id": "USER_ID", ...}}
   
   tool_manager.add_tool(twitter_app.users_id_like, name="like_tweet")
   tool_manager.add_tool(twitter_app.users_id_unlike, name="unlike_tweet") # Not used in this specific task
   tool_manager.add_tool(twitter_app.find_my_user, name="find_my_user_id")
   
   Self-note: The actual method names in twitter_app might differ; this is illustrative based on the like_tweet.py example’s tool naming.
4. Get Tools in Langchain Format:
   Copy

   langchain_tools = tool_manager.list_tools(format=ToolFormat.LANGCHAIN)
   
5. Create the Langchain Agent:
   Copy

   from langgraph.prebuilt import create_react_agent
   
   agent_executor = create_react_agent(
       llm,
       tools=langchain_tools,
       prompt="You are a helpful assistant that can use tools to help the user."
   )
   
6. Define the Agent’s Task Prompt: This prompt instructs the agent on the steps to like a tweet.
   Copy

   agent_task_prompt_template = """
   You are a helpful assistant. Your task is to like a specific tweet on behalf of the current user.
   
   Steps:
   1. Retrieve the current user's Twitter user ID using the 'find_my_user_id' tool.
   2. Given the tweet URL or tweet ID ({tweet_id_or_url}), extract the tweet ID if necessary.
   3. Use the 'like_tweet' tool (which requires the user ID and tweet ID) to like the tweet as the current user.
   4. Confirm to the user that the tweet has been liked, or report any errors.
   
   Only perform the like action; do not reply, retweet, or perform any other operation.
   """
   
7. Invoke the Agent: Get the tweet ID/URL from the user and run the agent.
   Copy

   async def main():
       # ... (LLM, ToolManager, App setup) ...
   
       tweet_id_or_url = input("Enter the tweet id / url: ")
   
       filled_prompt = agent_task_prompt_template.format(tweet_id_or_url=tweet_id_or_url)
   
       result = await agent_executor.ainvoke(
           input={"messages": [{"role": "user", "content": filled_prompt}]}
       )
       print(result["messages"][-1].content)
   
   import asyncio
   asyncio.run(main())
   

​

How It Works

• The agent first calls find_my_user_id to get the necessary user ID.
• The LLM extracts the tweet ID from the input tweet_id_or_url.
• It then calls the like_tweet tool, passing the retrieved user ID and the extracted tweet ID.
• The twitter_app.users_id_like method (aliased as like_tweet) interacts with the Twitter API.
• The agent reports the success or failure of the operation.