agent-factory: generate agent auto-retrieve-files-from-a-reposito

app/nodes/core_node.py

@@ -1,33 +1,35 @@
+"""
+This module defines a LangGraph node function for retrieving files from a repository.
+The function interacts with an LLM to determine the workflow files available in the
+specified repository and scope.
+"""
+
 async def process(state: dict) -> dict:
-    """
-    This node retrieves files from a repository based on the provided URL and update scope.
-    It uses an LLM to analyze and extract relevant workflow files from the repository.
-
-    Input Schema:
-    - repository_url: The URL of the repository to retrieve files from.
-    - update_scope: The scope or context for the update (e.g., branch, directory).
-
-    Output Schema:
-    - workflow_files: A list of workflow file names retrieved from the repository.
-    """
     from app.agent import llm
     from langchain_core.messages import SystemMessage, HumanMessage
 
     try:
+        # Extract input fields from state
         repository_url = state.get("repository_url", "")
         update_scope = state.get("update_scope", "")
 
+        # Validate inputs
         if not repository_url or not update_scope:
             raise ValueError("Both 'repository_url' and 'update_scope' must be provided.")
 
+        # Prepare messages for the LLM
         messages = [
-            SystemMessage(content="You are an assistant that retrieves workflow files from a repository."),
+            SystemMessage(content="You are an expert in repository management and file retrieval."),
-            HumanMessage(content=f"Retrieve workflow files from the repository at {repository_url} within the scope '{update_scope}'.")
+            HumanMessage(content=f"Retrieve all workflow files from the repository at '{repository_url}' "
+                                 f"within the scope '{update_scope}'. Provide the list of file names.")
         ]
 
+        # Invoke the LLM
         response = await llm.ainvoke(messages)
-        workflow_files = response.content.splitlines()  # Assuming the response lists files line by line.
 
+        # Parse the response and return the output
+        workflow_files = response.content.splitlines()  # Assuming the LLM returns file names as newline-separated text
         return {"workflow_files": workflow_files, "phase": "complete"}
+
     except Exception as exc:
         return {"error": str(exc), "phase": "failed"}