Add a planner agent

src/main_agent.py

@@ -3,6 +3,9 @@
 import requests
 from openai_utils import get_code_from_open_ai
 from html_cleaner import get_cleaned_html
+from planner_agent.planner_agent import get_plan
+import queue
+import os
 
 
 def send_message(message):
@@ -20,11 +23,24 @@ def send_message(message):
 
 
 def main():
+    is_test_mode = os.getenv("TEST_MODE", False)
+
     history_messages = []
+
+    if is_test_mode:
+        goal = "Search a blue car"
+    else:
+        goal = input("Enter a goal for the web browser agent or exit to quit: ")
+
+    if goal.lower() == "exit":
+        return
+
+    actions = queue.Queue()
+    for action in get_plan(goal, test_mode=is_test_mode):
+        actions.put(action)
+
     while True:
-        action = input("Enter a web action or exit to quit: ")
-        if action.lower() == "exit":
-            break
+        action = actions.get()
         is_error = input("Is this an error? (y/n): ")
         is_error = is_error.lower() == "y"
         history_messages, message = get_code_from_open_ai(

src/openai_utils.py

@@ -1,6 +1,7 @@
 import os
 import openai
 from dotenv import load_dotenv, find_dotenv
+from typing import Literal
 
 _ = load_dotenv(find_dotenv())  # read local .env file
 
@@ -31,6 +32,16 @@
     "content": """{response}""",
 }
 
+TypeMessage = dict[Literal["role", "content"], str]
+
+def format_message(message: str, role:Literal["user", "system", "assistant"]="user") -> TypeMessage:
+    return {
+        "role": role,
+        "content": message
+    }
+
+def compile_messages(*messages: tuple[TypeMessage]) -> list[TypeMessage]: return list(messages)
+
 
 def replace_action_in_message(message, action):
     new_message = message.copy()

src/planner_agent/constants.py

@@ -0,0 +1,10 @@
+DEFAULT_RAW_PLAN = """To achieve the goal I will need to perform the following actions:
+1. Open the web browser.
+2. Navigate to a search engine or a car dealership website.
+3. Locate the search bar on the webpage.
+4. Enter the keyword "blue car" into the search bar.
+5. Submit the search query.
+6. Wait for the search results to load.
+7. Click on the most relevant search result webpage.
+8. Wait for the selected webpage to load.
+9. Close the web browser."""

src/planner_agent/planner_agent.py

@@ -0,0 +1,52 @@
+import os
+import sys
+import re
+
+os.chdir('./src')
+# setting path
+sys.path.append(os.getcwd())
+
+# importing project modules
+from openai_utils import compile_messages, format_message,  get_completion_from_messages
+from planner_agent.prompts import CREATE_PLAN
+from planner_agent.constants import DEFAULT_RAW_PLAN
+
+
+def get_raw_plan(goal: str) -> str:
+    create_plan_prompt = format_message(CREATE_PLAN.format(goal=goal))
+    context = compile_messages(create_plan_prompt)
+    plan = get_completion_from_messages(context)
+    return plan, context
+
+def parse_plan(plan: str) -> list[str]:
+    # Use regex to find all numbered items in the text
+    numbered_items = re.findall(r'\d+\.\s+(.+)', plan)
+
+    # Get rid of the first item, which is open the web browser
+    numbered_items = numbered_items[1:]
+
+    return [item.strip() for item in numbered_items]
+
+def get_plan(goal: str, test_mode = False) -> list[str]:
+    if test_mode:
+        raw_plan = DEFAULT_RAW_PLAN
+        print(raw_plan)
+    else:
+        raw_plan, context = get_raw_plan(goal)
+
+    # Get user feedback on the plan
+    while True and not test_mode:
+        print("Here is the plan:")
+        print(raw_plan)
+        print("Does this plan look good?")
+        user_feedback = input("Press enter to continue with the plan or provide some feedback: ")
+        if user_feedback == "":
+            break
+        else:
+            # Improve the plan based on the user feedback
+            context.append(format_message(raw_plan, role="assistant"))
+            context.append(format_message(user_feedback))
+            raw_plan = get_completion_from_messages(context)
+
+    actions = parse_plan(raw_plan)
+    return actions

src/planner_agent/prompts.py

@@ -0,0 +1,6 @@
+PLAN_FORMAT = """Give me your answer only in the following format:
+To achieve the goal I will need to perform the following actions:
+1. Open the web browser
+2. ... """
+
+CREATE_PLAN = """Given that you are an AI agent using selenium to navigate a web browser. What are the high level actions you would need to take to achieve the goal "{goal}"?""" + PLAN_FORMAT