Structured Output Revisions

.gitignore

@@ -14,7 +14,7 @@ logs/*
 !logs/skip-morning_2024-05-15_13-54-44.txt
 
 ### Add simulations to keep here ###
-!environment/frontend_server/storage/base_the_ville_isabella_maria_klaus/*
+!environment/frontend_server/storage/base_the_ville_isabella_maria_klaus/
 !environment/frontend_server/storage/base_the_ville_n25/*
 !environment/frontend_server/storage/July1_the_ville_isabella_maria_klaus-step-3-*/*
 !environment/frontend_server/storage/skip-morning-s-14/

convo_parser.py

@@ -0,0 +1,62 @@
+import os
+import glob
+import os
+import sys
+import json
+
+def get_unique_conversations(simulation_name):
+    step_folder = "environment/frontend_server/storage"
+
+    # Use glob to find all files that start with the simulation_name
+    search_pattern = os.path.join(step_folder, f"{simulation_name}*/movement/*")
+    filepaths = glob.glob(search_pattern)
+
+    observed_conversations = set()
+    file_contents = []
+
+    # Iterate over all matching files
+    for filepath in filepaths:
+        with open(filepath, "r") as file:
+            data = json.load(file)
+            personas = data.get("persona", {})
+
+            # Loop over all personas except one, since conversations are always
+            # between two people
+            for name in list(personas.keys())[:-1]:
+                persona = personas[name]
+                chat = persona.get("chat")
+
+                if chat:
+                    chat_string = str(chat)
+
+                    # Add unique conversations
+                    if chat_string not in observed_conversations:
+                        observed_conversations.add(chat_string)
+                        file_contents.append(data)
+                        break
+
+    return file_contents
+
+def write_conversations_to_file(conversations, simulation_name):
+   output_directory = "logs/conversations"
+   if not os.path.exists(output_directory):
+        os.makedirs(output_directory)
+   output_filename = f"{simulation_name}_highlights.json"
+   full_path = os.path.join(output_directory, output_filename);
+   with open(full_path, "w") as file:
+        for conversation in conversations:
+            json.dump(conversation, file, indent=4)
+
+if __name__ == "__main__":
+    if len(sys.argv) < 2:
+        print("Please provide the simulation name as a command line argument.")
+        sys.exit(1)
+
+    simulation_name = sys.argv[1]
+    unique_conversations = get_unique_conversations(simulation_name)
+
+    if unique_conversations:
+        write_conversations_to_file(unique_conversations, simulation_name)
+        print(f"Unique conversations written to {simulation_name}_highlights.txt")
+    else:
+        print("No unique conversations found.")

environment/frontend_server/storage/base_the_ville_smol_elections_5_voters/personas/Klaus Mueller/bootstrap_memory/scratch.json

@@ -11,7 +11,7 @@
   "age": 20,
   "innate": "kind, inquisitive, passionate", 
   "learned": "Klaus Mueller is a student at Oak Hill College studying sociology. He is passionate about social justice and loves to explore different perspectives.", 
-  "currently": "Klaus Mueller is writing a research paper on the effects of gentrification in low-income communities. Klauss is thinking about whom to vote for in the upcoming town election. He knows that Adam Smith and Isabella Rodriguez are running for the office of town mayor. The election takes place next week.", 
+  "currently": "Klaus Mueller is writing a research paper on the effects of gentrification in low-income communities. Klaus is thinking about whom to vote for in the upcoming town election. He knows that Adam Smith and Isabella Rodriguez are running for the office of town mayor. The election takes place next week.", 
   "lifestyle": "Klaus Mueller goes to bed around 11pm, awakes up around 7am, eats dinner around 5pm.", 
   "living_area": "the Ville:Dorm for Oak Hill College:Klaus Mueller's room",
   "concept_forget": 100,

environment/frontend_server/storage/base_the_ville_smol_elections_5_voters_swapped_personalities/personas/Klaus Mueller/bootstrap_memory/scratch.json

@@ -11,7 +11,7 @@
   "age": 20,
   "innate": "kind, inquisitive, passionate", 
   "learned": "Klaus Mueller is a student at Oak Hill College studying sociology. He is passionate about social justice and loves to explore different perspectives.", 
-  "currently": "Klaus Mueller is writing a research paper on the effects of gentrification in low-income communities. Klauss is thinking about whom to vote for in the upcoming town election. He knows that Adam Smith and Isabella Rodriguez are running for the office of town mayor. The election takes place next week.", 
+  "currently": "Klaus Mueller is writing a research paper on the effects of gentrification in low-income communities. Klaus is thinking about whom to vote for in the upcoming town election. He knows that Adam Smith and Isabella Rodriguez are running for the office of town mayor. The election takes place next week.", 
   "lifestyle": "Klaus Mueller goes to bed around 11pm, awakes up around 7am, eats dinner around 5pm.", 
   "living_area": "the Ville:Dorm for Oak Hill College:Klaus Mueller's room",
   "concept_forget": 100,

nlp/openai_convo_summary.py

@@ -33,7 +33,7 @@ def main():
     print(file_chunks[0])
 
     response = client.chat.completions.create(
-        model="gpt-4",
+        model="gpt-4o-mini",
         messages=[
             {
                 "role": "system",

print_all_sim.py

@@ -0,0 +1,59 @@
+import os
+import json
+import re
+import sys
+
+def get_unique_conversations(simulation_name):
+    sim_folder = os.path.join("environment", "frontend_server", "storage", "base_search_and_rescue")
+
+    regex_name = re.compile(re.escape(simulation_name + '-'))
+    for file_name in os.listdir(sim_folder):
+        step=0
+        output = []  
+        if regex_name.search(file_name):
+            step_folder = os.path.join(sim_folder, file_name, "movement")
+            for filename in os.listdir(step_folder):
+                filepath = os.path.join(step_folder, filename)
+                output.append(f"Step {str(step)}:")
+                try:
+                    with open(filepath, "r") as file:
+                        data = json.load(file)
+                        for k, v in data.items():
+                            output.append(k)
+                            if k == 'persona':
+                                for key, value in v.items():
+                                    output.append(f'   {key}')
+                                    for attribute, val in value.items():
+                                        if attribute != 'chat' or (attribute == 'chat' and val is None):
+                                            output.append(f'      {attribute}: {val}')
+                                        else:
+                                            output.append(f'      {attribute}:')
+                                            for convo in val:
+                                                output.append(f'         {convo[0]}: {convo[1]}')
+                            else:
+                                for key, value in v.items():
+                                    output.append(f'   {key}: {value}')
+                            output.append('\n')
+                except json.JSONDecodeError:
+                    continue
+                except Exception as e:
+                    print(f"Error processing file {filename}: {e}")
+                    continue
+                step+=1
+
+            output_filename = os.path.join(sim_folder, file_name, f"output_0-{file_name.split('-')[5]}.txt", )
+            with open(output_filename, "w") as output_file:
+                output_file.write('\n'.join(output))
+
+
+if __name__ == "__main__":
+    if len(sys.argv) < 2:
+        print("Please provide the simulation name as a command line argument.")
+        sys.exit(1)
+
+    simulation_name = sys.argv[1]
+    '''
+    unique_conversations = get_unique_conversations(simulation_name)
+    print(json.dumps(unique_conversations, indent=2))
+    '''
+    get_unique_conversations(simulation_name)

reverie/backend_server/persona/cognitive_modules/converse.py

@@ -176,7 +176,7 @@ def agent_chat_v2(maze, init_persona, target_persona):
     focal_points = [f"{target_persona.scratch.name}"]
     retrieved = new_retrieve(init_persona, focal_points, 50) 
     relationship = generate_summarize_agent_relationship(init_persona, target_persona, retrieved)
-    print ("-------- relationshopadsjfhkalsdjf", relationship)
+    print ("-------- relationship", relationship)
     last_chat = ""
     for i in curr_chat[-4:]:
       last_chat += ": ".join(i) + "\n"

reverie/backend_server/persona/prompt_template/gpt_structure.py

@@ -188,7 +188,40 @@ def ChatGPT_request(prompt):
     traceback.print_exc()
     return "LLM ERROR"
 
+def ChatGPT_structured_request(prompt, response_format):
+  """
+  Given a prompt and a dictionary of GPT parameters, make a request to OpenAI
+  server and returns the response. 
+  ARGS:
+    prompt: a str prompt
+    gpt_parameter: a python dictionary with the keys indicating the names of  
+                   the parameter and the values indicating the parameter 
+                   values.   
+  RETURNS: 
+    a str of GPT-3's response. 
+  """
+  # temp_sleep()
+  print("--- ChatGPT_request() ---")
+  print("Prompt:", prompt)
 
+  try: 
+    completion = client.beta.chat.completions.parse(
+      model=openai_config["model"],
+      response_format=response_format,
+      messages=[{"role": "user", "content": prompt}]
+    )
+    content = completion.choices[0].message.content
+    print("Response content:", content)
+    cost_logger.update_cost(
+      completion, input_cost=openai_config["model-costs"]["input"], output_cost=openai_config["model-costs"]["output"]
+    )
+    return content
+
+  except Exception as e: 
+    print(f"Error: {e}")
+    traceback.print_exc()
+    return "LLM ERROR"
+
 # def GPT4_safe_generate_response(
 #   prompt,
 #   example_output,
@@ -234,7 +267,6 @@ def ChatGPT_request(prompt):
 
 #   return False
 
-
 def ChatGPT_safe_generate_response(
   prompt,
   example_output,
@@ -246,26 +278,51 @@ def ChatGPT_safe_generate_response(
   verbose=False,
 ):
   if func_validate and func_clean_up:
-    # prompt = 'GPT-3 Prompt:\n"""\n' + prompt + '\n"""\n'
-    prompt = '"""\n' + prompt + '\n"""\n'
-    prompt += (
-      f"Output the response to the prompt above in json. {special_instruction}\n"
-    )
-    prompt += "Example output json:\n"
-    prompt += '{"output": "' + str(example_output) + '"}'
+    # Constructing the new prompt using the structured output format
+    prompt_structure = {
+      "model": "gpt-4o-2024-08-06",
+      "messages": [
+        {
+          "role": "system",
+          "content": special_instruction
+        },
+        {
+          "role": "user",
+          "content": prompt
+        }
+      ],
+      "response_format": {
+        "type": "json_schema",
+        "json_schema": {
+          "name": "output_response",
+          "strict": True,
+          "schema": {
+            "type": "object",
+            "properties": {
+              "output": {
+                "type": "string"
+              }
+            },
+            "required": ["output"],
+            "additionalProperties": False
+          }
+        }
+      }
+    }
 
     if verbose:
-      print("LLM PROMPT")
-      print(prompt)
+      print("LLM PROMPT STRUCTURE")
+      print(json.dumps(prompt_structure, indent=2))
 
     for i in range(repeat):
       try:
-        chatgpt_response = ChatGPT_request(prompt)
+        chatgpt_response = ChatGPT_request(json.dumps(prompt_structure))
         if not chatgpt_response:
           raise Exception("No valid response from LLM.")
         curr_gpt_response = chatgpt_response.strip()
         end_index = curr_gpt_response.rfind("}") + 1
         curr_gpt_response = curr_gpt_response[:end_index]
+        print(curr_gpt_response)
         curr_gpt_response = json.loads(curr_gpt_response)["output"]
 
         if verbose:
@@ -491,8 +548,9 @@ def generate_structured_response(
           prompt=prompt
         ):
           return func_clean_up(curr_gpt_response, prompt=prompt)
-        print("Response validation failed.")
-      except:
+        print("Response validation failed.", func_validate(curr_gpt_response,
+          prompt=prompt), curr_gpt_response.decision)
+      except Exception as e:
         print("Could not process response.")
       if verbose:
         print("---- repeat count: ", i, curr_gpt_response)