Feature/test summarization against memory challenge (#3567)

Co-authored-by: Toran Bruce Richards <toran.richards@gmail.com>
2026-01-08 14:53:53 -05:00 · 2023-04-30 01:56:57 -07:00
parent 064ac5c742
commit dd96d98fa1
12 changed files with 1847 additions and 2494 deletions
--- a/autogpt/agent/agent.py
+++ b/autogpt/agent/agent.py
@@ -56,6 +56,10 @@ class Agent:
        cfg = Config()
        self.ai_name = ai_name
        self.memory = memory
+        self.summary_memory = (
+            "I was created."  # Initial memory necessary to avoid hilucination
+        )
+        self.last_memory_index = 0
        self.full_message_history = full_message_history
        self.next_action_count = next_action_count
        self.command_registry = command_registry
--- a/autogpt/llm/chat.py
+++ b/autogpt/llm/chat.py
@@ -12,6 +12,10 @@ from autogpt.logs import logger
 from autogpt.memory_management.store_memory import (
    save_memory_trimmed_from_context_window,
 )
+from autogpt.memory_management.summary_memory import (
+    get_newly_trimmed_messages,
+    update_running_summary,
+)

 cfg = Config()

@@ -36,10 +40,10 @@ def generate_context(prompt, relevant_memory, full_message_history, model):
        create_chat_message(
            "system", f"The current time and date is {time.strftime('%c')}"
        ),
-        create_chat_message(
-            "system",
-            f"This reminds you of these events from your past:\n{relevant_memory}\n\n",
-        ),
+        # create_chat_message(
+        #     "system",
+        #     f"This reminds you of these events from your past:\n{relevant_memory}\n\n",
+        # ),
    ]

    # Add messages from the full message history until we reach the token limit
@@ -81,21 +85,21 @@ def chat_with_ai(
            """
            model = cfg.fast_llm_model  # TODO: Change model from hardcode to argument
            # Reserve 1000 tokens for the response
-
            logger.debug(f"Token limit: {token_limit}")
            send_token_limit = token_limit - 1000
-            if len(full_message_history) == 0:
-                relevant_memory = ""
-            else:
-                recent_history = full_message_history[-5:]
-                shuffle(recent_history)
-                relevant_memories = permanent_memory.get_relevant(
-                    str(recent_history), 5
-                )
-                if relevant_memories:
-                    shuffle(relevant_memories)
-                relevant_memory = str(relevant_memories)

+            # if len(full_message_history) == 0:
+            #     relevant_memory = ""
+            # else:
+            #     recent_history = full_message_history[-5:]
+            #     shuffle(recent_history)
+            #     relevant_memories = permanent_memory.get_relevant(
+            #         str(recent_history), 5
+            #     )
+            #     if relevant_memories:
+            #         shuffle(relevant_memories)
+            #     relevant_memory = str(relevant_memories)
+            relevant_memory = ""
            logger.debug(f"Memory Stats: {permanent_memory.get_stats()}")

            (
@@ -105,33 +109,36 @@ def chat_with_ai(
                current_context,
            ) = generate_context(prompt, relevant_memory, full_message_history, model)

-            while current_tokens_used > 2500:
-                # remove memories until we are under 2500 tokens
-                relevant_memory = relevant_memory[:-1]
-                (
-                    next_message_to_add_index,
-                    current_tokens_used,
-                    insertion_index,
-                    current_context,
-                ) = generate_context(
-                    prompt, relevant_memory, full_message_history, model
-                )
+            # while current_tokens_used > 2500:
+            #     # remove memories until we are under 2500 tokens
+            #     relevant_memory = relevant_memory[:-1]
+            #     (
+            #         next_message_to_add_index,
+            #         current_tokens_used,
+            #         insertion_index,
+            #         current_context,
+            #     ) = generate_context(
+            #         prompt, relevant_memory, full_message_history, model
+            #     )

            current_tokens_used += count_message_tokens(
                [create_chat_message("user", user_input)], model
            )  # Account for user input (appended later)

+            current_tokens_used += 500  # Account for memory (appended later) TODO: The final memory may be less than 500 tokens
+
+            # Add Messages until the token limit is reached or there are no more messages to add.
            while next_message_to_add_index >= 0:
                # print (f"CURRENT TOKENS USED: {current_tokens_used}")
                message_to_add = full_message_history[next_message_to_add_index]

                tokens_to_add = count_message_tokens([message_to_add], model)
                if current_tokens_used + tokens_to_add > send_token_limit:
-                    save_memory_trimmed_from_context_window(
-                        full_message_history,
-                        next_message_to_add_index,
-                        permanent_memory,
-                    )
+                    # save_memory_trimmed_from_context_window(
+                    #     full_message_history,
+                    #     next_message_to_add_index,
+                    #     permanent_memory,
+                    # )
                    break

                # Add the most recent message to the start of the current context,
@@ -146,6 +153,22 @@ def chat_with_ai(
                # Move to the next most recent message in the full message history
                next_message_to_add_index -= 1

+            # Insert Memories
+            if len(full_message_history) > 0:
+                (
+                    newly_trimmed_messages,
+                    agent.last_memory_index,
+                ) = get_newly_trimmed_messages(
+                    full_message_history=full_message_history,
+                    current_context=current_context,
+                    last_memory_index=agent.last_memory_index,
+                )
+                agent.summary_memory = update_running_summary(
+                    current_memory=agent.summary_memory,
+                    new_events=newly_trimmed_messages,
+                )
+                current_context.insert(insertion_index, agent.summary_memory)
+
            api_manager = ApiManager()
            # inform the AI about its remaining budget (if it has one)
            if api_manager.get_total_budget() > 0.0:
--- a/autogpt/memory_management/summary_memory.py
+++ b/autogpt/memory_management/summary_memory.py
@@ -0,0 +1,112 @@
+import json
+from typing import Dict, List, Tuple
+
+from autogpt.config import Config
+from autogpt.llm.llm_utils import create_chat_completion
+
+cfg = Config()
+
+
+def get_newly_trimmed_messages(
+    full_message_history: List[Dict[str, str]],
+    current_context: List[Dict[str, str]],
+    last_memory_index: int,
+) -> Tuple[List[Dict[str, str]], int]:
+    """
+    This function returns a list of dictionaries contained in full_message_history
+    with an index higher than prev_index that are absent from current_context.
+
+    Args:
+        full_message_history (list): A list of dictionaries representing the full message history.
+        current_context (list): A list of dictionaries representing the current context.
+        last_memory_index (int): An integer representing the previous index.
+
+    Returns:
+        list: A list of dictionaries that are in full_message_history with an index higher than last_memory_index and absent from current_context.
+        int: The new index value for use in the next loop.
+    """
+    # Select messages in full_message_history with an index higher than last_memory_index
+    new_messages = [
+        msg for i, msg in enumerate(full_message_history) if i > last_memory_index
+    ]
+
+    # Remove messages that are already present in current_context
+    new_messages_not_in_context = [
+        msg for msg in new_messages if msg not in current_context
+    ]
+
+    # Find the index of the last message processed
+    new_index = last_memory_index
+    if new_messages_not_in_context:
+        last_message = new_messages_not_in_context[-1]
+        new_index = full_message_history.index(last_message)
+
+    return new_messages_not_in_context, new_index
+
+
+def update_running_summary(current_memory: str, new_events: List[Dict]) -> str:
+    """
+    This function takes a list of dictionaries representing new events and combines them with the current summary,
+    focusing on key and potentially important information to remember. The updated summary is returned in a message
+    formatted in the 1st person past tense.
+
+    Args:
+        new_events (List[Dict]): A list of dictionaries containing the latest events to be added to the summary.
+
+    Returns:
+        str: A message containing the updated summary of actions, formatted in the 1st person past tense.
+
+    Example:
+        new_events = [{"event": "entered the kitchen."}, {"event": "found a scrawled note with the number 7"}]
+        update_running_summary(new_events)
+        # Returns: "This reminds you of these events from your past: \nI entered the kitchen and found a scrawled note saying 7."
+    """
+    # Replace "assistant" with "you". This produces much better first person past tense results.
+    for event in new_events:
+        if event["role"].lower() == "assistant":
+            event["role"] = "you"
+            # Remove "thoughts" dictionary from "content"
+            content_dict = json.loads(event["content"])
+            if "thoughts" in content_dict:
+                del content_dict["thoughts"]
+            event["content"] = json.dumps(content_dict)
+        elif event["role"].lower() == "system":
+            event["role"] = "your computer"
+        # Delete all user messages
+        elif event["role"] == "user":
+            new_events.remove(event)
+
+    # This can happen at any point during execturion, not just the beginning
+    if len(new_events) == 0:
+        new_events = "Nothing new happened."
+
+    prompt = f'''Your task is to create a concise running summary of actions and information results in the provided text, focusing on key and potentially important information to remember.
+
+You will receive the current summary and the your latest actions. Combine them, adding relevant key information from the latest development in 1st person past tense and keeping the summary concise.
+
+Summary So Far:
+"""
+{current_memory}
+"""
+
+Latest Development:
+"""
+{new_events}
+"""
+'''
+
+    messages = [
+        {
+            "role": "user",
+            "content": prompt,
+        }
+    ]
+
+    current_memory = create_chat_completion(messages, cfg.fast_llm_model)
+
+    message_to_return = {
+        "role": "system",
+        "content": f"This reminds you of these events from your past: \n{current_memory}",
+    }
+
+    return message_to_return
--- a/tests/integration/cassettes/test_memory_management/test_save_memory_trimmed_from_context_window.yaml
+++ b/tests/integration/cassettes/test_memory_management/test_save_memory_trimmed_from_context_window.yaml
@@ -332,4 +332,174 @@ interactions:
    status:
      code: 200
      message: OK
+- request:
+    body: '{"input": ["Assistant Reply: {     \"thoughts\": {         \"text\": \"thoughts\",         \"reasoning\":
+      \"reasoning\",         \"plan\": \"plan\",         \"criticism\": \"criticism\",         \"speak\":
+      \"speak\"     },     \"command\": {         \"name\": \"google\",         \"args\":
+      {             \"query\": \"google_query\"         }     } } Result: None Human
+      Feedback:Command Result: Important Information."], "model": "text-embedding-ada-002",
+      "encoding_format": "base64"}'
+    headers:
+      Accept:
+      - '*/*'
+      Accept-Encoding:
+      - gzip, deflate
+      Connection:
+      - keep-alive
+      Content-Length:
+      - '483'
+      Content-Type:
+      - application/json
+    method: POST
+    uri: https://api.openai.com/v1/embeddings
+  response:
+    body:
+      string: !!binary |
+        H4sIAAAAAAAAA1R6WxOyurbl+/4Vq9YrvUtEJWG9IfeLJihesKurCxC5iSCQBHLq/Pcu/E6d7n7x
+        AfFCMuaYY4yZ//jXX3/93SZVlo5///PX3+9yGP/+H8u1ZzzGf//z1//8119//fXXf/xe/787sybJ
+        ns/yk/9u/71Zfp7Z9Pc/f4n/feX/3vTPX3/fopdD0flZVNPUPC8yT+sT3gt8pY/uh6XKIzl8CTzP
+        74Rhggh4m3hNHTU6VhN7kC18XyqDotgb+zmfIhk+DKen3kq1uXg3HwJceS8R25JtA15/gkgZ7UnA
+        TptGFXfomsGpRFfq7XDr94G9VxVzjTbUY1RLlu+XQXJ53LE261NCuyg4w6QiDZHiXKrKTRMg+KJl
+        RuBWrkPyiQQGU2j55ASHhnM1HnIomOyK/UP46VlvDDGM50TFfqDyiqUNOyuv9Lahhye6hzPnWQTX
+        0DngC2N2zxHqDMiu8I6RNGxBvy4uKRxflxRn7lr0yfd0G6D9kXSMwuTT8yG7s13IOgHbO5FzPiY1
+        gUPXrSiGmtF3X3RSYWolBLEi6qohbVigrDwhJErOdX924KlWtIgcqepU34SIL3sLt1W6RYA/hpBd
+        XlYN7qJ2JXxb0rAtJOBAr1hv0EbURZ88Qa3CO68k7Hjw40+nMUeKEphbspncgg9QLSM4xqs3Vldq
+        2Q+N4auwuBY11W3fqDYwfjLZQOOZ7vX+0k8nZT9AZH0YdYzjPpl1ZEiwdpQNUjYbLZwf5gZBr0EE
+        yXfb1ScRHBFEb4QQUJTGZ6etZ0E8ZQeq7nDJSSh3AVTF3UR1cQQ9f/Rghrt9s8HqIbgl3dGzIazi
+        zZaqvQjC6fM+11AAnUH9G90l8yRLGdRbsSJSo0nhDx874yTL2DtEVcjelmDJT1utcDaLec/ZfmvA
+        q9po1EykIZnDzhRhfMgDnJovE4xm6NTQEsIj+azevJpv+HSAard6oM333erTls0GWGnUwK7DN9W4
+        G0wLNifhjI3tM+x5FRdQcTV5Qq0ajRV7R48cesn5RtYAv8NJj9sWyhZbY0MqPsm8CzADv/U/lPtJ
+        58q9HGQeHWO0RpHaSyvCNeW9eiUYqXcnmd9lxIAZrwKkMA2B6f18OHA9SAoBtI56LhFPlsVTOC37
+        U4SDaHcGkG+WgS1SQp9kzrlUhq5fYZWlnc+d/BLAfZhjanrVwOe5oRA+SHTCqLmvknnclNqqt8SC
+        PrEE+Dhf1Bk2gD6wE+GGz1BKLEj6JsbHfez5bGUypnjl7Ua9ZwX0Sex7AhJpeyTysy8BewRWBpUU
+        vwhJd2FPuDHJ8Ni+Juyvdmd/krtYkyV7ulJNqO468UPYwMo4R3SfayEgD3vnycA/a9R7H7Vqs5dg
+        Bm/9l1L9/IbVpE4lgsVoaTjQFKlqjuMjg7f0nqMdqu/JFL/cFFTxekujyH75v/oGIJ4q6th2mHDz
+        +4kg/OgFKRa804MxEvg9zh49qmmt840Xy2B0zk9qW49rwkmbOFAJ6js2d/6pkqxSJNAs9YAoJ6r3
+        kzp9EVhfZwGrRdT1k51oGiw0KaWOcDcr8aoEKVzwgrVCf3N2TcNSMYm3wQi1Rz6JwERwT9071a/p
+        N6GMpY2MOqPBL1xs9bFwijOM5Rbj+4m1nKNMa2C9P9rk/TG3Pnt9DwN0b1ZEfXx89Pw5PD14LjSb
+        yDa79mx3r6Bsf0Qd8chnCZ3dqYSZc3FJMSeFzp6lHgAxkjC1nncjHGCgy5CBtUXtjXnkE1lFIjSb
+        yMbJBuX6JB7aARhXfiav/vjWp1XHY8ir9kQtfPLAdNzwmyy9wQXJFLbJYH0DWV7qjXqm2PXvfDgF
+        UIZ9jTjTuoocTm4OdkOFyMZ4dsngzucblDo+4MN12FSD7YYQipvvih78wOxJJq1UmCm5TdXhc03m
+        /KWk8NrnD2qiSK1Y9Vx9wRoRAXt3u/PnPAgh9ByLUkshVTIJxz6TrbI1sakIfcIOm5xAaY0qatw8
+        J2HnS2wASyqf1Oi9kPPrVhXhFHfSgvfE5/k1rJXUehBqrjOz38jn9Ca3uxWhe20kFWvRqoRSLECs
+        CyHWqcwfZ+iO3p2sG/us98q9JPC7YgVF/u7DZ8N/zMBxpZii4/fDl342w+YxaljXbwFgZeh7snLJ
+        KMY38wQ2xf5ogU14KxEzh7QfLRqVcH1lAjZipfQndJPiHx4w/kSnkHu7yVHgtNphu2GXZFY/6gVe
+        lCimzjXPOVvpkgh2R5LRvQwswPZP14Af/J6wrl7uOnNuWgzkx0vBfgfUipw0N4LwYLvYzldiz/N6
+        bUH3ARn1TXoE8+68J4pyjCk2Nd32e+IGDsw3jzeNEncLSK7OKRDM+UpxbyrJV+G5BovhXWEru36T
+        WX1bMTQPTo5NfT/qfE4tA0p3o6aHVef5ZK8HN3jt9zLe36ioT+uWfeGzEABGluPobGKrA6wvMcH7
+        9GqD4RX3B3BzvxE+HmXK5yGJLVlzYwEfsVD1VC4UCU6NMWLrLEh85CwgCtiCErvtgVWt8662CkPy
+        HW220hGw2yNP4b19Z+RtstDnJWEZHJPMoOb82XDii1MOtnPYkt0pVMJB6FoGX5K9wobmlLwvj6AE
+        76AdcLgtcULWaSrCQGq93//3OV3vG8iFuqX7jvOeDF/ewMIaRGp06APGBDxi+MkOGGuf2e8JausU
+        quQWYLyJJT6sdEEEXzMH+IilBLBG7FKoSdIB47TzK/4ZX7J8bJ8T3UvOUZ/XIzsr95paGEXWA/Dp
+        UNfwZGYyVY8oDbnnngcZefCD1dqmVafvqhaEN2Ri119PITkmkgZHVUfY2vf3ZNA+6xo4JtTwq6qU
+        P3gBjzL40iDuqorI2daD/WnekTk7RDrHj2sDz+28IcL+SkJO7NsBiunwodqqqwFd9AJY+IKsiHzS
+        6VtcCZAe64nMq67m/VWytsr3yDycLHphoNtXAOyAGdT8mJE+m1Y9K5aRW/R4j5Jq4UNBeU6Bt/QD
+        Ho5XJrYQN/qVwPbtcdFD3xvklTuTebN6huyqRCmkFVSp8RxafT6dShk8DK/HnpRrfp9LQQmOg9hS
+        zXicQbeK7hDGssvQSrXVZD3Z8Q1UzZliJznlPW9ldIH6cevTxM+PlfTZgQFe4l1NUZ2EIa3iQgC9
+        9vwifo9uIV+d1wN0+LukjisUPnXy9AzyTfJGz12rhRsv+Zwh/5QddS3jVJEvbRtwFY4Fja/6OpkN
+        /8RgkcOEquWN9ryYqhKa9tSRKcovPquemy/YyuWGuhp+Av7a6Q0kJj5QdKhjn0MQE3AKkvOfz7M5
+        K1vF3mdrJFbtxR/vMtOU7bFxkLD4iW9OMwi6NKsRsPsdmL5wXYI6KG/YfzQfn83K+6vUczki/s4o
+        6OazJcK7NlBq4Ab1FMZXBndPvCK7avAq3s/NGQYX94JE1DoVHQ86gnorVWitKrtw8gwnBZMInngv
+        tGdAUeY10E6TA2EGkXpuRbEE1fLQIUF/lDqtvnomE9euETDpkfO19kiBgI4SNaIdCZlyDCQ4Cv6B
+        VGveJQufDUr9MSVq80DW5wM75NB9v3dk/epVf1P5TgYPfaVR1RxgNY1zGUH3jFTyDty8mstdfIEn
+        93omKVKmnt22fQZKH+3JdtFXvVa+BTB8nYyej6smIQLeavBPP86uXjgDr/hC13dU+uPb8Zq1FsRM
+        GSl6rorkexxPKWisKqF+EYZ6W9HEgJ8zf2C0PC/fWzcV5quviPXvye03fcAOskbnA6psHQL+rQz2
+        01dYj7Ghc7ZnhrL0M7wvnKKnWfc1YBz5D4zI3uK8liYZuhZ3yXolbxKiWnoEvcm3iLTxumoQdt4Z
+        yKh4Yzu9lH1XWp8bXNYXY30/h81aOTEQnqsV9fy3m7CWwgyGxRrhoxYFPqnfvgiFDRvoLclg9X1d
+        VajYwWwgmB2UhFy1coaege5k3uzvCdOvnQrXloCwxa9rPudTICvZ95KjMeiuYMr0eAupOzywfau0
+        iknnaw1HmwuLf+l0XpJtKluqi7HabWky3Qa73Sr0YVGNXvqEi+tSgEUtcuq/HQu0YeGUMKmGhh4K
+        3dene3epYXnzHti5ZRYQv+4pVqbGGolSwHVS39csUqrkcscnMxWr+Tg+Ujjf43B5XqYv+BTBhMeM
+        rJ3XQ59zeoNQf1QXtE1Oas/w6uvB8+PVEvDMViEr3i8ChgJgqn7cLuS1laSy3PID9jZZ69Ov4c7A
+        V7MrPSq5XI07QBoonk4T9fvNoZrCyzAAKNg5PeZc1zfppDkQIWmF1lc5D6crg1+QEbTFqBiO4eB/
+        5zO8CoOK789zW01FpLGfHqUvTR9ClkonBuWnJJLVTGvA3NMzgsAPNDQnZ5NvhlH1lMva9ala27if
+        b8U4w5dwRxhd0L6X/HlVA/O9a7G+y9fJbOf7r3w9AICNXfKouAtrBMPYelJPUxJAgEMH8A6+w8/v
+        6iTqOwd2VVkTZR9//blB3hl0fXghEzJKfURrJ4DgBnwiJsDRx/M2HaBFFIEiDxSgrZNDA/eVzjC+
+        2j5Y+9ktg6e03KGJ3ceEX5sxAj2fAuqxj9yzraeUYNH/2L9OPBykYmWAvhZybNpiwydDlFoIJmem
+        etFBfXxePxl8SeYK+9tV7lN19ylhaW08/OO7yQsLAQqrIqWaH+Qhf+ttJL+PT4NmdcLDrjWuLZS3
+        VoL3F+74U1R/Gnh+Si/qnkIlIdnnWcLr3DywRm5myNMi/8Ky4CX2x/ZQzcXQbyEP0i8+2n7Haez4
+        M4Ty0aHGuioSlsR7SdH2X0Td3SvQKUKFofz46YteST9rdD8r7vuzI1BvMp+l0mOG2zTe0r0VxxWr
+        /d6Ac7MTqd19Sp+VetEqb34XqRraFZ+f/rGBw9qQ8D7UHH/hx/inf5d6aQBrg10GR8E9/MmP6PMY
+        e5BI+Uz44vfyzctQ4Qd/JrLkJ/1M2R4CdR0Fv/WrJrR2znCpL6qNwaYf7N0cw0uZ6Wi3ewX+FJyC
+        M2wQ+yJZUxI+tbOrwsW/UJfv0pD89OAFKweyWfKb2V4ZX7Dw4+In8op9wroEBtycSOevp4QDUgfw
+        tK9LqhaRWzEY+FvI0PZOnyda9VMVPmdwRlWAJNq6YP20TAFKxKko3ptW8k2FlwX53k/QuoNvTsHL
+        Y3CXJx8EorbwGWbqF4p7FuDjwh+0e68kqCZOR7PFH9DKRTl8m+RC91apJZLjAwdepWzCtv7FiYhX
+        XwfGezPG1Pk0/vAgwUVxt1OOzZ9eDbMKwuzu7LAty/uKH9aGB8LCe1L19XH7GblFCpf+SJ01jXzy
+        MFcHeHkFCdmF98mfrmqG/uh999D6yXrlbSOwZtobbR/HAYys125Qr6GPo6P0SHiIvQyGt4NJ1be5
+        rsgNPw5gdZV7ii2p8VnQJy1Q3PJOVoPw9kdlSjLgPIcZ+3TIQ17sTQNOZ2CRpf/3w7e9xPB2LBlp
+        9fiTzAA4Bvwo6mrJD8uEb+/uDdqCrJFNFqCKb7qrDFfaaGBLp76+IRuxVcJzsSJzzU5g0eslgNKF
+        Ud3WU8DKcmjBJXTEBY8fzk5t3gBzZ2P882stuxWHH38TLo5JP2fSSgMfBiIEPe8O6OG298BJmnMC
+        +09djY8ClnCvXlp6cdeIr6P6U//05sLXjr7x3DOB7pbnJFj0Qz8pvgZ+eaDdFmLCrY94UWjAHtiL
+        tq4+tcazBbt6zqjx2FySce2RGTBtZtRY+I1rM07Bc6fJSIi9Y8Wcdy/DpX6R/D5qPTtfzobyW8/5
+        TTacdclbgB7evAgrb7iiJWEpzMhhS8JtSZORZKEMxWcSoQ/Y2n2VFnkLzVgJ6GEKIKCwW523Gyf2
+        fn7Cn9/B+gYvShxjx9Vf/XSXmQqgmu2xO6y+Cdsn2IPkgyuqM83tZydJ/2v/+cOVwkV/WErIeoFc
+        jXn02TYNINxevg6CU1RXk7wFOXgFOCDi4qemqClyeAR9hC3adqDwwk6A+BsfsHk2h55stgMDzenF
+        MarY3herIDDgq28tmsnFoVry7QHUqZtR20OdPifqyVA0ehoJS98YTMP6mQGLrAR0nboTEPW3V+6e
+        O1XG6g5rnMBm10Dn+qqoeTYPPVj6Mdwk5hmbxjzqTUKUGrKrcEfwzvpkxNpHhhGCA6lNbe0Pevgx
+        4LGFAdWliSZTs/l4MBqnlCIlOFd9/EEzIJcyIehdFPq8P+9EyE6fHCORDWDG710E05P4xv5bfenf
+        X5658DkR7Z1WLev7BdZF/1DD2bSAd5vkDIon41hzD0PFSTS1ytKPiOhsWj5/ImkGwcW/0IPeCDqf
+        xiJXyt1Fxd7FxOG08li8e1XRhp7bpwLI5TYFUEubN9aDzEho+bxqsO6zkShouwaMNRqBi/6nP7wv
+        eaMKNcoOCBQh19n67jVQSY8vNITgXv34E17lq4s976Mns7nREfzVozU0gS9mQnIAPz1gy/ODj5E5
+        NPIzs0oyPZqCT6SwIOi3no3y/Aw4rT9BrAD7Df/U2+K/blsihCWiggs5D7NegCczlbHpvhveCccq
+        hc5GeGC8+DeedaUFsyZzyIL/amMIyhZsmtz95U096y/m9sc32MfXIpzNmxXALmoOhOAm5Fy8EA1Y
+        nyNa8Ff47PZoU6Aaxhcb4LLXf35febxsjYjNa883lYtKeAtIROZKufNf3gTAbefjfbPbJ5K9OrRQ
+        aJ9P6q92s88jQkR5yQvw8R6BaoLnIIP4EByxrRQvPgHzq0Hqkgda3/W5X28upacseTJ2dvViR+s4
+        Anh9qbEZiXNPkXKsf/MOsi15Hs6K7jagTTyBVBAn4ThqPpRHIzKJuORl618e4e2iKwFLvstPpm1A
+        4po1mvevvGdsPRx+9T72XrwCVMqKCJYjcKn2RLM+7tK+hukTHn75lT7fM5lAK6lNjM+RCUTMnC8M
+        +mOI1den63nQHLRfP8JPciI//MdgiDcYrQwlq9jZ2DZQ2Z8cNP346LtXEOzcs4I9MSXhSFaBBC5f
+        xKl/2+x1vpGnG7gGZkXNtLNC8XZv8p+/wE5kr/xh28sZlAstxX7t5aD9uo8Ymm/QUj22YciNobvJ
+        y/NgVZEvYH7nmQRHIzaxdcAnf7qcpi/45W+4jP2EqS/owB24CtQ+grXfit3dAc6o7ukyvwIUf3ca
+        XPQD1cV1Fc7+O9LgE/GQ7Epi9O/Uk6FsPLov2XkbPxlfwijJhgVLmmrt1v+jhy93/UAWfusnn+5k
+        uBeyBu/v6z1YL3iQf/m0q2EF8LRov9A5fXw0XMip3wj65QB3/IPQz9/x80nJ5R9fRsFo6zzEWgbt
+        z70m4rJf7G1JFqw/toRdcbglE0FyDJEnfOiiRxPuufEA3+CjU5X0nc5uzlEFnJ4e2JjI2LPEIRE0
+        toGz5CmlTw77XFV0UPXUaF4Fnwxp28C2sW8Ur7Mvnw33LsFfvnPrNl99qsTzAOzPtcY6vFf6/KrL
+        VFnyanr81hqXztnppqDMQfSy9Asaf6wZtmGE/sz/eHx/n5W2RALWssNWn9nhAQFODAUJs5hXjN0K
+        pHi7+EqPxiVNxtXuIsMXzTMkZus4XPqTAIVnZaJ5MvtwEgaSysVoaPiw8P3PH8BcEHTsxP4ZzOte
+        uoFJcQZqvB/in3xFLnc3lTqfTqhm81MMP/7A2Pk0+h9/u/Q/spXGTG8Pm/bPfA6j74H7LdCfGSw1
+        YUYf67EOJ+MKWji/fJ2i5v5KFjyKwK/8J5mcsEx++wUWvqUH+GHh3IVRC08SywkQwMFn1WtqYNKe
+        KD7KTRvyylfTP/MziyVLPOCPX+iIxp5Gy/xrWPIjeNnlGtUuyTYca/oeYHTTBvQVtAIsec0ByJr8
+        xD+8MamaNFCez3uKHodPz9GBer/8Eu1OyVuf78qaAHHTrvAprm1Oln6m7JPzCmOAvWpCdirDjb3r
+        qbbot0XfB7DP3x3VJrNPFr5BP71O/Vp76AR09Re2hxvCep13YB5cqVWM2HOxpykADPpby8GSj5Lx
+        KFPQGaLQ/vILROvbvR9Xu1SG7/39iarFn/LAdlWIDU2gP30+uRYo5fal38gUqCScwNUv5V0gW9Sp
+        6zxc+t8AHFPQsJnf5nAUSB4rzOoc6l68Xh/HSpWhDARO7cndg01EGhGyUbthLQeZz1dq58GVSBOy
+        zFNC+tuPxf+h0yOTdfYaNjNcdcYZe1tXDtn+ubeUhhknetklu/4rnvILNA9eTo3mLvvvG9k3yt+/
+        UwH/+a+//vpfvxMGTfvM3svBgDGbxn//91GBf8fP+N+iKP2bSn9OIpAhzrO///mvQwh/d33bdOP/
+        Hts6+wx///OXIv85bvD32I7x+/+9/q/l1/7zX/8HAAD//wMAPe/DkOMgAAA=
+    headers:
+      CF-Cache-Status:
+      - DYNAMIC
+      CF-RAY:
+      - 7bfdedf56918cfbc-SJC
+      Connection:
+      - keep-alive
+      Content-Encoding:
+      - gzip
+      Content-Type:
+      - application/json
+      Date:
+      - Sun, 30 Apr 2023 07:04:15 GMT
+      Server:
+      - cloudflare
+      access-control-allow-origin:
+      - '*'
+      alt-svc:
+      - h3=":443"; ma=86400, h3-29=":443"; ma=86400
+      openai-organization:
+      - user-adtx4fhfg1qsiyzdoaxciooj
+      openai-processing-ms:
+      - '25'
+      openai-version:
+      - '2020-10-01'
+      strict-transport-security:
+      - max-age=15724800; includeSubDomains
+      x-ratelimit-limit-requests:
+      - '3000'
+      x-ratelimit-remaining-requests:
+      - '2999'
+      x-ratelimit-reset-requests:
+      - 20ms
+      x-request-id:
+      - f2850337441001a8e5bd70bacb4dede8
+    status:
+      code: 200
+      message: OK
 version: 1
--- a/tests/integration/challenges/memory/cassettes/test_memory_challenge_a/test_memory_challenge_a.yaml
+++ b/tests/integration/challenges/memory/cassettes/test_memory_challenge_a/test_memory_challenge_a.yaml
--- a/tests/integration/challenges/memory/cassettes/test_memory_challenge_b/test_memory_challenge_b.yaml
+++ b/tests/integration/challenges/memory/cassettes/test_memory_challenge_b/test_memory_challenge_b.yaml
--- a/tests/integration/challenges/memory/test_memory_challenge_a.py
+++ b/tests/integration/challenges/memory/test_memory_challenge_a.py
@@ -6,8 +6,8 @@ from tests.integration.agent_utils import run_interaction_loop
 from tests.integration.challenges.utils import get_level_to_run
 from tests.utils import requires_api_key

-LEVEL_CURRENTLY_BEATEN = 3
-MAX_LEVEL = 5
+LEVEL_CURRENTLY_BEATEN = 3  # real level beaten 30 and maybe more, but we can't record it, the cassette is too big
+MAX_LEVEL = 3


@pytest.mark.vcr
@@ -30,7 +30,7 @@ def test_memory_challenge_a(
    create_instructions_files(memory_management_agent, num_files, task_id)

    try:
-        run_interaction_loop(memory_management_agent, 40)
+        run_interaction_loop(memory_management_agent, 180)
    # catch system exit exceptions
    except SystemExit:
        file_path = str(memory_management_agent.workspace.get_path("output.txt"))
--- a/tests/integration/challenges/memory/test_memory_challenge_b.py
+++ b/tests/integration/challenges/memory/test_memory_challenge_b.py
@@ -6,7 +6,7 @@ from tests.integration.agent_utils import run_interaction_loop
 from tests.integration.challenges.utils import generate_noise, get_level_to_run
 from tests.utils import requires_api_key

-LEVEL_CURRENTLY_BEATEN = 1
+LEVEL_CURRENTLY_BEATEN = 2
 MAX_LEVEL = 5
 NOISE = 1000

--- a/tests/integration/goal_oriented/cassettes/test_browse_website/test_browse_website.yaml
+++ b/tests/integration/goal_oriented/cassettes/test_browse_website/test_browse_website.yaml
--- a/tests/integration/goal_oriented/cassettes/test_write_file/test_write_file.yaml
+++ b/tests/integration/goal_oriented/cassettes/test_write_file/test_write_file.yaml
@@ -39,10 +39,9 @@ interactions:
      to user\"\n    },\n    \"command\": {\n        \"name\": \"command name\",\n        \"args\":
      {\n            \"arg name\": \"value\"\n        }\n    }\n} \nEnsure the response
      can be parsed by Python json.loads"}, {"role": "system", "content": "The current
-      time and date is Tue Jan 01 00:00:00 2000"}, {"role": "system", "content": "This
-      reminds you of these events from your past:\n\n\n"}, {"role": "user", "content":
-      "Determine which next command to use, and respond using the format specified
-      above:"}], "temperature": 0, "max_tokens": 3223}'
+      time and date is Tue Jan 01 00:00:00 2000"}, {"role": "user", "content": "Determine
+      which next command to use, and respond using the format specified above:"}],
+      "temperature": 0, "max_tokens": 2738}'
    headers:
      Accept:
      - '*/*'
@@ -51,7 +50,7 @@ interactions:
      Connection:
      - keep-alive
      Content-Length:
-      - '3490'
+      - '3401'
      Content-Type:
      - application/json
    method: POST
@@ -59,20 +58,20 @@ interactions:
  response:
    body:
      string: !!binary |
-        H4sIAAAAAAAAA7yTT2/TQBDF73yK0VxycSI3f0jjWwUHisQFAQVhFG3tib3NesfdHSutIn93tLbT
-        tEY9gbjOG7/3m9nxEXWOCWalkqyqzXS9ubt53H+68u9+FPeON2vzcb654m/v4/vP3zFCvr2jTIYv
-        ZhlXtSHRbDHCzJESyjG5eHs5X62Wi/llhBXnZDDBopbpYraaSuNueRov4guMsPGqIEyOWDuuatkK
-        78l6TNbrZYRn76f6PI4jFBZlnkqb9bKNMCtZZ+Qx+XnEivzJ1rEhTFB5r70oKwGSrZANAxxTCwCQ
-        opTcFKX4FBMYioNADxKKKV7DQRsDjSeQkuDgtNBWeLvThiDjqlI2B+FegMkHMobhhp3JJ6CtMCjo
-        Wq2qKIdJGfTtIegzeZDJLMXoebAj5dlqW/TpX0qCgpUB7f8iJIIO8lX8YF4SVOwFVF07rp1WQrBj
-        B1IGVfn9mLQ2yvaQU/j6n7aTOS060746vY0l6vzJ+sYFBiVwDaqCxmtbdEwZO0eZnHOGZSjp5PDS
-        YQGBUMie2s3jONvXpPav30Sg+HdT98FtdDrUwfSPOw0OPdOL1Y/QlSvGJ94LofdsMeJ4YTL+LZ7N
-        dKLtiAfw1LbYRrjTVvty2x81JuiFa4xQ25weMInbX+2b3wAAAP//AwAxpIznhgQAAA==
+        H4sIAAAAAAAAA7yTT2/bMAzF7/sUBC+5KEHSLG3q21a0a6/DiqGoh0C1GVuLLHoSDRsL/N0H2/nT
+        ethp2K581Hs/StQeTYoRJrmWpCjt9OpDfdd8+ni7+Hz3ePOzWa+emqf77Cb1tdzWqJBfvlMihxOz
+        hIvSkhh2qDDxpIVSjBaX64v1+8Xqaqmw4JQsRpiVMl3OVlOp/AtP58v5AhVWQWeE0R5Lz0UpG+Ed
+        uYDR1epa4dn7VL+YLxQKi7an0vXlvFWY5GwSChg977GgcLT1bAkj1CGYINpJB8lOyHUD7GMHABCj
+        5FxluYQYIzgUDwI10hVjfABHlIIwVIFAcoLaG6GN8GZrLEHCRaFd39ALMLknaxm+srfpBIwTBg19
+        q9MFpTDJO31Td/pMGpnMYlSvsz3pwM64bAD4khOIDjvw9KMyngIU9BdpCnraP85hQi8WHAR0WXou
+        vdFCsGUPkneqDrsxcmm1G2in8PifrinxRkxiQjF+J3Kh8h2DFng4vVrC3lMi54jhHqiRM5Nw39q1
+        jNNCSXp3TKqNtf9yHYbgVh3X9GD625Z2DgPTG4gRuvbZeMEHoes9W4w43piMP8WrmY60PfEBPHYt
+        tgq3xpmQb4Z9xgiDcIkKjUupwWjefmvf/QIAAP//AwCaXwR3hAQAAA==
    headers:
      CF-Cache-Status:
      - DYNAMIC
      CF-RAY:
-      - 7be2da273c20ce54-SJC
+      - 7bfe3ef73b52fac2-SJC
      Cache-Control:
      - no-cache, must-revalidate
      Connection:
@@ -82,7 +81,7 @@ interactions:
      Content-Type:
      - application/json
      Date:
-      - Thu, 27 Apr 2023 00:12:22 GMT
+      - Sun, 30 Apr 2023 07:59:46 GMT
      Server:
      - cloudflare
      access-control-allow-origin:
@@ -94,19 +93,103 @@ interactions:
      openai-organization:
      - user-adtx4fhfg1qsiyzdoaxciooj
      openai-processing-ms:
-      - '14714'
+      - '13160'
      openai-version:
      - '2020-10-01'
      strict-transport-security:
      - max-age=15724800; includeSubDomains
      x-ratelimit-limit-requests:
      - '3500'
+      x-ratelimit-limit-tokens:
+      - '90000'
      x-ratelimit-remaining-requests:
      - '3499'
+      x-ratelimit-remaining-tokens:
+      - '86497'
      x-ratelimit-reset-requests:
      - 17ms
+      x-ratelimit-reset-tokens:
+      - 2.335s
      x-request-id:
-      - 23366aac1ddcc141fc735b4a1a8b268c
+      - f665162ae22af897be24f632a031d434
+    status:
+      code: 200
+      message: OK
+- request:
+    body: '{"model": "gpt-3.5-turbo", "messages": [{"role": "user", "content": "Your
+      task is to create a concise running summary of actions and information results
+      in the provided text, focusing on key and potentially important information
+      to remember.\n\n\nYou will receive the current summary and the latest development.
+      Combine them, adding relevant key information from the latest development in
+      1st person past tense and keeping the summary concise.\n\nSummary So Far:\n\"\"\"\nI
+      was created.\n\"\"\"\n\nLatest Development:\n\"\"\"\nNothing new happened.\n\"\"\"\n"}],
+      "temperature": 0, "max_tokens": null}'
+    headers:
+      Accept:
+      - '*/*'
+      Accept-Encoding:
+      - gzip, deflate
+      Connection:
+      - keep-alive
+      Content-Length:
+      - '600'
+      Content-Type:
+      - application/json
+    method: POST
+    uri: https://api.openai.com/v1/chat/completions
+  response:
+    body:
+      string: !!binary |
+        H4sIAAAAAAAAA0SOQUsDMRQG7/6K8J2zZdNa3eZmeygiPSqISEmTZze6mxc2r1Qt+9+lUPU6MMOc
+        EAMsfOvE97mrbu+O69X3lJrdks2D2Sy/5o/rp+fgu9ViAw3evZOXizHx3OeOJHKChh/ICQVYc9NM
+        m2szbxqNngN1sNhnqWaTeSWHYcdVPasNNA7F7Qn2hDxwn2Ur/EGpwC6Mxn/6H2sIi+v+gKnrUcO3
+        HD0V2JcTeiq/zYE7goUrJRZxSc6HnITS+f5eHV1Rl2PlUlCJpY1prxIdVeuKal3OlChMMGq8xRRL
+        ux3IFU6wKMIZGjEF+oStx9fx6gcAAP//AwCGJ6JPSQEAAA==
+    headers:
+      CF-Cache-Status:
+      - DYNAMIC
+      CF-RAY:
+      - 7bfe3f5a5fb3fac2-SJC
+      Cache-Control:
+      - no-cache, must-revalidate
+      Connection:
+      - keep-alive
+      Content-Encoding:
+      - gzip
+      Content-Type:
+      - application/json
+      Date:
+      - Sun, 30 Apr 2023 07:59:49 GMT
+      Server:
+      - cloudflare
+      access-control-allow-origin:
+      - '*'
+      alt-svc:
+      - h3=":443"; ma=86400, h3-29=":443"; ma=86400
+      openai-model:
+      - gpt-3.5-turbo-0301
+      openai-organization:
+      - user-adtx4fhfg1qsiyzdoaxciooj
+      openai-processing-ms:
+      - '634'
+      openai-version:
+      - '2020-10-01'
+      strict-transport-security:
+      - max-age=15724800; includeSubDomains
+      x-ratelimit-limit-requests:
+      - '3500'
+      x-ratelimit-limit-tokens:
+      - '90000'
+      x-ratelimit-remaining-requests:
+      - '3499'
+      x-ratelimit-remaining-tokens:
+      - '89866'
+      x-ratelimit-reset-requests:
+      - 17ms
+      x-ratelimit-reset-tokens:
+      - 88ms
+      x-request-id:
+      - 0bc738f58569b4b3635fe0c8384ebcfb
    status:
      code: 200
      message: OK
@@ -151,23 +234,23 @@ interactions:
      {\n            \"arg name\": \"value\"\n        }\n    }\n} \nEnsure the response
      can be parsed by Python json.loads"}, {"role": "system", "content": "The current
      time and date is Tue Jan 01 00:00:00 2000"}, {"role": "system", "content": "This
-      reminds you of these events from your past:\nNone\n\n"}, {"role": "user", "content":
-      "Determine which next command to use, and respond using the format specified
-      above:"}, {"role": "assistant", "content": "{\n    \"thoughts\": {\n        \"text\":
+      reminds you of these events from your past: \nI was created and nothing new
+      has happened."}, {"role": "user", "content": "Determine which next command to
+      use, and respond using the format specified above:"}, {"role": "assistant",
+      "content": "{\n    \"thoughts\": {\n        \"text\": \"I need to use the write_to_file
+      command to write ''Hello World'' into a file named ''hello_world.txt''.\",\n        \"reasoning\":
+      \"The task requires me to write ''Hello World'' into a file named ''hello_world.txt'',
+      and the write_to_file command is the most appropriate for this task.\",\n        \"plan\":
+      \"- Use the write_to_file command to write ''Hello World'' into a file named
+      ''hello_world.txt''.\",\n        \"criticism\": \"I need to ensure that I use
+      the correct file name and text to write to the file.\",\n        \"speak\":
      \"I will use the write_to_file command to write ''Hello World'' into a file
-      named ''hello_world.txt''.\",\n        \"reasoning\": \"The goal is to write
-      ''Hello World'' into a file named ''hello_world.txt'', and the write_to_file
-      command is the most appropriate for this task.\",\n        \"plan\": \"- Use
-      the write_to_file command to write ''Hello World'' into a file named ''hello_world.txt''.\",\n        \"criticism\":
-      \"I need to ensure that I am using the correct file name and that the text is
-      written correctly.\",\n        \"speak\": \"I will use the write to file command
-      to write ''Hello World'' into a file named ''hello_world.txt''.\"\n    },\n    \"command\":
-      {\n        \"name\": \"write_to_file\",\n        \"args\": {\n            \"filename\":
-      \"hello_world.txt\",\n            \"text\": \"Hello World\"\n        }\n    }\n}"},
-      {"role": "system", "content": "Command write_to_file returned: File written
-      to successfully."}, {"role": "user", "content": "Determine which next command
-      to use, and respond using the format specified above:"}], "temperature": 0,
-      "max_tokens": 2970}'
+      named ''hello_world.txt''.\"\n    },\n    \"command\": {\n        \"name\":
+      \"write_to_file\",\n        \"args\": {\n            \"filename\": \"hello_world.txt\",\n            \"text\":
+      \"Hello World\"\n        }\n    }\n}"}, {"role": "system", "content": "Command
+      write_to_file returned: File written to successfully."}, {"role": "user", "content":
+      "Determine which next command to use, and respond using the format specified
+      above:"}], "temperature": 0, "max_tokens": 2486}'
    headers:
      Accept:
      - '*/*'
@@ -176,7 +259,7 @@ interactions:
      Connection:
      - keep-alive
      Content-Length:
-      - '4612'
+      - '4646'
      Content-Type:
      - application/json
    method: POST
@@ -184,19 +267,21 @@ interactions:
  response:
    body:
      string: !!binary |
-        H4sIAAAAAAAAA6SSQW8TMRCF7/yK0ZydKkuSttkbAiGFA+JQBIKtItc7Wbvx2pZnFoqi/e9om00T
-        QishcZ3x++bNPO/Q1ViisVpMm/zkann/9b3JZnm1XMnHmzfLd7b4IrL6xB++vUWF8e6ejIyKCxPb
-        5ElcDKjQZNJCNZbF5fXrxWI+my8UtrEmjyU2SSazi8VEunwXJ9PZtECFHeuGsNxhyrFNspa4pcBY
-        FtPiWuERfmzMLhVKFO2PpWIx7xUaG50hxvL7DlviAzhHT1iiZnYsOshgMwahMKywqwIAQIViY9dY
-        4QpLGItjgx5kKFa4gp/Oe+iYQCyBaN6uR4MEJratDjVIhKfa4dVFheoUmUlzDC40e+6NJWii9uD4
-        WbmCR/CLIweZJWgjC+iUckzZaSHYxAxih+4zHpLXYT9+Ap//eyOTnTjjuD1cKhA9KilwlwedFliB
-        1T/oCXZcCbgzhpg3nfe/ztGcSG9fDAD+1e4e2atD3uPzv+IOuqX9tD+ucWZK5+b8p5xGO+Z66q8+
-        X/Ko7EdvVeixV7hxwbFd70lYIktMqNCFmh6wnPa3/avfAAAA//8DAKZ6myqyAwAA
+        H4sIAAAAAAAAA7SSTW/TQBCG7/yK0VxycaIkJST4VqmijYSEQEWIYhRt1xN7yXp3uztOgiL/d7Sx
+        TUpJOYC4zuf7zLwHVDmmKEvBsnJ6OL/cXV+9mX2+0+aag7u6vHlvH+7myw/bd2+XmKC9/0aSu46R
+        tJXTxMoaTFB6Ekw5ppNXi+ni5WS2eJ1gZXPSmGLheHgxmg259vd2OL4YTzDBOoiCMD2g87ZyvGK7
+        IRMwnYyn8wRPw0+JxSJBtiz0KTSdzJoEZWmVpIDplwNWFPrB3mrCFEUIKrAwHGVaw2QiwiEzAAAZ
+        cmnrouSQYQpdsEvQnmMwwyWUYksQaikphHWt9XfYecVMBgY3pLWFT9brfADKsAUBa6UJjKgoh0EZ
+        86tdzI94z4MR3JYEhvYMgcmBCsAW6kDAJQGLsFl18ATSVpUweSz4GeurRhkmj+V6EsEaZYpW821X
+        Bp4eauUpQEVn54BYM/kjjzLFOZ5Y+SeiBI4an1UfCUuCygYG4Zy3zivBBGvrgcuYPYPjtDAtyRA+
+        /vNxZKSTKlT9Qw3RsZNMqH3sEwxLcN5uVU4gYCu0yqG96VFob0hTPLskOBKbfsFOaf13X21HNklv
+        0K78N3/Gb7Tbfhn/RJTwxVNrP/bL/zB4j3DE6Ggy02CT4FoZFcpVuxtTDGwdJqhMTntMx83X5sUP
+        AAAA//8DAK0qY5KVBAAA
    headers:
      CF-Cache-Status:
      - DYNAMIC
      CF-RAY:
-      - 7be2da943bd5ce54-SJC
+      - 7bfe3f5edbd9fac2-SJC
      Cache-Control:
      - no-cache, must-revalidate
      Connection:
@@ -206,7 +291,7 @@ interactions:
      Content-Type:
      - application/json
      Date:
-      - Thu, 27 Apr 2023 00:12:36 GMT
+      - Sun, 30 Apr 2023 08:00:00 GMT
      Server:
      - cloudflare
      access-control-allow-origin:
@@ -218,19 +303,25 @@ interactions:
      openai-organization:
      - user-adtx4fhfg1qsiyzdoaxciooj
      openai-processing-ms:
-      - '10505'
+      - '10898'
      openai-version:
      - '2020-10-01'
      strict-transport-security:
      - max-age=15724800; includeSubDomains
      x-ratelimit-limit-requests:
      - '3500'
+      x-ratelimit-limit-tokens:
+      - '90000'
      x-ratelimit-remaining-requests:
      - '3499'
+      x-ratelimit-remaining-tokens:
+      - '86482'
      x-ratelimit-reset-requests:
      - 17ms
+      x-ratelimit-reset-tokens:
+      - 2.345s
      x-request-id:
-      - f593fff7d415793b53e5124283ad25c9
+      - f294799d2564196ef4852c84fdc8fb17
    status:
      code: 200
      message: OK
--- a/tests/unit/test_chat.py
+++ b/tests/unit/test_chat.py
@@ -33,18 +33,14 @@ def test_generate_context_empty_inputs(mocker):
    # Assert
    expected_result = (
        -1,
-        47,
-        3,
+        32,
+        2,
        [
            {"role": "system", "content": ""},
            {
                "role": "system",
                "content": f"The current time and date is {time.strftime('%c')}",
            },
-            {
-                "role": "system",
-                "content": f"This reminds you of these events from your past:\n\n\n",
-            },
        ],
    )
    assert result == expected_result
@@ -78,5 +74,5 @@ def test_generate_context_valid_inputs():
    assert result[0] >= 0
    assert result[2] >= 0
    assert result[1] >= 0
-    assert len(result[3]) >= 3  # current_context should have at least 3 messages
+    assert len(result[3]) >= 2  # current_context should have at least 2 messages
    assert result[1] <= 2048  # token limit for GPT-3.5-turbo-0301 is 2048 tokens
--- a/tests/vcr/openai_filter.py
+++ b/tests/vcr/openai_filter.py
@@ -41,7 +41,10 @@ def before_record_request(request):


 def filter_hostnames(request):
-    allowed_hostnames = ["api.openai.com"]  # List of hostnames you want to allow
+    allowed_hostnames = [
+        "api.openai.com",
+        "localhost:50337",
+    ]  # List of hostnames you want to allow

    if any(hostname in request.url for hostname in allowed_hostnames):
        return request