Upgrade OpenAI Dependency (to fix "Quota Reached" error) & some lint …

…fixes (#21)

* openai-version-bump

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* fixes

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* run fmt

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* s

* not going to bother

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* update disc

* str

---------

Co-authored-by: Josh <[email protected]>
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>

app/adapters/openai/gpt.py

@@ -5,9 +5,15 @@
 from typing import Required
 from typing import TypedDict
 
-import backoff
-import openai.error
-from openai.openai_object import OpenAIObject
+import openai
+from openai.types.chat import ChatCompletion
+
+from app import settings
+
+
+openai_client = openai.AsyncOpenAI(
+    api_key=settings.OPENAI_API_KEY,
+)
 
 
 class OpenAIModel(StrEnum):
@@ -55,56 +61,11 @@ class FunctionSchema(TypedDict):
     parameters: Parameters
 
 
-class GPTResponse(TypedDict):
-    choices: Sequence[Message]
-
-
-MAX_BACKOFF_TIME = 16
-
-
-def _is_non_retriable_error(error: Exception) -> bool:
-    """\
-    Determine whether an error is non-retriable.
-    """
-    if isinstance(error, openai.error.APIConnectionError):
-        return error.should_retry  # TODO: confirm this
-    elif isinstance(
-        error,
-        (
-            openai.error.APIError,  # TODO: confirm this
-            openai.error.TryAgain,
-            openai.error.Timeout,
-            openai.error.RateLimitError,
-            openai.error.ServiceUnavailableError,
-        ),
-    ):
-        return True
-    elif isinstance(
-        error,
-        (
-            openai.error.InvalidRequestError,
-            openai.error.AuthenticationError,
-            openai.error.PermissionError,
-            openai.error.InvalidAPIType,
-            openai.error.SignatureVerificationError,
-        ),
-    ):
-        return False
-    else:
-        raise NotImplementedError(f"Unknown error type: {error}")
-
-
-@backoff.on_exception(
-    backoff.expo,
-    openai.error.OpenAIError,
-    max_time=MAX_BACKOFF_TIME,
-    giveup=_is_non_retriable_error,
-)
 async def send(
     model: OpenAIModel,
     messages: Sequence[Message],
     functions: Sequence[FunctionSchema] | None = None,
-) -> OpenAIObject:
+) -> ChatCompletion:
     """\
     Send a message to the OpenAI API, as a given model.
 
@@ -114,6 +75,5 @@ async def send(
     if functions is not None:
         kwargs["functions"] = functions
 
-    response = await openai.ChatCompletion.acreate(**kwargs)
-    assert isinstance(response, OpenAIObject)
+    response = await openai_client.chat.completions.create(**kwargs)
     return response

app/main.py

@@ -37,7 +37,7 @@ class Bot(discord.Client):
     async def start(self, *args, **kwargs) -> None:
         state.read_database = database.Database(
             database.dsn(
-                scheme="postgresql",
+                scheme=settings.READ_DB_SCHEME,
                 user=settings.READ_DB_USER,
                 password=settings.READ_DB_PASS,
                 host=settings.READ_DB_HOST,
@@ -52,7 +52,7 @@ async def start(self, *args, **kwargs) -> None:
 
         state.write_database = database.Database(
             database.dsn(
-                scheme="postgresql",
+                scheme=settings.WRITE_DB_SCHEME,
                 user=settings.WRITE_DB_USER,
                 password=settings.WRITE_DB_PASS,
                 host=settings.WRITE_DB_HOST,
@@ -117,11 +117,6 @@ async def on_ready():
     # NOTE: we can't use this as a lifecycle hook because
     # it may be called more than a single time.
     # our lifecycle hook is in our Bot class definition
-
-    import openai
-
-    openai.api_key = settings.OPENAI_API_KEY
-
     await command_tree.sync()
 
 
@@ -212,22 +207,32 @@ async def on_message(message: discord.Message):
         )
         if not gpt_response:
             await message.channel.send(
-                f"Request failed after multiple retries.\n"
-                f"Please try again after some time.\n"
-                f"If this issue persists, please contact cmyui#0425 on discord."
+                "Request failed after multiple retries.\n"
+                "Please try again after some time.\n"
+                'If this issue persists, please contact "cmyui" on discord.'
             )
             return
 
-        gpt_choice = gpt_response["choices"][0]
-        gpt_message = gpt_choice["message"]
-        message_history.append(gpt_message)
+        gpt_choice = gpt_response.choices[0]
+        gpt_message = gpt_choice.message
+        assert gpt_message.content is not None
 
-        if gpt_choice["finish_reason"] == "stop":
-            gpt_response_content = gpt_message["content"]
+        message_history.append(
+            gpt.Message(
+                role=gpt_message.role,
+                content=gpt_message.content,
+            )
+        )
+
+        if gpt_choice.finish_reason == "stop":
+            gpt_response_content: str = gpt_message.content
 
-        elif gpt_choice["finish_reason"] == "function_call":
-            function_name = gpt_message["function_call"]["name"]
-            function_kwargs = json.loads(gpt_message["function_call"]["arguments"])
+        elif (
+            gpt_choice.finish_reason == "function_call"
+            and gpt_message.function_call is not None
+        ):
+            function_name = gpt_message.function_call.name
+            function_kwargs = json.loads(gpt_message.function_call.arguments)
 
             ai_function = openai_functions.ai_functions[function_name]
             function_response = await ai_function["callback"](**function_kwargs)
@@ -243,13 +248,15 @@ async def on_message(message: discord.Message):
                 }
             )
             gpt_response = await gpt.send(tracked_thread["model"], message_history)
-            gpt_response_content = gpt_response["choices"][0]["message"]["content"]
+            assert gpt_response.choices[0].message.content is not None
+            gpt_response_content = gpt_response.choices[0].message.content
 
         else:
             raise NotImplementedError(
-                f"Unknown chatgpt finish reason: {gpt_choice['finish_reason']}"
+                f"Unknown chatgpt finish reason: {gpt_choice.finish_reason}"
             )
 
+        assert gpt_response.usage is not None
         input_tokens = gpt_response.usage.prompt_tokens
         output_tokens = gpt_response.usage.completion_tokens
 
@@ -344,8 +351,8 @@ async def monthlycost(interaction: discord.Interaction):
     response_cost = sum(per_user_cost.values())
 
     message_chunks = [
-        f"**Monthly Cost Breakdown**",
-        f"**----------------------**",
+        "**Monthly Cost Breakdown**",
+        "**----------------------**",
         "",
     ]
     for user_id, cost in per_user_cost.items():
@@ -380,8 +387,8 @@ async def threadcost(interaction: discord.Interaction):
     response_cost = sum(per_user_cost.values())
 
     message_chunks = [
-        f"**Thread Cost Breakdown**",
-        f"**---------------------**",
+        "**Thread Cost Breakdown**",
+        "**---------------------**",
         "",
     ]
     for user_id, cost in per_user_cost.items():
@@ -472,7 +479,7 @@ async def context(
         content="\n".join(
             (
                 f"**Context length (messages length preserved) updated to {context_length}**",
-                f"NOTE: longer context costs linearly more tokens, so please take care.",
+                "NOTE: longer context costs linearly more tokens, so please take care.",
             )
         )
     )
@@ -542,6 +549,8 @@ async def summarize(
     gpt_response = await gpt.send(gpt.OpenAIModel.GPT_4_OMNI, messages)
 
     gpt_response_content = gpt_response.choices[0].message.content
+    assert gpt_response_content is not None
+
     # tokens_spent = gpt_response.usage.total_tokens
 
     for chunk in split_message(gpt_response_content, 2000):

app/openai_functions.py

@@ -61,7 +61,7 @@ def get_function_openai_schema(f: Callable[..., Awaitable[str]]) -> FunctionSche
             or len(param_type.__metadata__) != 1
         ):
             logging.warning(
-                f"Function decorated with @ai_function lacks parameter description annotation(s)",
+                "Function decorated with @ai_function lacks parameter description annotation(s)",
                 extra={
                     "param_name": param_name,
                     "param_type": param_type,

requirements.txt

@@ -1,7 +1,6 @@
 asyncpg
-backoff
 databases
 discord.py
 httpx
-openai<1.0.0
+openai
 python-dotenv