chore(forge): Upgrade OpenAI client lib and LiteLLM from v0 to v1

* Update `openai` dependency from `^0.27.8` to `^1.7.2`
* Update `litellm` dependency from `^0.1.821` to `^1.17.9`
* Migrate llm.py from OpenAI module-level client to client instance
* Update return types in llm.py for new OpenAI and LiteLLM versions
   * Also remove `Exception` as a return type because they are raised, not returned
   * Update tutorials/003_crafting_agent_logic.md accordingly

Note: this changes the output types of the functions in `forge.llm`: `chat_completion_request`, `create_embedding_request`, `transcribe_audio`

autogpts/forge/forge/llm.py

@@ -1,18 +1,18 @@
-import typing
+from pathlib import Path
 
-import openai
+from litellm import AuthenticationError, InvalidRequestError, ModelResponse, acompletion
+from openai import OpenAI
+from openai.types import CreateEmbeddingResponse
+from openai.types.audio import Transcription
 from tenacity import retry, stop_after_attempt, wait_random_exponential
 
 from .sdk.forge_log import ForgeLogger
-from litellm import completion, acompletion, AuthenticationError, InvalidRequestError
 
 LOG = ForgeLogger(__name__)
 
 
 @retry(wait=wait_random_exponential(min=1, max=40), stop=stop_after_attempt(3))
-async def chat_completion_request(
+async def chat_completion_request(model, messages, **kwargs) -> ModelResponse:
-    model, messages, **kwargs
-) -> typing.Union[typing.Dict[str, typing.Any], Exception]:
     """Generate a response to a list of messages using OpenAI's API"""
     try:
         kwargs["model"] = model
@@ -22,8 +22,10 @@ async def chat_completion_request(
         return resp
     except AuthenticationError as e:
         LOG.exception("Authentication Error")
+        raise
     except InvalidRequestError as e:
         LOG.exception("Invalid Request Error")
+        raise
     except Exception as e:
         LOG.error("Unable to generate ChatCompletion response")
         LOG.error(f"Exception: {e}")
@@ -33,12 +35,12 @@ async def chat_completion_request(
 @retry(wait=wait_random_exponential(min=1, max=40), stop=stop_after_attempt(3))
 async def create_embedding_request(
     messages, model="text-embedding-ada-002"
-) -> typing.Union[typing.Dict[str, typing.Any], Exception]:
+) -> CreateEmbeddingResponse:
     """Generate an embedding for a list of messages using OpenAI's API"""
     try:
-        return await openai.Embedding.acreate(
+        return OpenAI().embeddings.create(
             input=[f"{m['role']}: {m['content']}" for m in messages],
-            engine=model,
+            model=model,
         )
     except Exception as e:
         LOG.error("Unable to generate ChatCompletion response")
@@ -47,12 +49,12 @@ async def create_embedding_request(
 
 
 @retry(wait=wait_random_exponential(min=1, max=40), stop=stop_after_attempt(3))
-async def transcribe_audio(
+async def transcribe_audio(audio_file: Path) -> Transcription:
-    audio_file: str,
-) -> typing.Union[typing.Dict[str, typing.Any], Exception]:
     """Transcribe an audio file using OpenAI's API"""
     try:
-        return await openai.Audio.transcribe(model="whisper-1", file=audio_file)
+        return OpenAI().audio.transcriptions.create(
+            model="whisper-1", file=audio_file.open(mode="rb")
+        )
     except Exception as e:
         LOG.error("Unable to generate ChatCompletion response")
         LOG.error(f"Exception: {e}")

autogpts/forge/pyproject.toml

@@ -10,7 +10,7 @@ packages = [{ include = "forge" }]
 [tool.poetry.dependencies]
 python = "^3.10"
 python-dotenv = "^1.0.0"
-openai = "^0.27.8"
+openai = "^1.7.2"
 tenacity = "^8.2.2"
 sqlalchemy = "^2.0.19"
 aiohttp = "^3.8.5"
@@ -20,7 +20,7 @@ python-multipart = "^0.0.6"
 toml = "^0.10.2"
 jinja2 = "^3.1.2"
 uvicorn = "^0.23.2"
-litellm = "^0.1.821"
+litellm = "^1.17.9"
 duckduckgo-search = "^4.0.0"
 selenium = "^4.13.0"
 bs4 = "^0.0.1"

autogpts/forge/tutorials/003_crafting_agent_logic.md

@@ -242,7 +242,7 @@ try:
     }
     # Get the LLM's response and interpret it
     chat_response = await chat_completion_request(**chat_completion_kwargs)
-    answer = json.loads(chat_response["choices"][0]["message"]["content"])
+    answer = json.loads(chat_response.choices[0].message.content)
 
     # Log the answer for reference
     LOG.info(pprint.pformat(answer))
@@ -403,7 +403,7 @@ async def execute_step(self, task_id: str, step_request: StepRequestBody) -> Ste
         }
         # Make the chat completion request and parse the response
         chat_response = await chat_completion_request(**chat_completion_kwargs)
-        answer = json.loads(chat_response["choices"][0]["message"]["content"])
+        answer = json.loads(chat_response.choices[0].message.content)
 
         # Log the answer for debugging purposes
         LOG.info(pprint.pformat(answer))