Migrate to latest openai code #13
Conversation
| import openai | ||
| from openai import OpenAI | ||
|
|
||
| client = OpenAI() |
There was a problem hiding this comment.
This appears to only be used from AzureAIService; if so I'd suggest making it a class variable.
| import os | ||
| from openai import OpenAI | ||
|
|
||
| client = OpenAI(api_key=os.getenv("OPENAI_API_KEY")) |
There was a problem hiding this comment.
This appears to only be used from OpenAIService; if so, I'd suggest making it a class variable. It also looks like our earlier reference to the key was using the variable name OPEN_AI_KEY - was this meant to be that?
| image = client.images.generate(api_type = 'openai', | ||
| api_version = '2020-11-07', | ||
| api_base = "https://api.openai.com/v1", | ||
| api_key = os.getenv("OPEN_AI_KEY"), |
There was a problem hiding this comment.
If we specify the API key when constructing the client, I'd presume this might no longer be needed. (let's clean up the indentation as well!)
| if chunk["choices"][0]["delta"]["content"] != {}: #streaming a content chunk | ||
| next_chunk = chunk["choices"][0]["delta"]["content"] | ||
| if chunk.choices[0].delta.content: | ||
| if chunk.choices[0].delta.content != {}: #streaming a content chunk |
There was a problem hiding this comment.
This is more a note for us to follow up on as I realize it is not in scope of this PR (so more for @chadbailey59 and @Moishe maybe!), but the repeated list and property access here is redundant - I suggest, in a follow-up PR, accessing the required elements once for reuse (here and below). e.g.:
choices = chunk.choices
if len(choices):
continue
delta = chunk.choices[0].delta
if delta.content:
//... etc(Hastily written just to give an idea of what I mean)
| stream=stream, | ||
| messages=messages | ||
| ) | ||
| response = client.chat.completions.create(api_type = 'azure', |
There was a problem hiding this comment.
Suggest just cleaning up the indentation here for readability (and below)
|
It's great to have this updated code. I'm testing locally and the EDIT: @lazeratops already noted this: https://github.com/daily-demos/llm-talk/pull/13/files#r1412789861 |
3 participants
Fixed a bunch of errors that was breaking the demo. Almost all of it is around handle updated data structures in the openai sdk.
(credits to @jmtame for pairing on this together)