Skip to content

Commit 5ef20d7

Browse files
stainless-botRobertCraigie
stainless-bot
authored and
RobertCraigie
committed
fix(streaming): add additional overload for ambiguous stream param
1 parent 7b4c5f7 commit 5ef20d7

File tree

3 files changed

+628
-0
lines changed

3 files changed

+628
-0
lines changed

‎src/openai/resources/chat/completions.py‎

Lines changed: 250 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -271,6 +271,131 @@ def create(
271271
"""
272272
...
273273

274+
@overload
275+
def create(
276+
self,
277+
*,
278+
messages: List[ChatCompletionMessageParam],
279+
model: Union[
280+
str,
281+
Literal[
282+
"gpt-4",
283+
"gpt-4-0314",
284+
"gpt-4-0613",
285+
"gpt-4-32k",
286+
"gpt-4-32k-0314",
287+
"gpt-4-32k-0613",
288+
"gpt-3.5-turbo",
289+
"gpt-3.5-turbo-16k",
290+
"gpt-3.5-turbo-0301",
291+
"gpt-3.5-turbo-0613",
292+
"gpt-3.5-turbo-16k-0613",
293+
],
294+
],
295+
stream: bool,
296+
frequency_penalty: Optional[float] | NotGiven = NOT_GIVEN,
297+
function_call: completion_create_params.FunctionCall | NotGiven = NOT_GIVEN,
298+
functions: List[completion_create_params.Function] | NotGiven = NOT_GIVEN,
299+
logit_bias: Optional[Dict[str, int]] | NotGiven = NOT_GIVEN,
300+
max_tokens: Optional[int] | NotGiven = NOT_GIVEN,
301+
n: Optional[int] | NotGiven = NOT_GIVEN,
302+
presence_penalty: Optional[float] | NotGiven = NOT_GIVEN,
303+
stop: Union[Optional[str], List[str]] | NotGiven = NOT_GIVEN,
304+
temperature: Optional[float] | NotGiven = NOT_GIVEN,
305+
top_p: Optional[float] | NotGiven = NOT_GIVEN,
306+
user: str | NotGiven = NOT_GIVEN,
307+
# Use the following arguments if you need to pass additional parameters to the API that aren't available via kwargs.
308+
# The extra values given here take precedence over values defined on the client or passed to this method.
309+
extra_headers: Headers | None = None,
310+
extra_query: Query | None = None,
311+
extra_body: Body | None = None,
312+
timeout: float | None | NotGiven = NOT_GIVEN,
313+
) -> ChatCompletion | Stream[ChatCompletionChunk]:
314+
"""
315+
Creates a model response for the given chat conversation.
316+
317+
Args:
318+
messages: A list of messages comprising the conversation so far.
319+
[Example Python code](https://cookbook.openai.com/examples/how_to_format_inputs_to_chatgpt_models).
320+
321+
model: ID of the model to use. See the
322+
[model endpoint compatibility](https://platform.openai.com/docs/models/model-endpoint-compatibility)
323+
table for details on which models work with the Chat API.
324+
325+
stream: If set, partial message deltas will be sent, like in ChatGPT. Tokens will be
326+
sent as data-only
327+
[server-sent events](https://developer.mozilla.org/en-US/docs/Web/API/Server-sent_events/Using_server-sent_events#Event_stream_format)
328+
as they become available, with the stream terminated by a `data: [DONE]`
329+
message.
330+
[Example Python code](https://cookbook.openai.com/examples/how_to_stream_completions).
331+
332+
frequency_penalty: Number between -2.0 and 2.0. Positive values penalize new tokens based on their
333+
existing frequency in the text so far, decreasing the model's likelihood to
334+
repeat the same line verbatim.
335+
336+
[See more information about frequency and presence penalties.](https://platform.openai.com/docs/guides/gpt/parameter-details)
337+
338+
function_call: Controls how the model calls functions. "none" means the model will not call a
339+
function and instead generates a message. "auto" means the model can pick
340+
between generating a message or calling a function. Specifying a particular
341+
function via `{"name": "my_function"}` forces the model to call that function.
342+
"none" is the default when no functions are present. "auto" is the default if
343+
functions are present.
344+
345+
functions: A list of functions the model may generate JSON inputs for.
346+
347+
logit_bias: Modify the likelihood of specified tokens appearing in the completion.
348+
349+
Accepts a json object that maps tokens (specified by their token ID in the
350+
tokenizer) to an associated bias value from -100 to 100. Mathematically, the
351+
bias is added to the logits generated by the model prior to sampling. The exact
352+
effect will vary per model, but values between -1 and 1 should decrease or
353+
increase likelihood of selection; values like -100 or 100 should result in a ban
354+
or exclusive selection of the relevant token.
355+
356+
max_tokens: The maximum number of [tokens](/tokenizer) to generate in the chat completion.
357+
358+
The total length of input tokens and generated tokens is limited by the model's
359+
context length.
360+
[Example Python code](https://cookbook.openai.com/examples/how_to_count_tokens_with_tiktoken)
361+
for counting tokens.
362+
363+
n: How many chat completion choices to generate for each input message.
364+
365+
presence_penalty: Number between -2.0 and 2.0. Positive values penalize new tokens based on
366+
whether they appear in the text so far, increasing the model's likelihood to
367+
talk about new topics.
368+
369+
[See more information about frequency and presence penalties.](https://platform.openai.com/docs/guides/gpt/parameter-details)
370+
371+
stop: Up to 4 sequences where the API will stop generating further tokens.
372+
373+
temperature: What sampling temperature to use, between 0 and 2. Higher values like 0.8 will
374+
make the output more random, while lower values like 0.2 will make it more
375+
focused and deterministic.
376+
377+
We generally recommend altering this or `top_p` but not both.
378+
379+
top_p: An alternative to sampling with temperature, called nucleus sampling, where the
380+
model considers the results of the tokens with top_p probability mass. So 0.1
381+
means only the tokens comprising the top 10% probability mass are considered.
382+
383+
We generally recommend altering this or `temperature` but not both.
384+
385+
user: A unique identifier representing your end-user, which can help OpenAI to monitor
386+
and detect abuse.
387+
[Learn more](https://platform.openai.com/docs/guides/safety-best-practices/end-user-ids).
388+
389+
extra_headers: Send extra headers
390+
391+
extra_query: Add additional query parameters to the request
392+
393+
extra_body: Add additional JSON properties to the request
394+
395+
timeout: Override the client-level default timeout for this request, in seconds
396+
"""
397+
...
398+
274399
@required_args(["messages", "model"], ["messages", "model", "stream"])
275400
def create(
276401
self,
@@ -592,6 +717,131 @@ async def create(
592717
"""
593718
...
594719

720+
@overload
721+
async def create(
722+
self,
723+
*,
724+
messages: List[ChatCompletionMessageParam],
725+
model: Union[
726+
str,
727+
Literal[
728+
"gpt-4",
729+
"gpt-4-0314",
730+
"gpt-4-0613",
731+
"gpt-4-32k",
732+
"gpt-4-32k-0314",
733+
"gpt-4-32k-0613",
734+
"gpt-3.5-turbo",
735+
"gpt-3.5-turbo-16k",
736+
"gpt-3.5-turbo-0301",
737+
"gpt-3.5-turbo-0613",
738+
"gpt-3.5-turbo-16k-0613",
739+
],
740+
],
741+
stream: bool,
742+
frequency_penalty: Optional[float] | NotGiven = NOT_GIVEN,
743+
function_call: completion_create_params.FunctionCall | NotGiven = NOT_GIVEN,
744+
functions: List[completion_create_params.Function] | NotGiven = NOT_GIVEN,
745+
logit_bias: Optional[Dict[str, int]] | NotGiven = NOT_GIVEN,
746+
max_tokens: Optional[int] | NotGiven = NOT_GIVEN,
747+
n: Optional[int] | NotGiven = NOT_GIVEN,
748+
presence_penalty: Optional[float] | NotGiven = NOT_GIVEN,
749+
stop: Union[Optional[str], List[str]] | NotGiven = NOT_GIVEN,
750+
temperature: Optional[float] | NotGiven = NOT_GIVEN,
751+
top_p: Optional[float] | NotGiven = NOT_GIVEN,
752+
user: str | NotGiven = NOT_GIVEN,
753+
# Use the following arguments if you need to pass additional parameters to the API that aren't available via kwargs.
754+
# The extra values given here take precedence over values defined on the client or passed to this method.
755+
extra_headers: Headers | None = None,
756+
extra_query: Query | None = None,
757+
extra_body: Body | None = None,
758+
timeout: float | None | NotGiven = NOT_GIVEN,
759+
) -> ChatCompletion | AsyncStream[ChatCompletionChunk]:
760+
"""
761+
Creates a model response for the given chat conversation.
762+
763+
Args:
764+
messages: A list of messages comprising the conversation so far.
765+
[Example Python code](https://cookbook.openai.com/examples/how_to_format_inputs_to_chatgpt_models).
766+
767+
model: ID of the model to use. See the
768+
[model endpoint compatibility](https://platform.openai.com/docs/models/model-endpoint-compatibility)
769+
table for details on which models work with the Chat API.
770+
771+
stream: If set, partial message deltas will be sent, like in ChatGPT. Tokens will be
772+
sent as data-only
773+
[server-sent events](https://developer.mozilla.org/en-US/docs/Web/API/Server-sent_events/Using_server-sent_events#Event_stream_format)
774+
as they become available, with the stream terminated by a `data: [DONE]`
775+
message.
776+
[Example Python code](https://cookbook.openai.com/examples/how_to_stream_completions).
777+
778+
frequency_penalty: Number between -2.0 and 2.0. Positive values penalize new tokens based on their
779+
existing frequency in the text so far, decreasing the model's likelihood to
780+
repeat the same line verbatim.
781+
782+
[See more information about frequency and presence penalties.](https://platform.openai.com/docs/guides/gpt/parameter-details)
783+
784+
function_call: Controls how the model calls functions. "none" means the model will not call a
785+
function and instead generates a message. "auto" means the model can pick
786+
between generating a message or calling a function. Specifying a particular
787+
function via `{"name": "my_function"}` forces the model to call that function.
788+
"none" is the default when no functions are present. "auto" is the default if
789+
functions are present.
790+
791+
functions: A list of functions the model may generate JSON inputs for.
792+
793+
logit_bias: Modify the likelihood of specified tokens appearing in the completion.
794+
795+
Accepts a json object that maps tokens (specified by their token ID in the
796+
tokenizer) to an associated bias value from -100 to 100. Mathematically, the
797+
bias is added to the logits generated by the model prior to sampling. The exact
798+
effect will vary per model, but values between -1 and 1 should decrease or
799+
increase likelihood of selection; values like -100 or 100 should result in a ban
800+
or exclusive selection of the relevant token.
801+
802+
max_tokens: The maximum number of [tokens](/tokenizer) to generate in the chat completion.
803+
804+
The total length of input tokens and generated tokens is limited by the model's
805+
context length.
806+
[Example Python code](https://cookbook.openai.com/examples/how_to_count_tokens_with_tiktoken)
807+
for counting tokens.
808+
809+
n: How many chat completion choices to generate for each input message.
810+
811+
presence_penalty: Number between -2.0 and 2.0. Positive values penalize new tokens based on
812+
whether they appear in the text so far, increasing the model's likelihood to
813+
talk about new topics.
814+
815+
[See more information about frequency and presence penalties.](https://platform.openai.com/docs/guides/gpt/parameter-details)
816+
817+
stop: Up to 4 sequences where the API will stop generating further tokens.
818+
819+
temperature: What sampling temperature to use, between 0 and 2. Higher values like 0.8 will
820+
make the output more random, while lower values like 0.2 will make it more
821+
focused and deterministic.
822+
823+
We generally recommend altering this or `top_p` but not both.
824+
825+
top_p: An alternative to sampling with temperature, called nucleus sampling, where the
826+
model considers the results of the tokens with top_p probability mass. So 0.1
827+
means only the tokens comprising the top 10% probability mass are considered.
828+
829+
We generally recommend altering this or `temperature` but not both.
830+
831+
user: A unique identifier representing your end-user, which can help OpenAI to monitor
832+
and detect abuse.
833+
[Learn more](https://platform.openai.com/docs/guides/safety-best-practices/end-user-ids).
834+
835+
extra_headers: Send extra headers
836+
837+
extra_query: Add additional query parameters to the request
838+
839+
extra_body: Add additional JSON properties to the request
840+
841+
timeout: Override the client-level default timeout for this request, in seconds
842+
"""
843+
...
844+
595845
@required_args(["messages", "model"], ["messages", "model", "stream"])
596846
async def create(
597847
self,

0 commit comments

Comments
 (0)