feat: introducing a "Generate" API for prompt-based content generation with saved prompts

CHANGELOG.md

@@ -1,14 +1,22 @@
 ### Changelog
+#### Unreleased
 
-All notable changes to this project will be documented in this file. Dates are displayed in UTC.
+- feat: introducing a "Generate" API for prompt-based content generation with saved prompts
 
-Generated by [`auto-changelog`](https://github.com/CookPete/auto-changelog).
+<!-- auto-changelog-below -->
+
+#### [v0.1.49](https://github.com/Portkey-AI/portkey-python-sdk/compare/v0.1.48...v0.1.49)
+
+> 16 September 2023
+
+- fix: Fixing the issue with the capturing of env variables [`#13`](https://github.com/Portkey-AI/portkey-python-sdk/pull/13)
 
 #### [v0.1.48](https://github.com/Portkey-AI/portkey-python-sdk/compare/v0.1.45...v0.1.48)
 
 > 14 September 2023
 
 - fix: cache age type fix [`#11`](https://github.com/Portkey-AI/portkey-python-sdk/pull/11)
+- docs: updated change log [`#10`](https://github.com/Portkey-AI/portkey-python-sdk/pull/10)
 
 #### [v0.1.45](https://github.com/Portkey-AI/portkey-python-sdk/compare/v0.1.44...v0.1.45)
 

portkey/__init__.py

@@ -19,6 +19,7 @@
  ChatCompletionChunk,
  TextCompletion,
  TextCompletionChunk,
+ Generations,
 )
 from portkey.version import VERSION
 from portkey.api_resources.global_constants import (
@@ -50,6 +51,7 @@
  "ChatCompletionChunk",
  "TextCompletion",
  "TextCompletionChunk",
+ "Generations",
  "Config",
  "api_key",
  "base_url",

portkey/api_resources/__init__.py

@@ -1,5 +1,5 @@
 """"""
-from .apis import ChatCompletions, Completions
+from .apis import ChatCompletions, Completions, Generations
 from .utils import (
  Modes,
  ModesLiteral,
@@ -41,4 +41,5 @@
  "ChatCompletionChunk",
  "TextCompletion",
  "TextCompletionChunk",
+ "Generations",
 ]

portkey/api_resources/apis.py

@@ -10,7 +10,7 @@
  ChatCompletion,
  TextCompletion,
  TextCompletionChunk,
- ApiType,
+ GenericResponse,
 )
 
 from .streaming import Stream
@@ -49,7 +49,7 @@ def create(
  max_tokens: Optional[int] = None,
  top_k: Optional[int] = None,
  top_p: Optional[float] = None,
- **kwargs
+ **kwargs,
  ) -> Stream[TextCompletionChunk]:
  ...
 
@@ -65,7 +65,7 @@ def create(
  max_tokens: Optional[int] = None,
  top_k: Optional[int] = None,
  top_p: Optional[float] = None,
- **kwargs
+ **kwargs,
  ) -> TextCompletion:
  ...
 
@@ -81,7 +81,7 @@ def create(
  max_tokens: Optional[int] = None,
  top_k: Optional[int] = None,
  top_p: Optional[float] = None,
- **kwargs
+ **kwargs,
  ) -> Union[TextCompletion, Stream[TextCompletionChunk]]:
  ...
 
@@ -96,7 +96,7 @@ def create(
  max_tokens: Optional[int] = None,
  top_k: Optional[int] = None,
  top_p: Optional[float] = None,
- **kwargs
+ **kwargs,
  ) -> Union[TextCompletion, Stream[TextCompletionChunk]]:
  if config is None:
  config = retrieve_config()
@@ -107,7 +107,7 @@ def create(
  max_tokens=max_tokens,
  top_k=top_k,
  top_p=top_p,
- **kwargs
+ **kwargs,
  )
  if config.mode == Modes.SINGLE.value:
  return cls(_client)._post(
@@ -118,7 +118,6 @@ def create(
  cast_to=TextCompletion,
  stream_cls=Stream[TextCompletionChunk],
  stream=stream,
- _type=ApiType.COMPLETIONS,
  )
  if config.mode == Modes.FALLBACK.value:
  return cls(_client)._post(
@@ -129,7 +128,6 @@ def create(
  cast_to=TextCompletion,
  stream_cls=Stream[TextCompletionChunk],
  stream=stream,
- _type=ApiType.COMPLETIONS,
  )
  if config.mode == Modes.AB_TEST.value:
  return cls(_client)._post(
@@ -140,7 +138,6 @@ def create(
  cast_to=TextCompletion,
  stream_cls=Stream[TextCompletionChunk],
  stream=stream,
- _type=ApiType.COMPLETIONS,
  )
  raise NotImplementedError("Mode not implemented.")
 
@@ -158,7 +155,7 @@ def create(
  max_tokens: Optional[int] = None,
  top_k: Optional[int] = None,
  top_p: Optional[float] = None,
- **kwargs
+ **kwargs,
  ) -> Stream[ChatCompletionChunk]:
  ...
 
@@ -174,7 +171,7 @@ def create(
  max_tokens: Optional[int] = None,
  top_k: Optional[int] = None,
  top_p: Optional[float] = None,
- **kwargs
+ **kwargs,
  ) -> ChatCompletion:
  ...
 
@@ -190,7 +187,7 @@ def create(
  max_tokens: Optional[int] = None,
  top_k: Optional[int] = None,
  top_p: Optional[float] = None,
- **kwargs
+ **kwargs,
  ) -> Union[ChatCompletion, Stream[ChatCompletionChunk]]:
  ...
 
@@ -205,7 +202,7 @@ def create(
  max_tokens: Optional[int] = None,
  top_k: Optional[int] = None,
  top_p: Optional[float] = None,
- **kwargs
+ **kwargs,
  ) -> Union[ChatCompletion, Stream[ChatCompletionChunk]]:
  if config is None:
  config = retrieve_config()
@@ -216,7 +213,7 @@ def create(
  max_tokens=max_tokens,
  top_k=top_k,
  top_p=top_p,
- **kwargs
+ **kwargs,
  )
  if config.mode == Modes.SINGLE.value:
  return cls(_client)._post(
@@ -227,7 +224,6 @@ def create(
  cast_to=ChatCompletion,
  stream_cls=Stream[ChatCompletionChunk],
  stream=stream,
- _type=ApiType.CHAT_COMPLETION,
  )
  if config.mode == Modes.FALLBACK.value:
  return cls(_client)._post(
@@ -238,7 +234,6 @@ def create(
  cast_to=ChatCompletion,
  stream_cls=Stream[ChatCompletionChunk],
  stream=stream,
- _type=ApiType.CHAT_COMPLETION,
  )
  if config.mode == Modes.AB_TEST.value:
  return cls(_client)._post(
@@ -249,6 +244,24 @@ def create(
  cast_to=ChatCompletion,
  stream_cls=Stream[ChatCompletionChunk],
  stream=stream,
- _type=ApiType.CHAT_COMPLETION,
  )
  raise NotImplementedError("Mode not implemented.")
+
+
+class Generations(APIResource):
+ @classmethod
+ def create(
+ cls, *, prompt_id: str, config: Optional[Config] = None, **kwargs
+ ) -> Union[GenericResponse, Stream[GenericResponse]]:
+ if config is None:
+ config = retrieve_config()
+ _client = APIClient(api_key=config.api_key, base_url=config.base_url)
+ return cls(_client)._post(
+ f"/v1/prompts/{prompt_id}/generate",
+ body=config.llms,
+ mode=None,
+ params=None,
+ cast_to=GenericResponse,
+ stream_cls=Stream[GenericResponse],
+ stream=False,
+ )

portkey/api_resources/base_client.py

@@ -13,6 +13,7 @@
  Type,
  overload,
  Literal,
+ get_args,
 )
 import httpx
 import platform
@@ -26,7 +27,7 @@
  ProviderOptions,
  Params,
  Constructs,
- ApiType,
+ PortkeyApiPaths,
 )
 from .exceptions import (
  APIStatusError,
@@ -91,7 +92,6 @@ def post(
  stream: Literal[True],
  stream_cls: type[StreamT],
  params: Params,
- _type: ApiType,
  ) -> StreamT:
  ...
 
@@ -106,7 +106,6 @@ def post(
  stream: Literal[False],
  stream_cls: type[StreamT],
  params: Params,
- _type: ApiType,
  ) -> ResponseT:
  ...
 
@@ -121,35 +120,68 @@ def post(
  stream: bool,
  stream_cls: type[StreamT],
  params: Params,
- _type: ApiType,
  ) -> Union[ResponseT, StreamT]:
  ...
 
  def post(
  self,
  path: str,
  *,
- body: List[Body],
+ body: Union[List[Body], Any],
  mode: str,
  cast_to: Type[ResponseT],
  stream: bool,
  stream_cls: type[StreamT],
  params: Params,
- _type: ApiType,
  ) -> Union[ResponseT, StreamT]:
- body = cast(List[Body], body)
- opts = self._construct(
- method="post", url=path, body=body, mode=mode, stream=stream, params=params
- )
+ if path in [PortkeyApiPaths.CHAT_COMPLETION, PortkeyApiPaths.COMPLETION]:
+ body = cast(List[Body], body)
+ opts = self._construct(
+ method="post",
+ url=path,
+ body=body,
+ mode=mode,
+ stream=stream,
+ params=params,
+ )
+ elif path.endswith("/generate"):
+ opts = self._construct_generate_options(
+ method="post",
+ url=path,
+ body=body,
+ mode=mode,
+ stream=stream,
+ params=params,
+ )
+ else:
+ raise NotImplementedError(f"This API path `{path}` is not implemented.")
+
  res = self._request(
  options=opts,
  stream=stream,
  cast_to=cast_to,
  stream_cls=stream_cls,
- _type=_type,
  )
  return res
 
+ def _construct_generate_options(
+ self,
+ *,
+ method: str,
+ url: str,
+ body: Any,
+ mode: str,
+ stream: bool,
+ params: Params,
+ ) -> Options:
+ opts = Options.construct()
+ opts.method = method
+ opts.url = url
+ json_body = body
+ opts.json_body = remove_empty_values(json_body)
+ opts.headers = None
+ return opts
+
  def _construct(
  self,
  *,
@@ -252,7 +284,6 @@ def _request(
  stream: Literal[False],
  cast_to: Type[ResponseT],
  stream_cls: Type[StreamT],
- _type: ApiType,
  ) -> ResponseT:
  ...
 
@@ -264,7 +295,6 @@ def _request(
  stream: Literal[True],
  cast_to: Type[ResponseT],
  stream_cls: Type[StreamT],
- _type: ApiType,
  ) -> StreamT:
  ...
 
@@ -276,7 +306,6 @@ def _request(
  stream: bool,
  cast_to: Type[ResponseT],
  stream_cls: Type[StreamT],
- _type: ApiType,
  ) -> Union[ResponseT, StreamT]:
  ...
 
@@ -287,7 +316,6 @@ def _request(
  stream: bool,
  cast_to: Type[ResponseT],
  stream_cls: Type[StreamT],
- _type: ApiType,
  ) -> Union[ResponseT, StreamT]:
  request = self._build_request(options)
  try:
@@ -302,18 +330,28 @@ def _request(
  raise APITimeoutError(request=request) from err
  except Exception as err:
  raise APIConnectionError(request=request) from err
- if stream:
- # stream_cls = stream_cls
+ if stream or res.headers["content-type"] == "text/event-stream":
  if stream_cls is None:
  raise MissingStreamClassError()
- stream_response = stream_cls(response=res, _type=_type)
+ stream_response = stream_cls(
+ response=res, cast_to=self._extract_stream_chunk_type(stream_cls)
+ )
  return stream_response
+
  response = cast(
  ResponseT,
  cast_to(**res.json()),
  )
  return response
 
+ def _extract_stream_chunk_type(self, stream_cls: Type) -> type:
+ args = get_args(stream_cls)
+ if not args:
+ raise TypeError(
+ f"Expected stream_cls to have been given a generic type argument, e.g. Stream[Foo] but received {stream_cls}",
+ )
+ return cast(type, args[0])
+
  def _make_status_error_from_response(
  self,
  request: httpx.Request,