Merge branch 'main' into add-documentation

libs/client_infinity/infinity_client/infinity_client/models/open_ai_embedding_input_audio.py

@@ -18,13 +18,15 @@ class OpenAIEmbeddingInputAudio:
         model (Union[Unset, str]):  Default: 'default/not-specified'.
         encoding_format (Union[Unset, EmbeddingEncodingFormat]):
         user (Union[None, Unset, str]):
+        dimensions (Union[Unset, int]):  Default: 0.
         modality (Union[Unset, OpenAIEmbeddingInputAudioModality]):  Default: OpenAIEmbeddingInputAudioModality.AUDIO.
     """
 
     input_: Union[List[str], str]
     model: Union[Unset, str] = "default/not-specified"
     encoding_format: Union[Unset, EmbeddingEncodingFormat] = UNSET
     user: Union[None, Unset, str] = UNSET
+    dimensions: Union[Unset, int] = 0
     modality: Union[Unset, OpenAIEmbeddingInputAudioModality] = OpenAIEmbeddingInputAudioModality.AUDIO
     additional_properties: Dict[str, Any] = _attrs_field(init=False, factory=dict)
 
@@ -52,6 +54,8 @@ def to_dict(self) -> Dict[str, Any]:
         else:
             user = self.user
 
+        dimensions = self.dimensions
+
         modality: Union[Unset, str] = UNSET
         if not isinstance(self.modality, Unset):
             modality = self.modality.value
@@ -69,6 +73,8 @@ def to_dict(self) -> Dict[str, Any]:
             field_dict["encoding_format"] = encoding_format
         if user is not UNSET:
             field_dict["user"] = user
+        if dimensions is not UNSET:
+            field_dict["dimensions"] = dimensions
         if modality is not UNSET:
             field_dict["modality"] = modality
 
@@ -118,6 +124,8 @@ def _parse_user(data: object) -> Union[None, Unset, str]:
 
         user = _parse_user(d.pop("user", UNSET))
 
+        dimensions = d.pop("dimensions", UNSET)
+
         _modality = d.pop("modality", UNSET)
         modality: Union[Unset, OpenAIEmbeddingInputAudioModality]
         if isinstance(_modality, Unset):
@@ -130,6 +138,7 @@ def _parse_user(data: object) -> Union[None, Unset, str]:
             model=model,
             encoding_format=encoding_format,
             user=user,
+            dimensions=dimensions,
             modality=modality,
         )
 

libs/client_infinity/infinity_client/infinity_client/models/open_ai_embedding_input_image.py

@@ -18,13 +18,15 @@ class OpenAIEmbeddingInputImage:
         model (Union[Unset, str]):  Default: 'default/not-specified'.
         encoding_format (Union[Unset, EmbeddingEncodingFormat]):
         user (Union[None, Unset, str]):
+        dimensions (Union[Unset, int]):  Default: 0.
         modality (Union[Unset, OpenAIEmbeddingInputImageModality]):  Default: OpenAIEmbeddingInputImageModality.IMAGE.
     """
 
     input_: Union[List[str], str]
     model: Union[Unset, str] = "default/not-specified"
     encoding_format: Union[Unset, EmbeddingEncodingFormat] = UNSET
     user: Union[None, Unset, str] = UNSET
+    dimensions: Union[Unset, int] = 0
     modality: Union[Unset, OpenAIEmbeddingInputImageModality] = OpenAIEmbeddingInputImageModality.IMAGE
     additional_properties: Dict[str, Any] = _attrs_field(init=False, factory=dict)
 
@@ -52,6 +54,8 @@ def to_dict(self) -> Dict[str, Any]:
         else:
             user = self.user
 
+        dimensions = self.dimensions
+
         modality: Union[Unset, str] = UNSET
         if not isinstance(self.modality, Unset):
             modality = self.modality.value
@@ -69,6 +73,8 @@ def to_dict(self) -> Dict[str, Any]:
             field_dict["encoding_format"] = encoding_format
         if user is not UNSET:
             field_dict["user"] = user
+        if dimensions is not UNSET:
+            field_dict["dimensions"] = dimensions
         if modality is not UNSET:
             field_dict["modality"] = modality
 
@@ -118,6 +124,8 @@ def _parse_user(data: object) -> Union[None, Unset, str]:
 
         user = _parse_user(d.pop("user", UNSET))
 
+        dimensions = d.pop("dimensions", UNSET)
+
         _modality = d.pop("modality", UNSET)
         modality: Union[Unset, OpenAIEmbeddingInputImageModality]
         if isinstance(_modality, Unset):
@@ -130,6 +138,7 @@ def _parse_user(data: object) -> Union[None, Unset, str]:
             model=model,
             encoding_format=encoding_format,
             user=user,
+            dimensions=dimensions,
             modality=modality,
         )
 

libs/client_infinity/infinity_client/infinity_client/models/open_ai_embedding_input_text.py

@@ -19,13 +19,15 @@ class OpenAIEmbeddingInputText:
         model (Union[Unset, str]):  Default: 'default/not-specified'.
         encoding_format (Union[Unset, EmbeddingEncodingFormat]):
         user (Union[None, Unset, str]):
+        dimensions (Union[Unset, int]):  Default: 0.
         modality (Union[Unset, OpenAIEmbeddingInputTextModality]):  Default: OpenAIEmbeddingInputTextModality.TEXT.
     """
 
     input_: Union[List[str], str]
     model: Union[Unset, str] = "default/not-specified"
     encoding_format: Union[Unset, EmbeddingEncodingFormat] = UNSET
     user: Union[None, Unset, str] = UNSET
+    dimensions: Union[Unset, int] = 0
     modality: Union[Unset, OpenAIEmbeddingInputTextModality] = OpenAIEmbeddingInputTextModality.TEXT
     additional_properties: Dict[str, Any] = _attrs_field(init=False, factory=dict)
 
@@ -49,6 +51,8 @@ def to_dict(self) -> Dict[str, Any]:
         else:
             user = self.user
 
+        dimensions = self.dimensions
+
         modality: Union[Unset, str] = UNSET
         if not isinstance(self.modality, Unset):
             modality = self.modality.value
@@ -66,6 +70,8 @@ def to_dict(self) -> Dict[str, Any]:
             field_dict["encoding_format"] = encoding_format
         if user is not UNSET:
             field_dict["user"] = user
+        if dimensions is not UNSET:
+            field_dict["dimensions"] = dimensions
         if modality is not UNSET:
             field_dict["modality"] = modality
 
@@ -106,6 +112,8 @@ def _parse_user(data: object) -> Union[None, Unset, str]:
 
         user = _parse_user(d.pop("user", UNSET))
 
+        dimensions = d.pop("dimensions", UNSET)
+
         _modality = d.pop("modality", UNSET)
         modality: Union[Unset, OpenAIEmbeddingInputTextModality]
         if isinstance(_modality, Unset):
@@ -118,6 +126,7 @@ def _parse_user(data: object) -> Union[None, Unset, str]:
             model=model,
             encoding_format=encoding_format,
             user=user,
+            dimensions=dimensions,
             modality=modality,
         )
 

libs/client_infinity/infinity_client/pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "infinity_client"
-version = "0.0.71"
+version = "0.0.73"
 description = "A client library for accessing ♾️ Infinity - Embedding Inference Server"
 authors = []
 readme = "README.md"

libs/infinity_emb/infinity_emb/engine.py

@@ -1,5 +1,6 @@
 # SPDX-License-Identifier: MIT
 # Copyright (c) 2023-now michaelfeilfeil
+from __future__ import annotations
 
 from asyncio import Semaphore
 from typing import Iterable, Iterator, Optional, Union
@@ -130,11 +131,14 @@ def capabilities(self) -> set[ModelCapabilites]:
     def engine_args(self) -> EngineArgs:
         return self._engine_args
 
-    async def embed(self, sentences: list[str]) -> tuple[list["EmbeddingReturnType"], int]:
+    async def embed(
+        self, sentences: list[str], matryoshka_dim: int | None = None
+    ) -> tuple[list["EmbeddingReturnType"], int]:
         """embed multiple sentences
 
         Kwargs:
             sentences (list[str]): sentences to be embedded
+            matryoshka_dim (int): Length of matryoshka embedding
 
         Raises:
             ValueError: raised if engine is not started yet
@@ -148,7 +152,9 @@ async def embed(self, sentences: list[str]) -> tuple[list["EmbeddingReturnType"]
         """
 
         self._assert_running()
-        embeddings, usage = await self._batch_handler.embed(sentences=sentences)
+        embeddings, usage = await self._batch_handler.embed(
+            sentences=sentences, matryoshka_dim=matryoshka_dim
+        )
         return embeddings, usage
 
     async def rerank(
@@ -213,12 +219,16 @@ async def classify(
         return scores, usage
 
     async def image_embed(
-        self, *, images: list[Union[str, "ImageClassType", bytes]]
+        self,
+        *,
+        images: list[Union[str, "ImageClassType", bytes]],
+        matryoshka_dim: int | None = None,
     ) -> tuple[list["EmbeddingReturnType"], int]:
         """embed multiple images
 
         Kwargs:
             images (list[Union[str, ImageClassType]]): list of image urls or ImageClassType objects, to be embedded
+            matryoshka_dim (int): Length of matryoshka embedding
 
         Raises:
             ValueError: raised if engine is not started yet
@@ -232,16 +242,19 @@ async def image_embed(
         """
 
         self._assert_running()
-        embeddings, usage = await self._batch_handler.image_embed(images=images)
+        embeddings, usage = await self._batch_handler.image_embed(
+            images=images, matryoshka_dim=matryoshka_dim
+        )
         return embeddings, usage
 
     async def audio_embed(
-        self, *, audios: list[Union[str, bytes]]
+        self, *, audios: list[Union[str, bytes]], matryoshka_dim: int | None = None
     ) -> tuple[list["EmbeddingReturnType"], int]:
         """embed multiple audios
 
         Kwargs:
             audios (list[Union[str, Audiobytes]]): list of audio data, to be embedded
+            matryoshka_dim (int): Length of matryoshka embedding
 
         Raises:
             ValueError: raised if engine is not started yet
@@ -255,7 +268,9 @@ async def audio_embed(
         """
 
         self._assert_running()
-        embeddings, usage = await self._batch_handler.audio_embed(audios=audios)
+        embeddings, usage = await self._batch_handler.audio_embed(
+            audios=audios, matryoshka_dim=matryoshka_dim
+        )
         return embeddings, usage
 
     def _assert_running(self):
@@ -304,13 +319,14 @@ async def astop(self):
             await engine.astop()
 
     async def embed(
-        self, *, model: str, sentences: list[str]
+        self, *, model: str, sentences: list[str], matryoshka_dim: Optional[int] = None
     ) -> tuple[list["EmbeddingReturnType"], int]:
         """embed multiple sentences
 
         Kwargs:
             model (str): model name to be used
             sentences (list[str]): sentences to be embedded
+            matryoshka_dim (int): Length of matryoshka embedding
 
         Raises:
             ValueError: raised if engine is not started yet
@@ -322,7 +338,7 @@ async def embed(
                 2D list-array of shape( len(sentences),embed_dim )
             int: token usage
         """
-        return await self[model].embed(sentences)
+        return await self[model].embed(sentences, matryoshka_dim=matryoshka_dim)
 
     def is_running(self) -> bool:
         return all(engine.is_running for engine in self.engines_dict.values())
@@ -378,13 +394,18 @@ async def classify(
         return await self[model].classify(sentences=sentences, raw_scores=raw_scores)
 
     async def image_embed(
-        self, *, model: str, images: list[Union[str, "ImageClassType"]]
+        self,
+        *,
+        model: str,
+        images: list[Union[str, "ImageClassType"]],
+        matryoshka_dim: Optional[int] = None,
     ) -> tuple[list["EmbeddingReturnType"], int]:
         """embed multiple images
 
         Kwargs:
             model (str): model name to be used
             images (list[Union[str, ImageClassType]]): list of image urls or ImageClassType objects, to be embedded
+            matryoshka_dim (int): Length of matryoshka embedding
 
         Raises:
             ValueError: raised if engine is not started yet
@@ -396,7 +417,7 @@ async def image_embed(
                 2D list-array of shape( len(sentences),embed_dim )
             int: token usage
         """
-        return await self[model].image_embed(images=images)
+        return await self[model].image_embed(images=images, matryoshka_dim=matryoshka_dim)
 
     def __getitem__(self, index_or_name: Union[str, int]) -> "AsyncEmbeddingEngine":
         """resolve engine by model name -> Auto resolve if only one engine is present
@@ -416,13 +437,14 @@ def __getitem__(self, index_or_name: Union[str, int]) -> "AsyncEmbeddingEngine":
         )
 
     async def audio_embed(
-        self, *, model: str, audios: list[Union[str, bytes]]
+        self, *, model: str, audios: list[Union[str, bytes]], matryoshka_dim: Optional[int] = None
     ) -> tuple[list["EmbeddingReturnType"], int]:
         """embed multiple audios
 
         Kwargs:
             model (str): model name to be used
             audios (list[Union[str, bytes]]): list of audio data, to be embedded
+            matryoshka_dim (int): Length of matryoshka embedding
 
         Raises:
             ValueError: raised if engine is not started yet
@@ -434,4 +456,4 @@ async def audio_embed(
                 2D list-array of shape( len(sentences),embed_dim )
             int: token usage
         """
-        return await self[model].audio_embed(audios=audios)
+        return await self[model].audio_embed(audios=audios, matryoshka_dim=matryoshka_dim)

libs/infinity_emb/infinity_emb/env.py

@@ -189,7 +189,15 @@ def cache_dir(self) -> Path:
 
     @cached_property
     def queue_size(self) -> int:
-        return int(self._optional_infinity_var("queue_size", default="32000"))
+        size = int(self._optional_infinity_var("queue_size", default="32000"))
+        assert size > 0, "INFINITY_QUEUE_SIZE must be a positive number"
+        return size
+
+    @cached_property
+    def max_client_batch_size(self) -> int:
+        size = int(self._optional_infinity_var("max_client_batch_size", default="2048"))
+        assert size > 0, "INFINITY_MAX_CLIENT_BATCH_SIZE must be a positive number"
+        return size
 
     @cached_property
     def permissive_cors(self):

libs/infinity_emb/infinity_emb/fastapi_schemas/docs.py

@@ -17,7 +17,9 @@ def startup_message(host: str, port: int, prefix: str) -> str:
     return f"""
 
 ♾️  Infinity - Embedding Inference Server
-MIT License; Copyright (c) 2023-now Michael Feil
+MIT License; Copyright (c) 2023-now 
+Infinity OSS-Project: github.com/michaelfeil.infinity
+Maintained by @michaelfeil @wirthual
 Version {__version__}
 
 Open the Docs via Swagger UI:

libs/infinity_emb/infinity_emb/fastapi_schemas/pydantic_v2.py

@@ -1,4 +1,5 @@
 from pydantic import AnyUrl, HttpUrl, StringConstraints
+from infinity_emb.env import MANAGER
 
 __all__ = [
     "INPUT_STRING",
@@ -14,9 +15,9 @@
 INPUT_STRING = StringConstraints(max_length=8192 * 15, strip_whitespace=True)
 ITEMS_LIMIT = {
     "min_length": 1,
-    "max_length": 2048,
+    "max_length": MANAGER.max_client_batch_size,
 }
 ITEMS_LIMIT_SMALL = {
     "min_length": 1,
-    "max_length": 32,
+    "max_length": min(32, MANAGER.max_client_batch_size),
 }

libs/infinity_emb/infinity_emb/fastapi_schemas/pymodels.py

@@ -54,6 +54,7 @@ class _OpenAIEmbeddingInput(BaseModel):
     model: str = "default/not-specified"
     encoding_format: EmbeddingEncodingFormat = EmbeddingEncodingFormat.float
     user: Optional[str] = None
+    dimensions: int = 0
 
 
 class _OpenAIEmbeddingInput_Text(_OpenAIEmbeddingInput):