phoenix-oss/llama-stack-mirror
0
0
Fork 1
mirror of https://github.com/meta-llama/llama-stack.git synced 2025-12-12 04:00:42 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

feat(api)!: support passing extra_body to embeddings and vector_stores APIs

Browse source 
Applies the same pattern from #3777 to embeddings and vector_stores.create() endpoints.

Breaking change: Method signatures now accept a single params object with Pydantic extra="allow" instead of individual parameters. Provider-specific params can be passed via extra_body and accessed through params.model_extra.

Updated APIs: openai_embeddings(), openai_create_vector_store(), openai_create_vector_store_file_batch()
This commit is contained in:
Ashwin Bharambe 2025-10-11 15:27:47 -07:00
parent cfd2e303db
commit 74e2976c1e
20 changed files with 364 additions and 297 deletions
Download patch file Download diff file Expand all files Collapse all files

49
docs/static/deprecated-llama-stack-spec.html vendored
View file

@ -1662,7 +1662,7 @@
"content": {
"application/json": {
"schema": {
"$ref": "#/components/schemas/OpenaiEmbeddingsRequest"
"$ref": "#/components/schemas/OpenAIEmbeddingsRequestWithExtraBody"
}
}
},
@ -2436,13 +2436,13 @@
"VectorIO"
],
"summary": "Creates a vector store.",
"description": "Creates a vector store.",
"description": "Creates a vector store.\nGenerate an OpenAI-compatible vector store with the given parameters.",
"parameters": [],
"requestBody": {
"content": {
"application/json": {
"schema": {
"$ref": "#/components/schemas/OpenaiCreateVectorStoreRequest"
"$ref": "#/components/schemas/OpenAICreateVectorStoreRequestWithExtraBody"
}
}
},
@ -2622,7 +2622,7 @@
"VectorIO"
],
"summary": "Create a vector store file batch.",
"description": "Create a vector store file batch.",
"description": "Create a vector store file batch.\nGenerate an OpenAI-compatible vector store file batch for the given vector store.",
"parameters": [
{
"name": "vector_store_id",
@ -2638,7 +2638,7 @@
"content": {
"application/json": {
"schema": {
"$ref": "#/components/schemas/OpenaiCreateVectorStoreFileBatchRequest"
"$ref": "#/components/schemas/OpenAICreateVectorStoreFileBatchRequestWithExtraBody"
}
}
},
@ -8174,7 +8174,7 @@
"title": "OpenAICompletionChoice",
"description": "A choice from an OpenAI-compatible completion response."
},
"OpenaiEmbeddingsRequest": {
"OpenAIEmbeddingsRequestWithExtraBody": {
"type": "object",
"properties": {
"model": {
@ -8197,6 +8197,7 @@
},
"encoding_format": {
"type": "string",
"default": "float",
"description": "(Optional) The format to return the embeddings in. Can be either \"float\" or \"base64\". Defaults to \"float\"."
},
"dimensions": {
@ -8213,7 +8214,8 @@
"model",
"input"
],
"title": "OpenaiEmbeddingsRequest"
"title": "OpenAIEmbeddingsRequestWithExtraBody",
"description": "Request parameters for OpenAI-compatible embeddings endpoint."
},
"OpenAIEmbeddingData": {
"type": "object",
@ -12061,19 +12063,19 @@
"title": "VectorStoreObject",
"description": "OpenAI Vector Store object."
},
"OpenaiCreateVectorStoreRequest": {
"OpenAICreateVectorStoreRequestWithExtraBody": {
"type": "object",
"properties": {
"name": {
"type": "string",
"description": "A name for the vector store."
"description": "(Optional) A name for the vector store"
},
"file_ids": {
"type": "array",
"items": {
"type": "string"
},
"description": "A list of File IDs that the vector store should use. Useful for tools like `file_search` that can access files."
"description": "List of file IDs to include in the vector store"
},
"expires_after": {
"type": "object",
@ -12099,7 +12101,7 @@
}
]
},
"description": "The expiration policy for a vector store."
"description": "(Optional) Expiration policy for the vector store"
},
"chunking_strategy": {
"type": "object",
@ -12125,7 +12127,7 @@
}
]
},
"description": "The chunking strategy used to chunk the file(s). If not set, will use the `auto` strategy."
"description": "(Optional) Strategy for splitting files into chunks"
},
"metadata": {
"type": "object",
@ -12151,23 +12153,25 @@
}
]
},
"description": "Set of 16 key-value pairs that can be attached to an object."
"description": "Set of key-value pairs that can be attached to the vector store"
},
"embedding_model": {
"type": "string",
"description": "The embedding model to use for this vector store."
"description": "(Optional) The embedding model to use for this vector store"
},
"embedding_dimension": {
"type": "integer",
"description": "The dimension of the embedding vectors (default: 384)."
"default": 384,
"description": "(Optional) The dimension of the embedding vectors (default: 384)"
},
"provider_id": {
"type": "string",
"description": "The ID of the provider to use for this vector store."
"description": "(Optional) The ID of the provider to use for this vector store"
}
},
"additionalProperties": false,
"title": "OpenaiCreateVectorStoreRequest"
"title": "OpenAICreateVectorStoreRequestWithExtraBody",
"description": "Request to create a vector store with extra_body support."
},
"OpenaiUpdateVectorStoreRequest": {
"type": "object",
@ -12337,7 +12341,7 @@
"title": "VectorStoreChunkingStrategyStaticConfig",
"description": "Configuration for static chunking strategy."
},
"OpenaiCreateVectorStoreFileBatchRequest": {
"OpenAICreateVectorStoreFileBatchRequestWithExtraBody": {
"type": "object",
"properties": {
"file_ids": {
@ -12345,7 +12349,7 @@
"items": {
"type": "string"
},
"description": "A list of File IDs that the vector store should use."
"description": "A list of File IDs that the vector store should use"
},
"attributes": {
"type": "object",
@ -12371,18 +12375,19 @@
}
]
},
"description": "(Optional) Key-value attributes to store with the files."
"description": "(Optional) Key-value attributes to store with the files"
},
"chunking_strategy": {
"$ref": "#/components/schemas/VectorStoreChunkingStrategy",
"description": "(Optional) The chunking strategy used to chunk the file(s). Defaults to auto."
"description": "(Optional) The chunking strategy used to chunk the file(s). Defaults to auto"
}
},
"additionalProperties": false,
"required": [
"file_ids"
],
"title": "OpenaiCreateVectorStoreFileBatchRequest"
"title": "OpenAICreateVectorStoreFileBatchRequestWithExtraBody",
"description": "Request to create a vector store file batch with extra_body support."
},
"VectorStoreFileBatchObject": {
"type": "object",

63
docs/static/deprecated-llama-stack-spec.yaml vendored
View file

@ -1203,7 +1203,7 @@ paths:
content:
application/json:
schema:
$ref: '#/components/schemas/OpenaiEmbeddingsRequest'
$ref: '#/components/schemas/OpenAIEmbeddingsRequestWithExtraBody'
required: true
deprecated: true
/v1/openai/v1/files:
@ -1792,13 +1792,16 @@ paths:
tags:
- VectorIO
summary: Creates a vector store.
description: Creates a vector store.
description: >-
Creates a vector store.
Generate an OpenAI-compatible vector store with the given parameters.
parameters: []
requestBody:
content:
application/json:
schema:
$ref: '#/components/schemas/OpenaiCreateVectorStoreRequest'
$ref: '#/components/schemas/OpenAICreateVectorStoreRequestWithExtraBody'
required: true
deprecated: true
/v1/openai/v1/vector_stores/{vector_store_id}:
@ -1924,7 +1927,11 @@ paths:
tags:
- VectorIO
summary: Create a vector store file batch.
description: Create a vector store file batch.
description: >-
Create a vector store file batch.
Generate an OpenAI-compatible vector store file batch for the given vector
store.
parameters:
- name: vector_store_id
in: path
@ -1937,7 +1944,7 @@ paths:
content:
application/json:
schema:
$ref: '#/components/schemas/OpenaiCreateVectorStoreFileBatchRequest'
$ref: '#/components/schemas/OpenAICreateVectorStoreFileBatchRequestWithExtraBody'
required: true
deprecated: true
/v1/openai/v1/vector_stores/{vector_store_id}/file_batches/{batch_id}:
@ -6035,7 +6042,7 @@ components:
title: OpenAICompletionChoice
description: >-
A choice from an OpenAI-compatible completion response.
OpenaiEmbeddingsRequest:
OpenAIEmbeddingsRequestWithExtraBody:
type: object
properties:
model:
@ -6054,6 +6061,7 @@ components:
multiple inputs in a single request, pass an array of strings.
encoding_format:
type: string
default: float
description: >-
(Optional) The format to return the embeddings in. Can be either "float"
or "base64". Defaults to "float".
@ -6071,7 +6079,9 @@ components:
required:
- model
- input
title: OpenaiEmbeddingsRequest
title: OpenAIEmbeddingsRequestWithExtraBody
description: >-
Request parameters for OpenAI-compatible embeddings endpoint.
OpenAIEmbeddingData:
type: object
properties:
@ -9147,19 +9157,18 @@ components:
- metadata
title: VectorStoreObject
description: OpenAI Vector Store object.
OpenaiCreateVectorStoreRequest:
"OpenAICreateVectorStoreRequestWithExtraBody":
type: object
properties:
name:
type: string
description: A name for the vector store.
description: (Optional) A name for the vector store
file_ids:
type: array
items:
type: string
description: >-
A list of File IDs that the vector store should use. Useful for tools
like `file_search` that can access files.
List of file IDs to include in the vector store
expires_after:
type: object
additionalProperties:
@ -9171,7 +9180,7 @@ components:
- type: array
- type: object
description: >-
The expiration policy for a vector store.
(Optional) Expiration policy for the vector store
chunking_strategy:
type: object
additionalProperties:
@ -9183,8 +9192,7 @@ components:
- type: array
- type: object
description: >-
The chunking strategy used to chunk the file(s). If not set, will use
the `auto` strategy.
(Optional) Strategy for splitting files into chunks
metadata:
type: object
additionalProperties:
@ -9196,21 +9204,25 @@ components:
- type: array
- type: object
description: >-
Set of 16 key-value pairs that can be attached to an object.
Set of key-value pairs that can be attached to the vector store
embedding_model:
type: string
description: >-
The embedding model to use for this vector store.
(Optional) The embedding model to use for this vector store
embedding_dimension:
type: integer
default: 384
description: >-
The dimension of the embedding vectors (default: 384).
(Optional) The dimension of the embedding vectors (default: 384)
provider_id:
type: string
description: >-
The ID of the provider to use for this vector store.
(Optional) The ID of the provider to use for this vector store
additionalProperties: false
title: OpenaiCreateVectorStoreRequest
title: >-
OpenAICreateVectorStoreRequestWithExtraBody
description: >-
Request to create a vector store with extra_body support.
OpenaiUpdateVectorStoreRequest:
type: object
properties:
@ -9331,7 +9343,7 @@ components:
title: VectorStoreChunkingStrategyStaticConfig
description: >-
Configuration for static chunking strategy.
OpenaiCreateVectorStoreFileBatchRequest:
"OpenAICreateVectorStoreFileBatchRequestWithExtraBody":
type: object
properties:
file_ids:
@ -9339,7 +9351,7 @@ components:
items:
type: string
description: >-
A list of File IDs that the vector store should use.
A list of File IDs that the vector store should use
attributes:
type: object
additionalProperties:
@ -9351,16 +9363,19 @@ components:
- type: array
- type: object
description: >-
(Optional) Key-value attributes to store with the files.
(Optional) Key-value attributes to store with the files
chunking_strategy:
$ref: '#/components/schemas/VectorStoreChunkingStrategy'
description: >-
(Optional) The chunking strategy used to chunk the file(s). Defaults to
auto.
auto
additionalProperties: false
required:
- file_ids
title: OpenaiCreateVectorStoreFileBatchRequest
title: >-
OpenAICreateVectorStoreFileBatchRequestWithExtraBody
description: >-
Request to create a vector store file batch with extra_body support.
VectorStoreFileBatchObject:
type: object
properties:

49
docs/static/llama-stack-spec.html vendored
View file

@ -765,7 +765,7 @@
"content": {
"application/json": {
"schema": {
"$ref": "#/components/schemas/OpenaiEmbeddingsRequest"
"$ref": "#/components/schemas/OpenAIEmbeddingsRequestWithExtraBody"
}
}
},
@ -3170,13 +3170,13 @@
"VectorIO"
],
"summary": "Creates a vector store.",
"description": "Creates a vector store.",
"description": "Creates a vector store.\nGenerate an OpenAI-compatible vector store with the given parameters.",
"parameters": [],
"requestBody": {
"content": {
"application/json": {
"schema": {
"$ref": "#/components/schemas/OpenaiCreateVectorStoreRequest"
"$ref": "#/components/schemas/OpenAICreateVectorStoreRequestWithExtraBody"
}
}
},
@ -3356,7 +3356,7 @@
"VectorIO"
],
"summary": "Create a vector store file batch.",
"description": "Create a vector store file batch.",
"description": "Create a vector store file batch.\nGenerate an OpenAI-compatible vector store file batch for the given vector store.",
"parameters": [
{
"name": "vector_store_id",
@ -3372,7 +3372,7 @@
"content": {
"application/json": {
"schema": {
"$ref": "#/components/schemas/OpenaiCreateVectorStoreFileBatchRequest"
"$ref": "#/components/schemas/OpenAICreateVectorStoreFileBatchRequestWithExtraBody"
}
}
},
@ -6324,7 +6324,7 @@
"title": "ConversationItemDeletedResource",
"description": "Response for deleted conversation item."
},
"OpenaiEmbeddingsRequest": {
"OpenAIEmbeddingsRequestWithExtraBody": {
"type": "object",
"properties": {
"model": {
@ -6347,6 +6347,7 @@
},
"encoding_format": {
"type": "string",
"default": "float",
"description": "(Optional) The format to return the embeddings in. Can be either \"float\" or \"base64\". Defaults to \"float\"."
},
"dimensions": {
@ -6363,7 +6364,8 @@
"model",
"input"
],
"title": "OpenaiEmbeddingsRequest"
"title": "OpenAIEmbeddingsRequestWithExtraBody",
"description": "Request parameters for OpenAI-compatible embeddings endpoint."
},
"OpenAIEmbeddingData": {
"type": "object",
@ -12587,19 +12589,19 @@
"title": "VectorStoreObject",
"description": "OpenAI Vector Store object."
},
"OpenaiCreateVectorStoreRequest": {
"OpenAICreateVectorStoreRequestWithExtraBody": {
"type": "object",
"properties": {
"name": {
"type": "string",
"description": "A name for the vector store."
"description": "(Optional) A name for the vector store"
},
"file_ids": {
"type": "array",
"items": {
"type": "string"
},
"description": "A list of File IDs that the vector store should use. Useful for tools like `file_search` that can access files."
"description": "List of file IDs to include in the vector store"
},
"expires_after": {
"type": "object",
@ -12625,7 +12627,7 @@
}
]
},
"description": "The expiration policy for a vector store."
"description": "(Optional) Expiration policy for the vector store"
},
"chunking_strategy": {
"type": "object",
@ -12651,7 +12653,7 @@
}
]
},
"description": "The chunking strategy used to chunk the file(s). If not set, will use the `auto` strategy."
"description": "(Optional) Strategy for splitting files into chunks"
},
"metadata": {
"type": "object",
@ -12677,23 +12679,25 @@
}
]
},
"description": "Set of 16 key-value pairs that can be attached to an object."
"description": "Set of key-value pairs that can be attached to the vector store"
},
"embedding_model": {
"type": "string",
"description": "The embedding model to use for this vector store."
"description": "(Optional) The embedding model to use for this vector store"
},
"embedding_dimension": {
"type": "integer",
"description": "The dimension of the embedding vectors (default: 384)."
"default": 384,
"description": "(Optional) The dimension of the embedding vectors (default: 384)"
},
"provider_id": {
"type": "string",
"description": "The ID of the provider to use for this vector store."
"description": "(Optional) The ID of the provider to use for this vector store"
}
},
"additionalProperties": false,
"title": "OpenaiCreateVectorStoreRequest"
"title": "OpenAICreateVectorStoreRequestWithExtraBody",
"description": "Request to create a vector store with extra_body support."
},
"OpenaiUpdateVectorStoreRequest": {
"type": "object",
@ -12863,7 +12867,7 @@
"title": "VectorStoreChunkingStrategyStaticConfig",
"description": "Configuration for static chunking strategy."
},
"OpenaiCreateVectorStoreFileBatchRequest": {
"OpenAICreateVectorStoreFileBatchRequestWithExtraBody": {
"type": "object",
"properties": {
"file_ids": {
@ -12871,7 +12875,7 @@
"items": {
"type": "string"
},
"description": "A list of File IDs that the vector store should use."
"description": "A list of File IDs that the vector store should use"
},
"attributes": {
"type": "object",
@ -12897,18 +12901,19 @@
}
]
},
"description": "(Optional) Key-value attributes to store with the files."
"description": "(Optional) Key-value attributes to store with the files"
},
"chunking_strategy": {
"$ref": "#/components/schemas/VectorStoreChunkingStrategy",
"description": "(Optional) The chunking strategy used to chunk the file(s). Defaults to auto."
"description": "(Optional) The chunking strategy used to chunk the file(s). Defaults to auto"
}
},
"additionalProperties": false,
"required": [
"file_ids"
],
"title": "OpenaiCreateVectorStoreFileBatchRequest"
"title": "OpenAICreateVectorStoreFileBatchRequestWithExtraBody",
"description": "Request to create a vector store file batch with extra_body support."
},
"VectorStoreFileBatchObject": {
"type": "object",

63
docs/static/llama-stack-spec.yaml vendored
View file

@ -617,7 +617,7 @@ paths:
content:
application/json:
schema:
$ref: '#/components/schemas/OpenaiEmbeddingsRequest'
$ref: '#/components/schemas/OpenAIEmbeddingsRequestWithExtraBody'
required: true
deprecated: false
/v1/files:
@ -2413,13 +2413,16 @@ paths:
tags:
- VectorIO
summary: Creates a vector store.
description: Creates a vector store.
description: >-
Creates a vector store.
Generate an OpenAI-compatible vector store with the given parameters.
parameters: []
requestBody:
content:
application/json:
schema:
$ref: '#/components/schemas/OpenaiCreateVectorStoreRequest'
$ref: '#/components/schemas/OpenAICreateVectorStoreRequestWithExtraBody'
required: true
deprecated: false
/v1/vector_stores/{vector_store_id}:
@ -2545,7 +2548,11 @@ paths:
tags:
- VectorIO
summary: Create a vector store file batch.
description: Create a vector store file batch.
description: >-
Create a vector store file batch.
Generate an OpenAI-compatible vector store file batch for the given vector
store.
parameters:
- name: vector_store_id
in: path
@ -2558,7 +2565,7 @@ paths:
content:
application/json:
schema:
$ref: '#/components/schemas/OpenaiCreateVectorStoreFileBatchRequest'
$ref: '#/components/schemas/OpenAICreateVectorStoreFileBatchRequestWithExtraBody'
required: true
deprecated: false
/v1/vector_stores/{vector_store_id}/file_batches/{batch_id}:
@ -4797,7 +4804,7 @@ components:
- deleted
title: ConversationItemDeletedResource
description: Response for deleted conversation item.
OpenaiEmbeddingsRequest:
OpenAIEmbeddingsRequestWithExtraBody:
type: object
properties:
model:
@ -4816,6 +4823,7 @@ components:
multiple inputs in a single request, pass an array of strings.
encoding_format:
type: string
default: float
description: >-
(Optional) The format to return the embeddings in. Can be either "float"
or "base64". Defaults to "float".
@ -4833,7 +4841,9 @@ components:
required:
- model
- input
title: OpenaiEmbeddingsRequest
title: OpenAIEmbeddingsRequestWithExtraBody
description: >-
Request parameters for OpenAI-compatible embeddings endpoint.
OpenAIEmbeddingData:
type: object
properties:
@ -9612,19 +9622,18 @@ components:
- metadata
title: VectorStoreObject
description: OpenAI Vector Store object.
OpenaiCreateVectorStoreRequest:
"OpenAICreateVectorStoreRequestWithExtraBody":
type: object
properties:
name:
type: string
description: A name for the vector store.
description: (Optional) A name for the vector store
file_ids:
type: array
items:
type: string
description: >-
A list of File IDs that the vector store should use. Useful for tools
like `file_search` that can access files.
List of file IDs to include in the vector store
expires_after:
type: object
additionalProperties:
@ -9636,7 +9645,7 @@ components:
- type: array
- type: object
description: >-
The expiration policy for a vector store.
(Optional) Expiration policy for the vector store
chunking_strategy:
type: object
additionalProperties:
@ -9648,8 +9657,7 @@ components:
- type: array
- type: object
description: >-
The chunking strategy used to chunk the file(s). If not set, will use
the `auto` strategy.
(Optional) Strategy for splitting files into chunks
metadata:
type: object
additionalProperties:
@ -9661,21 +9669,25 @@ components:
- type: array
- type: object
description: >-
Set of 16 key-value pairs that can be attached to an object.
Set of key-value pairs that can be attached to the vector store
embedding_model:
type: string
description: >-
The embedding model to use for this vector store.
(Optional) The embedding model to use for this vector store
embedding_dimension:
type: integer
default: 384
description: >-
The dimension of the embedding vectors (default: 384).
(Optional) The dimension of the embedding vectors (default: 384)
provider_id:
type: string
description: >-
The ID of the provider to use for this vector store.
(Optional) The ID of the provider to use for this vector store
additionalProperties: false
title: OpenaiCreateVectorStoreRequest
title: >-
OpenAICreateVectorStoreRequestWithExtraBody
description: >-
Request to create a vector store with extra_body support.
OpenaiUpdateVectorStoreRequest:
type: object
properties:
@ -9796,7 +9808,7 @@ components:
title: VectorStoreChunkingStrategyStaticConfig
description: >-
Configuration for static chunking strategy.
OpenaiCreateVectorStoreFileBatchRequest:
"OpenAICreateVectorStoreFileBatchRequestWithExtraBody":
type: object
properties:
file_ids:
@ -9804,7 +9816,7 @@ components:
items:
type: string
description: >-
A list of File IDs that the vector store should use.
A list of File IDs that the vector store should use
attributes:
type: object
additionalProperties:
@ -9816,16 +9828,19 @@ components:
- type: array
- type: object
description: >-
(Optional) Key-value attributes to store with the files.
(Optional) Key-value attributes to store with the files
chunking_strategy:
$ref: '#/components/schemas/VectorStoreChunkingStrategy'
description: >-
(Optional) The chunking strategy used to chunk the file(s). Defaults to
auto.
auto
additionalProperties: false
required:
- file_ids
title: OpenaiCreateVectorStoreFileBatchRequest
title: >-
OpenAICreateVectorStoreFileBatchRequestWithExtraBody
description: >-
Request to create a vector store file batch with extra_body support.
VectorStoreFileBatchObject:
type: object
properties:

49
docs/static/stainless-llama-stack-spec.html vendored
View file

@ -765,7 +765,7 @@
"content": {
"application/json": {
"schema": {
"$ref": "#/components/schemas/OpenaiEmbeddingsRequest"
"$ref": "#/components/schemas/OpenAIEmbeddingsRequestWithExtraBody"
}
}
},
@ -3170,13 +3170,13 @@
"VectorIO"
],
"summary": "Creates a vector store.",
"description": "Creates a vector store.",
"description": "Creates a vector store.\nGenerate an OpenAI-compatible vector store with the given parameters.",
"parameters": [],
"requestBody": {
"content": {
"application/json": {
"schema": {
"$ref": "#/components/schemas/OpenaiCreateVectorStoreRequest"
"$ref": "#/components/schemas/OpenAICreateVectorStoreRequestWithExtraBody"
}
}
},
@ -3356,7 +3356,7 @@
"VectorIO"
],
"summary": "Create a vector store file batch.",
"description": "Create a vector store file batch.",
"description": "Create a vector store file batch.\nGenerate an OpenAI-compatible vector store file batch for the given vector store.",
"parameters": [
{
"name": "vector_store_id",
@ -3372,7 +3372,7 @@
"content": {
"application/json": {
"schema": {
"$ref": "#/components/schemas/OpenaiCreateVectorStoreFileBatchRequest"
"$ref": "#/components/schemas/OpenAICreateVectorStoreFileBatchRequestWithExtraBody"
}
}
},
@ -8333,7 +8333,7 @@
"title": "ConversationItemDeletedResource",
"description": "Response for deleted conversation item."
},
"OpenaiEmbeddingsRequest": {
"OpenAIEmbeddingsRequestWithExtraBody": {
"type": "object",
"properties": {
"model": {
@ -8356,6 +8356,7 @@
},
"encoding_format": {
"type": "string",
"default": "float",
"description": "(Optional) The format to return the embeddings in. Can be either \"float\" or \"base64\". Defaults to \"float\"."
},
"dimensions": {
@ -8372,7 +8373,8 @@
"model",
"input"
],
"title": "OpenaiEmbeddingsRequest"
"title": "OpenAIEmbeddingsRequestWithExtraBody",
"description": "Request parameters for OpenAI-compatible embeddings endpoint."
},
"OpenAIEmbeddingData": {
"type": "object",
@ -14596,19 +14598,19 @@
"title": "VectorStoreObject",
"description": "OpenAI Vector Store object."
},
"OpenaiCreateVectorStoreRequest": {
"OpenAICreateVectorStoreRequestWithExtraBody": {
"type": "object",
"properties": {
"name": {
"type": "string",
"description": "A name for the vector store."
"description": "(Optional) A name for the vector store"
},
"file_ids": {
"type": "array",
"items": {
"type": "string"
},
"description": "A list of File IDs that the vector store should use. Useful for tools like `file_search` that can access files."
"description": "List of file IDs to include in the vector store"
},
"expires_after": {
"type": "object",
@ -14634,7 +14636,7 @@
}
]
},
"description": "The expiration policy for a vector store."
"description": "(Optional) Expiration policy for the vector store"
},
"chunking_strategy": {
"type": "object",
@ -14660,7 +14662,7 @@
}
]
},
"description": "The chunking strategy used to chunk the file(s). If not set, will use the `auto` strategy."
"description": "(Optional) Strategy for splitting files into chunks"
},
"metadata": {
"type": "object",
@ -14686,23 +14688,25 @@
}
]
},
"description": "Set of 16 key-value pairs that can be attached to an object."
"description": "Set of key-value pairs that can be attached to the vector store"
},
"embedding_model": {
"type": "string",
"description": "The embedding model to use for this vector store."
"description": "(Optional) The embedding model to use for this vector store"
},
"embedding_dimension": {
"type": "integer",
"description": "The dimension of the embedding vectors (default: 384)."
"default": 384,
"description": "(Optional) The dimension of the embedding vectors (default: 384)"
},
"provider_id": {
"type": "string",
"description": "The ID of the provider to use for this vector store."
"description": "(Optional) The ID of the provider to use for this vector store"
}
},
"additionalProperties": false,
"title": "OpenaiCreateVectorStoreRequest"
"title": "OpenAICreateVectorStoreRequestWithExtraBody",
"description": "Request to create a vector store with extra_body support."
},
"OpenaiUpdateVectorStoreRequest": {
"type": "object",
@ -14872,7 +14876,7 @@
"title": "VectorStoreChunkingStrategyStaticConfig",
"description": "Configuration for static chunking strategy."
},
"OpenaiCreateVectorStoreFileBatchRequest": {
"OpenAICreateVectorStoreFileBatchRequestWithExtraBody": {
"type": "object",
"properties": {
"file_ids": {
@ -14880,7 +14884,7 @@
"items": {
"type": "string"
},
"description": "A list of File IDs that the vector store should use."
"description": "A list of File IDs that the vector store should use"
},
"attributes": {
"type": "object",
@ -14906,18 +14910,19 @@
}
]
},
"description": "(Optional) Key-value attributes to store with the files."
"description": "(Optional) Key-value attributes to store with the files"
},
"chunking_strategy": {
"$ref": "#/components/schemas/VectorStoreChunkingStrategy",
"description": "(Optional) The chunking strategy used to chunk the file(s). Defaults to auto."
"description": "(Optional) The chunking strategy used to chunk the file(s). Defaults to auto"
}
},
"additionalProperties": false,
"required": [
"file_ids"
],
"title": "OpenaiCreateVectorStoreFileBatchRequest"
"title": "OpenAICreateVectorStoreFileBatchRequestWithExtraBody",
"description": "Request to create a vector store file batch with extra_body support."
},
"VectorStoreFileBatchObject": {
"type": "object",

63
docs/static/stainless-llama-stack-spec.yaml vendored
View file

@ -620,7 +620,7 @@ paths:
content:
application/json:
schema:
$ref: '#/components/schemas/OpenaiEmbeddingsRequest'
$ref: '#/components/schemas/OpenAIEmbeddingsRequestWithExtraBody'
required: true
deprecated: false
/v1/files:
@ -2416,13 +2416,16 @@ paths:
tags:
- VectorIO
summary: Creates a vector store.
description: Creates a vector store.
description: >-
Creates a vector store.
Generate an OpenAI-compatible vector store with the given parameters.
parameters: []
requestBody:
content:
application/json:
schema:
$ref: '#/components/schemas/OpenaiCreateVectorStoreRequest'
$ref: '#/components/schemas/OpenAICreateVectorStoreRequestWithExtraBody'
required: true
deprecated: false
/v1/vector_stores/{vector_store_id}:
@ -2548,7 +2551,11 @@ paths:
tags:
- VectorIO
summary: Create a vector store file batch.
description: Create a vector store file batch.
description: >-
Create a vector store file batch.
Generate an OpenAI-compatible vector store file batch for the given vector
store.
parameters:
- name: vector_store_id
in: path
@ -2561,7 +2568,7 @@ paths:
content:
application/json:
schema:
$ref: '#/components/schemas/OpenaiCreateVectorStoreFileBatchRequest'
$ref: '#/components/schemas/OpenAICreateVectorStoreFileBatchRequestWithExtraBody'
required: true
deprecated: false
/v1/vector_stores/{vector_store_id}/file_batches/{batch_id}:
@ -6242,7 +6249,7 @@ components:
- deleted
title: ConversationItemDeletedResource
description: Response for deleted conversation item.
OpenaiEmbeddingsRequest:
OpenAIEmbeddingsRequestWithExtraBody:
type: object
properties:
model:
@ -6261,6 +6268,7 @@ components:
multiple inputs in a single request, pass an array of strings.
encoding_format:
type: string
default: float
description: >-
(Optional) The format to return the embeddings in. Can be either "float"
or "base64". Defaults to "float".
@ -6278,7 +6286,9 @@ components:
required:
- model
- input
title: OpenaiEmbeddingsRequest
title: OpenAIEmbeddingsRequestWithExtraBody
description: >-
Request parameters for OpenAI-compatible embeddings endpoint.
OpenAIEmbeddingData:
type: object
properties:
@ -11057,19 +11067,18 @@ components:
- metadata
title: VectorStoreObject
description: OpenAI Vector Store object.
OpenaiCreateVectorStoreRequest:
"OpenAICreateVectorStoreRequestWithExtraBody":
type: object
properties:
name:
type: string
description: A name for the vector store.
description: (Optional) A name for the vector store
file_ids:
type: array
items:
type: string
description: >-
A list of File IDs that the vector store should use. Useful for tools
like `file_search` that can access files.
List of file IDs to include in the vector store
expires_after:
type: object
additionalProperties:
@ -11081,7 +11090,7 @@ components:
- type: array
- type: object
description: >-
The expiration policy for a vector store.
(Optional) Expiration policy for the vector store
chunking_strategy:
type: object
additionalProperties:
@ -11093,8 +11102,7 @@ components:
- type: array
- type: object
description: >-
The chunking strategy used to chunk the file(s). If not set, will use
the `auto` strategy.
(Optional) Strategy for splitting files into chunks
metadata:
type: object
additionalProperties:
@ -11106,21 +11114,25 @@ components:
- type: array
- type: object
description: >-
Set of 16 key-value pairs that can be attached to an object.
Set of key-value pairs that can be attached to the vector store
embedding_model:
type: string
description: >-
The embedding model to use for this vector store.
(Optional) The embedding model to use for this vector store
embedding_dimension:
type: integer
default: 384
description: >-
The dimension of the embedding vectors (default: 384).
(Optional) The dimension of the embedding vectors (default: 384)
provider_id:
type: string
description: >-
The ID of the provider to use for this vector store.
(Optional) The ID of the provider to use for this vector store
additionalProperties: false
title: OpenaiCreateVectorStoreRequest
title: >-
OpenAICreateVectorStoreRequestWithExtraBody
description: >-
Request to create a vector store with extra_body support.
OpenaiUpdateVectorStoreRequest:
type: object
properties:
@ -11241,7 +11253,7 @@ components:
title: VectorStoreChunkingStrategyStaticConfig
description: >-
Configuration for static chunking strategy.
OpenaiCreateVectorStoreFileBatchRequest:
"OpenAICreateVectorStoreFileBatchRequestWithExtraBody":
type: object
properties:
file_ids:
@ -11249,7 +11261,7 @@ components:
items:
type: string
description: >-
A list of File IDs that the vector store should use.
A list of File IDs that the vector store should use
attributes:
type: object
additionalProperties:
@ -11261,16 +11273,19 @@ components:
- type: array
- type: object
description: >-
(Optional) Key-value attributes to store with the files.
(Optional) Key-value attributes to store with the files
chunking_strategy:
$ref: '#/components/schemas/VectorStoreChunkingStrategy'
description: >-
(Optional) The chunking strategy used to chunk the file(s). Defaults to
auto.
auto
additionalProperties: false
required:
- file_ids
title: OpenaiCreateVectorStoreFileBatchRequest
title: >-
OpenAICreateVectorStoreFileBatchRequestWithExtraBody
description: >-
Request to create a vector store file batch with extra_body support.
VectorStoreFileBatchObject:
type: object
properties:

31
llama_stack/apis/inference/inference.py
View file

@ -1140,6 +1140,25 @@ class OpenAIChatCompletionRequestWithExtraBody(BaseModel, extra="allow"):
user: str | None = None
# extra_body can be accessed via .model_extra
@json_schema_type
class OpenAIEmbeddingsRequestWithExtraBody(BaseModel, extra="allow"):
"""Request parameters for OpenAI-compatible embeddings endpoint.
:param model: The identifier of the model to use. The model must be an embedding model registered with Llama Stack and available via the /models endpoint.
:param input: Input text to embed, encoded as a string or array of strings. To embed multiple inputs in a single request, pass an array of strings.
:param encoding_format: (Optional) The format to return the embeddings in. Can be either "float" or "base64". Defaults to "float".
:param dimensions: (Optional) The number of dimensions the resulting output embeddings should have. Only supported in text-embedding-3 and later models.
:param user: (Optional) A unique identifier representing your end-user, which can help OpenAI to monitor and detect abuse.
"""
model: str
input: str | list[str]
encoding_format: str | None = "float"
dimensions: int | None = None
user: str | None = None
@runtime_checkable
@trace_protocol
class InferenceProvider(Protocol):
@ -1200,21 +1219,11 @@ class InferenceProvider(Protocol):
@webmethod(route="/embeddings", method="POST", level=LLAMA_STACK_API_V1)
async def openai_embeddings(
self,
model: str,
input: str | list[str],
encoding_format: str | None = "float",
dimensions: int | None = None,
user: str | None = None,
params: Annotated[OpenAIEmbeddingsRequestWithExtraBody, Body(...)],
) -> OpenAIEmbeddingsResponse:
"""Create embeddings.
Generate OpenAI-compatible embeddings for the given input using the specified model.
:param model: The identifier of the model to use. The model must be an embedding model registered with Llama Stack and available via the /models endpoint.
:param input: Input text to embed, encoded as a string or array of strings. To embed multiple inputs in a single request, pass an array of strings.
:param encoding_format: (Optional) The format to return the embeddings in. Can be either "float" or "base64". Defaults to "float".
:param dimensions: (Optional) The number of dimensions the resulting output embeddings should have. Only supported in text-embedding-3 and later models.
:param user: (Optional) A unique identifier representing your end-user, which can help OpenAI to monitor and detect abuse.
:returns: An OpenAIEmbeddingsResponse containing the embeddings.
"""
...

67
llama_stack/apis/vector_io/vector_io.py
View file

@ -11,6 +11,7 @@
import uuid
from typing import Annotated, Any, Literal, Protocol, runtime_checkable
from fastapi import Body
from pydantic import BaseModel, Field
from llama_stack.apis.inference import InterleavedContent
@ -466,6 +467,46 @@ class VectorStoreFilesListInBatchResponse(BaseModel):
has_more: bool = False
# extra_body can be accessed via .model_extra
@json_schema_type
class OpenAICreateVectorStoreRequestWithExtraBody(BaseModel, extra="allow"):
"""Request to create a vector store with extra_body support.
:param name: (Optional) A name for the vector store
:param file_ids: List of file IDs to include in the vector store
:param expires_after: (Optional) Expiration policy for the vector store
:param chunking_strategy: (Optional) Strategy for splitting files into chunks
:param metadata: Set of key-value pairs that can be attached to the vector store
:param embedding_model: (Optional) The embedding model to use for this vector store
:param embedding_dimension: (Optional) The dimension of the embedding vectors (default: 384)
:param provider_id: (Optional) The ID of the provider to use for this vector store
"""
name: str | None = None
file_ids: list[str] | None = None
expires_after: dict[str, Any] | None = None
chunking_strategy: dict[str, Any] | None = None
metadata: dict[str, Any] | None = None
embedding_model: str | None = None
embedding_dimension: int | None = 384
provider_id: str | None = None
# extra_body can be accessed via .model_extra
@json_schema_type
class OpenAICreateVectorStoreFileBatchRequestWithExtraBody(BaseModel, extra="allow"):
"""Request to create a vector store file batch with extra_body support.
:param file_ids: A list of File IDs that the vector store should use
:param attributes: (Optional) Key-value attributes to store with the files
:param chunking_strategy: (Optional) The chunking strategy used to chunk the file(s). Defaults to auto
"""
file_ids: list[str]
attributes: dict[str, Any] | None = None
chunking_strategy: VectorStoreChunkingStrategy | None = None
class VectorDBStore(Protocol):
def get_vector_db(self, vector_db_id: str) -> VectorDB | None: ...
@ -516,25 +557,11 @@ class VectorIO(Protocol):
@webmethod(route="/vector_stores", method="POST", level=LLAMA_STACK_API_V1)
async def openai_create_vector_store(
self,
name: str | None = None,
file_ids: list[str] | None = None,
expires_after: dict[str, Any] | None = None,
chunking_strategy: dict[str, Any] | None = None,
metadata: dict[str, Any] | None = None,
embedding_model: str | None = None,
embedding_dimension: int | None = 384,
provider_id: str | None = None,
params: Annotated[OpenAICreateVectorStoreRequestWithExtraBody, Body(...)],
) -> VectorStoreObject:
"""Creates a vector store.
:param name: A name for the vector store.
:param file_ids: A list of File IDs that the vector store should use. Useful for tools like `file_search` that can access files.
:param expires_after: The expiration policy for a vector store.
:param chunking_strategy: The chunking strategy used to chunk the file(s). If not set, will use the `auto` strategy.
:param metadata: Set of 16 key-value pairs that can be attached to an object.
:param embedding_model: The embedding model to use for this vector store.
:param embedding_dimension: The dimension of the embedding vectors (default: 384).
:param provider_id: The ID of the provider to use for this vector store.
Generate an OpenAI-compatible vector store with the given parameters.
:returns: A VectorStoreObject representing the created vector store.
"""
...
@ -827,16 +854,12 @@ class VectorIO(Protocol):
async def openai_create_vector_store_file_batch(
self,
vector_store_id: str,
file_ids: list[str],
attributes: dict[str, Any] | None = None,
chunking_strategy: VectorStoreChunkingStrategy | None = None,
params: Annotated[OpenAICreateVectorStoreFileBatchRequestWithExtraBody, Body(...)],
) -> VectorStoreFileBatchObject:
"""Create a vector store file batch.
Generate an OpenAI-compatible vector store file batch for the given vector store.
:param vector_store_id: The ID of the vector store to create the file batch for.
:param file_ids: A list of File IDs that the vector store should use.
:param attributes: (Optional) Key-value attributes to store with the files.
:param chunking_strategy: (Optional) The chunking strategy used to chunk the file(s). Defaults to auto.
:returns: A VectorStoreFileBatchObject representing the created file batch.
"""
...

23
llama_stack/core/routers/inference.py
View file

@ -40,6 +40,7 @@ from llama_stack.apis.inference import (
OpenAICompletion,
OpenAICompletionRequestWithExtraBody,
OpenAICompletionWithInputMessages,
OpenAIEmbeddingsRequestWithExtraBody,
OpenAIEmbeddingsResponse,
OpenAIMessageParam,
Order,
@ -279,26 +280,18 @@ class InferenceRouter(Inference):
async def openai_embeddings(
self,
model: str,
input: str | list[str],
encoding_format: str | None = "float",
dimensions: int | None = None,
user: str | None = None,
params: Annotated[OpenAIEmbeddingsRequestWithExtraBody, Body(...)],
) -> OpenAIEmbeddingsResponse:
logger.debug(
f"InferenceRouter.openai_embeddings: {model=}, input_type={type(input)}, {encoding_format=}, {dimensions=}",
)
model_obj = await self._get_model(model, ModelType.embedding)
params = dict(
model=model_obj.identifier,
input=input,
encoding_format=encoding_format,
dimensions=dimensions,
user=user,
f"InferenceRouter.openai_embeddings: model={params.model}, input_type={type(params.input)}, encoding_format={params.encoding_format}, dimensions={params.dimensions}",
)
model_obj = await self._get_model(params.model, ModelType.embedding)
# Update model to use resolved identifier
params.model = model_obj.identifier
provider = await self.routing_table.get_provider_impl(model_obj.identifier)
return await provider.openai_embeddings(**params)
return await provider.openai_embeddings(params)
async def list_chat_completions(
self,

7
llama_stack/providers/remote/inference/bedrock/bedrock.py
View file

@ -14,6 +14,7 @@ from llama_stack.apis.inference import (
Inference,
OpenAIChatCompletionRequestWithExtraBody,
OpenAICompletionRequestWithExtraBody,
OpenAIEmbeddingsRequestWithExtraBody,
OpenAIEmbeddingsResponse,
)
from llama_stack.apis.inference.inference import (
@ -124,11 +125,7 @@ class BedrockInferenceAdapter(
async def openai_embeddings(
self,
model: str,
input: str | list[str],
encoding_format: str | None = "float",
dimensions: int | None = None,
user: str | None = None,
params: OpenAIEmbeddingsRequestWithExtraBody,
) -> OpenAIEmbeddingsResponse:
raise NotImplementedError()

11
llama_stack/providers/remote/inference/cerebras/cerebras.py
View file

@ -6,7 +6,10 @@
from urllib.parse import urljoin
from llama_stack.apis.inference import OpenAIEmbeddingsResponse
from llama_stack.apis.inference import (
OpenAIEmbeddingsRequestWithExtraBody,
OpenAIEmbeddingsResponse,
)
from llama_stack.providers.utils.inference.openai_mixin import OpenAIMixin
from .config import CerebrasImplConfig
@ -20,10 +23,6 @@ class CerebrasInferenceAdapter(OpenAIMixin):
async def openai_embeddings(
self,
model: str,
input: str | list[str],
encoding_format: str | None = "float",
dimensions: int | None = None,
user: str | None = None,
params: OpenAIEmbeddingsRequestWithExtraBody,
) -> OpenAIEmbeddingsResponse:
raise NotImplementedError()

7
llama_stack/providers/remote/inference/llama_openai_compat/llama.py
View file

@ -7,6 +7,7 @@
from llama_stack.apis.inference.inference import (
OpenAICompletion,
OpenAICompletionRequestWithExtraBody,
OpenAIEmbeddingsRequestWithExtraBody,
OpenAIEmbeddingsResponse,
)
from llama_stack.log import get_logger
@ -40,10 +41,6 @@ class LlamaCompatInferenceAdapter(OpenAIMixin):
async def openai_embeddings(
self,
model: str,
input: str | list[str],
encoding_format: str | None = "float",
dimensions: int | None = None,
user: str | None = None,
params: OpenAIEmbeddingsRequestWithExtraBody,
) -> OpenAIEmbeddingsResponse:
raise NotImplementedError()

17
llama_stack/providers/remote/inference/nvidia/nvidia.py
View file

@ -9,6 +9,7 @@ from openai import NOT_GIVEN
from llama_stack.apis.inference import (
OpenAIEmbeddingData,
OpenAIEmbeddingsRequestWithExtraBody,
OpenAIEmbeddingsResponse,
OpenAIEmbeddingUsage,
)
@ -78,11 +79,7 @@ class NVIDIAInferenceAdapter(OpenAIMixin):
async def openai_embeddings(
self,
model: str,
input: str | list[str],
encoding_format: str | None = "float",
dimensions: int | None = None,
user: str | None = None,
params: OpenAIEmbeddingsRequestWithExtraBody,
) -> OpenAIEmbeddingsResponse:
"""
OpenAI-compatible embeddings for NVIDIA NIM.
@ -99,11 +96,11 @@ class NVIDIAInferenceAdapter(OpenAIMixin):
)
response = await self.client.embeddings.create(
model=await self._get_provider_model_id(model),
input=input,
encoding_format=encoding_format if encoding_format is not None else NOT_GIVEN,
dimensions=dimensions if dimensions is not None else NOT_GIVEN,
user=user if user is not None else NOT_GIVEN,
model=await self._get_provider_model_id(params.model),
input=params.input,
encoding_format=params.encoding_format if params.encoding_format is not None else NOT_GIVEN,
dimensions=params.dimensions if params.dimensions is not None else NOT_GIVEN,
user=params.user if params.user is not None else NOT_GIVEN,
extra_body=extra_body,
)

7
llama_stack/providers/remote/inference/passthrough/passthrough.py
View file

@ -16,6 +16,7 @@ from llama_stack.apis.inference import (
OpenAIChatCompletionRequestWithExtraBody,
OpenAICompletion,
OpenAICompletionRequestWithExtraBody,
OpenAIEmbeddingsRequestWithExtraBody,
OpenAIEmbeddingsResponse,
)
from llama_stack.apis.models import Model
@ -69,11 +70,7 @@ class PassthroughInferenceAdapter(Inference):
async def openai_embeddings(
self,
model: str,
input: str | list[str],
encoding_format: str | None = "float",
dimensions: int | None = None,
user: str | None = None,
params: OpenAIEmbeddingsRequestWithExtraBody,
) -> OpenAIEmbeddingsResponse:
raise NotImplementedError()

11
llama_stack/providers/remote/inference/tgi/tgi.py
View file

@ -10,7 +10,10 @@ from collections.abc import Iterable
from huggingface_hub import AsyncInferenceClient, HfApi
from pydantic import SecretStr
from llama_stack.apis.inference import OpenAIEmbeddingsResponse
from llama_stack.apis.inference import (
OpenAIEmbeddingsRequestWithExtraBody,
OpenAIEmbeddingsResponse,
)
from llama_stack.log import get_logger
from llama_stack.providers.utils.inference.openai_mixin import OpenAIMixin
@ -40,11 +43,7 @@ class _HfAdapter(OpenAIMixin):
async def openai_embeddings(
self,
model: str,
input: str | list[str],
encoding_format: str | None = "float",
dimensions: int | None = None,
user: str | None = None,
params: OpenAIEmbeddingsRequestWithExtraBody,
) -> OpenAIEmbeddingsResponse:
raise NotImplementedError()

23
llama_stack/providers/remote/inference/together/together.py
View file

@ -11,6 +11,7 @@ from together import AsyncTogether
from together.constants import BASE_URL
from llama_stack.apis.inference import (
OpenAIEmbeddingsRequestWithExtraBody,
OpenAIEmbeddingsResponse,
)
from llama_stack.apis.inference.inference import OpenAIEmbeddingUsage
@ -62,11 +63,7 @@ class TogetherInferenceAdapter(OpenAIMixin, NeedsRequestProviderData):
async def openai_embeddings(
self,
model: str,
input: str | list[str],
encoding_format: str | None = "float",
dimensions: int | None = None,
user: str | None = None,
params: OpenAIEmbeddingsRequestWithExtraBody,
) -> OpenAIEmbeddingsResponse:
"""
Together's OpenAI-compatible embeddings endpoint is not compatible with
@ -78,25 +75,27 @@ class TogetherInferenceAdapter(OpenAIMixin, NeedsRequestProviderData):
- does not support dimensions param, returns 400 Unrecognized request arguments supplied: dimensions
"""
# Together support ticket #13332 -> will not fix
if user is not None:
if params.user is not None:
raise ValueError("Together's embeddings endpoint does not support user param.")
# Together support ticket #13333 -> escalated
if dimensions is not None:
if params.dimensions is not None:
raise ValueError("Together's embeddings endpoint does not support dimensions param.")
response = await self.client.embeddings.create(
model=await self._get_provider_model_id(model),
input=input,
encoding_format=encoding_format,
model=await self._get_provider_model_id(params.model),
input=params.input,
encoding_format=params.encoding_format,
)
response.model = model # return the user the same model id they provided, avoid exposing the provider model id
response.model = (
params.model
) # return the user the same model id they provided, avoid exposing the provider model id
# Together support ticket #13330 -> escalated
# - togethercomputer/m2-bert-80M-32k-retrieval *does not* return usage information
if not hasattr(response, "usage") or response.usage is None:
logger.warning(
f"Together's embedding endpoint for {model} did not return usage information, substituting -1s."
f"Together's embedding endpoint for {params.model} did not return usage information, substituting -1s."
)
response.usage = OpenAIEmbeddingUsage(prompt_tokens=-1, total_tokens=-1)

15
llama_stack/providers/utils/inference/embedding_mixin.py
View file

@ -17,6 +17,7 @@ if TYPE_CHECKING:
from llama_stack.apis.inference import (
ModelStore,
OpenAIEmbeddingData,
OpenAIEmbeddingsRequestWithExtraBody,
OpenAIEmbeddingsResponse,
OpenAIEmbeddingUsage,
)
@ -32,26 +33,22 @@ class SentenceTransformerEmbeddingMixin:
async def openai_embeddings(
self,
model: str,
input: str | list[str],
encoding_format: str | None = "float",
dimensions: int | None = None,
user: str | None = None,
params: OpenAIEmbeddingsRequestWithExtraBody,
) -> OpenAIEmbeddingsResponse:
# Convert input to list format if it's a single string
input_list = [input] if isinstance(input, str) else input
input_list = [params.input] if isinstance(params.input, str) else params.input
if not input_list:
raise ValueError("Empty list not supported")
# Get the model and generate embeddings
model_obj = await self.model_store.get_model(model)
model_obj = await self.model_store.get_model(params.model)
embedding_model = await self._load_sentence_transformer_model(model_obj.provider_resource_id)
embeddings = await asyncio.to_thread(embedding_model.encode, input_list, show_progress_bar=False)
# Convert embeddings to the requested format
data = []
for i, embedding in enumerate(embeddings):
if encoding_format == "base64":
if params.encoding_format == "base64":
# Convert float array to base64 string
float_bytes = struct.pack(f"{len(embedding)}f", *embedding)
embedding_value = base64.b64encode(float_bytes).decode("ascii")
@ -70,7 +67,7 @@ class SentenceTransformerEmbeddingMixin:
usage = OpenAIEmbeddingUsage(prompt_tokens=-1, total_tokens=-1)
return OpenAIEmbeddingsResponse(
data=data,
model=model,
model=params.model,
usage=usage,
)

15
llama_stack/providers/utils/inference/litellm_openai_mixin.py
View file

@ -20,6 +20,7 @@ from llama_stack.apis.inference import (
OpenAICompletion,
OpenAICompletionRequestWithExtraBody,
OpenAIEmbeddingData,
OpenAIEmbeddingsRequestWithExtraBody,
OpenAIEmbeddingsResponse,
OpenAIEmbeddingUsage,
ToolChoice,
@ -189,16 +190,12 @@ class LiteLLMOpenAIMixin(
async def openai_embeddings(
self,
model: str,
input: str | list[str],
encoding_format: str | None = "float",
dimensions: int | None = None,
user: str | None = None,
params: OpenAIEmbeddingsRequestWithExtraBody,
) -> OpenAIEmbeddingsResponse:
model_obj = await self.model_store.get_model(model)
model_obj = await self.model_store.get_model(params.model)
# Convert input to list if it's a string
input_list = [input] if isinstance(input, str) else input
input_list = [params.input] if isinstance(params.input, str) else params.input
# Call litellm embedding function
# litellm.drop_params = True
@ -207,11 +204,11 @@ class LiteLLMOpenAIMixin(
input=input_list,
api_key=self.get_api_key(),
api_base=self.api_base,
dimensions=dimensions,
dimensions=params.dimensions,
)
# Convert response to OpenAI format
data = b64_encode_openai_embeddings_response(response.data, encoding_format)
data = b64_encode_openai_embeddings_response(response.data, params.encoding_format)
usage = OpenAIEmbeddingUsage(
prompt_tokens=response["usage"]["prompt_tokens"],

31
llama_stack/providers/utils/inference/openai_mixin.py
View file

@ -21,6 +21,7 @@ from llama_stack.apis.inference import (
OpenAICompletion,
OpenAICompletionRequestWithExtraBody,
OpenAIEmbeddingData,
OpenAIEmbeddingsRequestWithExtraBody,
OpenAIEmbeddingsResponse,
OpenAIEmbeddingUsage,
OpenAIMessageParam,
@ -316,23 +317,27 @@ class OpenAIMixin(NeedsRequestProviderData, ABC, BaseModel):
async def openai_embeddings(
self,
model: str,
input: str | list[str],
encoding_format: str | None = "float",
dimensions: int | None = None,
user: str | None = None,
params: OpenAIEmbeddingsRequestWithExtraBody,
) -> OpenAIEmbeddingsResponse:
"""
Direct OpenAI embeddings API call.
"""
# Prepare request parameters
request_params = {
"model": await self._get_provider_model_id(params.model),
"input": params.input,
"encoding_format": params.encoding_format if params.encoding_format is not None else NOT_GIVEN,
"dimensions": params.dimensions if params.dimensions is not None else NOT_GIVEN,
"user": params.user if params.user is not None else NOT_GIVEN,
}
# Add extra_body if present
extra_body = params.model_extra
if extra_body:
request_params["extra_body"] = extra_body
# Call OpenAI embeddings API with properly typed parameters
response = await self.client.embeddings.create(
model=await self._get_provider_model_id(model),
input=input,
encoding_format=encoding_format if encoding_format is not None else NOT_GIVEN,
dimensions=dimensions if dimensions is not None else NOT_GIVEN,
user=user if user is not None else NOT_GIVEN,
)
response = await self.client.embeddings.create(**request_params)
data = []
for i, embedding_data in enumerate(response.data):
@ -350,7 +355,7 @@ class OpenAIMixin(NeedsRequestProviderData, ABC, BaseModel):
return OpenAIEmbeddingsResponse(
data=data,
model=model,
model=params.model,
usage=usage,
)

60
llama_stack/providers/utils/memory/openai_vector_store_mixin.py
View file

@ -19,6 +19,8 @@ from llama_stack.apis.files import Files, OpenAIFileObject
from llama_stack.apis.vector_dbs import VectorDB
from llama_stack.apis.vector_io import (
Chunk,
OpenAICreateVectorStoreFileBatchRequestWithExtraBody,
OpenAICreateVectorStoreRequestWithExtraBody,
QueryChunksResponse,
SearchRankingOptions,
VectorStoreChunkingStrategy,
@ -340,39 +342,37 @@ class OpenAIVectorStoreMixin(ABC):
async def openai_create_vector_store(
self,
name: str | None = None,
file_ids: list[str] | None = None,
expires_after: dict[str, Any] | None = None,
chunking_strategy: dict[str, Any] | None = None,
metadata: dict[str, Any] | None = None,
embedding_model: str | None = None,
embedding_dimension: int | None = 384,
provider_id: str | None = None,
provider_vector_db_id: str | None = None,
params: OpenAICreateVectorStoreRequestWithExtraBody,
) -> VectorStoreObject:
"""Creates a vector store."""
created_at = int(time.time())
# Extract provider_vector_db_id from extra_body if present
provider_vector_db_id = None
if params.model_extra and "provider_vector_db_id" in params.model_extra:
provider_vector_db_id = params.model_extra["provider_vector_db_id"]
# Derive the canonical vector_db_id (allow override, else generate)
vector_db_id = provider_vector_db_id or generate_object_id("vector_store", lambda: f"vs_{uuid.uuid4()}")
if provider_id is None:
if params.provider_id is None:
raise ValueError("Provider ID is required")
if embedding_model is None:
if params.embedding_model is None:
raise ValueError("Embedding model is required")
# Embedding dimension is required (defaulted to 384 if not provided)
if embedding_dimension is None:
if params.embedding_dimension is None:
raise ValueError("Embedding dimension is required")
# Register the VectorDB backing this vector store
vector_db = VectorDB(
identifier=vector_db_id,
embedding_dimension=embedding_dimension,
embedding_model=embedding_model,
provider_id=provider_id,
embedding_dimension=params.embedding_dimension,
embedding_model=params.embedding_model,
provider_id=params.provider_id,
provider_resource_id=vector_db_id,
vector_db_name=name,
vector_db_name=params.name,
)
await self.register_vector_db(vector_db)
@ -391,21 +391,21 @@ class OpenAIVectorStoreMixin(ABC):
"id": vector_db_id,
"object": "vector_store",
"created_at": created_at,
"name": name,
"name": params.name,
"usage_bytes": 0,
"file_counts": file_counts.model_dump(),
"status": status,
"expires_after": expires_after,
"expires_after": params.expires_after,
"expires_at": None,
"last_active_at": created_at,
"file_ids": [],
"chunking_strategy": chunking_strategy,
"chunking_strategy": params.chunking_strategy,
}
# Add provider information to metadata if provided
metadata = metadata or {}
if provider_id:
metadata["provider_id"] = provider_id
metadata = params.metadata or {}
if params.provider_id:
metadata["provider_id"] = params.provider_id
if provider_vector_db_id:
metadata["provider_vector_db_id"] = provider_vector_db_id
store_info["metadata"] = metadata
@ -417,7 +417,7 @@ class OpenAIVectorStoreMixin(ABC):
self.openai_vector_stores[vector_db_id] = store_info
# Now that our vector store is created, attach any files that were provided
file_ids = file_ids or []
file_ids = params.file_ids or []
tasks = [self.openai_attach_file_to_vector_store(vector_db_id, file_id) for file_id in file_ids]
await asyncio.gather(*tasks)
@ -976,15 +976,13 @@ class OpenAIVectorStoreMixin(ABC):
async def openai_create_vector_store_file_batch(
self,
vector_store_id: str,
file_ids: list[str],
attributes: dict[str, Any] | None = None,
chunking_strategy: VectorStoreChunkingStrategy | None = None,
params: OpenAICreateVectorStoreFileBatchRequestWithExtraBody,
) -> VectorStoreFileBatchObject:
"""Create a vector store file batch."""
if vector_store_id not in self.openai_vector_stores:
raise VectorStoreNotFoundError(vector_store_id)
chunking_strategy = chunking_strategy or VectorStoreChunkingStrategyAuto()
chunking_strategy = params.chunking_strategy or VectorStoreChunkingStrategyAuto()
created_at = int(time.time())
batch_id = generate_object_id("vector_store_file_batch", lambda: f"batch_{uuid.uuid4()}")
@ -996,8 +994,8 @@ class OpenAIVectorStoreMixin(ABC):
completed=0,
cancelled=0,
failed=0,
in_progress=len(file_ids),
total=len(file_ids),
in_progress=len(params.file_ids),
total=len(params.file_ids),
)
# Create batch object immediately with in_progress status
@ -1011,8 +1009,8 @@ class OpenAIVectorStoreMixin(ABC):
batch_info = {
**batch_object.model_dump(),
"file_ids": file_ids,
"attributes": attributes,
"file_ids": params.file_ids,
"attributes": params.attributes,
"chunking_strategy": chunking_strategy.model_dump(),
"expires_at": expires_at,
}