mirror of
https://github.com/meta-llama/llama-stack.git
synced 2025-10-12 21:58:38 +00:00
feat(responses)!: add reasoning and annotation added events (#3793)
Implements missing streaming events from OpenAI Responses API spec: - reasoning text/summary events for o1/o3 models, - refusal events for safety moderation - annotation events for citations, - and file search streaming events. Added optional reasoning_content field to chat completion chunks to support non-standard provider extensions. **NOTE:** OpenAI does _not_ fill reasoning_content when users use the chat_completion APIs. This means there is no way for us to implement Responses (with reasoning) by using OpenAI chat completions! We'd need to transparently punt to OpenAI's responses endpoints if we wish to do that. For others though (vLLM, etc.) we can use it. ## Test Plan File search streaming test passes: ``` ./scripts/integration-tests.sh --stack-config server:ci-tests \ --suite responses --setup gpt --inference-mode replay --pattern test_response_file_search_streaming_events ``` Need more complex setup and validation for reasoning tests (need a vLLM powered OSS model maybe gpt-oss which can return reasoning_content). I will do that in a followup PR.
This commit is contained in:
Ashwin Bharambe
GitHub
parent
f365961731
commit
7c63aebd64
25 changed files with 23530 additions and 2 deletions
506
docs/static/llama-stack-spec.yaml
vendored
506
docs/static/llama-stack-spec.yaml
vendored
|
|
@ -3950,6 +3950,11 @@ components:
|
|||
items:
|
||||
$ref: '#/components/schemas/OpenAIChatCompletionToolCall'
|
||||
description: (Optional) The tool calls of the delta
|
||||
reasoning_content:
|
||||
type: string
|
||||
description: >-
|
||||
(Optional) The reasoning content from the model (non-standard, for o1/o3
|
||||
models)
|
||||
additionalProperties: false
|
||||
title: OpenAIChoiceDelta
|
||||
description: >-
|
||||
|
|
@ -6224,6 +6229,26 @@ components:
|
|||
title: OpenAIResponseContentPartOutputText
|
||||
description: >-
|
||||
Text content within a streamed response part.
|
||||
"OpenAIResponseContentPartReasoningSummary":
|
||||
type: object
|
||||
properties:
|
||||
type:
|
||||
type: string
|
||||
const: summary_text
|
||||
default: summary_text
|
||||
description: >-
|
||||
Content part type identifier, always "summary_text"
|
||||
text:
|
||||
type: string
|
||||
description: Summary text
|
||||
additionalProperties: false
|
||||
required:
|
||||
- type
|
||||
- text
|
||||
title: >-
|
||||
OpenAIResponseContentPartReasoningSummary
|
||||
description: >-
|
||||
Reasoning summary part in a streamed response.
|
||||
OpenAIResponseContentPartReasoningText:
|
||||
type: object
|
||||
properties:
|
||||
|
|
@ -6285,6 +6310,18 @@ components:
|
|||
- $ref: '#/components/schemas/OpenAIResponseObjectStreamResponseMcpCallCompleted'
|
||||
- $ref: '#/components/schemas/OpenAIResponseObjectStreamResponseContentPartAdded'
|
||||
- $ref: '#/components/schemas/OpenAIResponseObjectStreamResponseContentPartDone'
|
||||
- $ref: '#/components/schemas/OpenAIResponseObjectStreamResponseReasoningTextDelta'
|
||||
- $ref: '#/components/schemas/OpenAIResponseObjectStreamResponseReasoningTextDone'
|
||||
- $ref: '#/components/schemas/OpenAIResponseObjectStreamResponseReasoningSummaryPartAdded'
|
||||
- $ref: '#/components/schemas/OpenAIResponseObjectStreamResponseReasoningSummaryPartDone'
|
||||
- $ref: '#/components/schemas/OpenAIResponseObjectStreamResponseReasoningSummaryTextDelta'
|
||||
- $ref: '#/components/schemas/OpenAIResponseObjectStreamResponseReasoningSummaryTextDone'
|
||||
- $ref: '#/components/schemas/OpenAIResponseObjectStreamResponseRefusalDelta'
|
||||
- $ref: '#/components/schemas/OpenAIResponseObjectStreamResponseRefusalDone'
|
||||
- $ref: '#/components/schemas/OpenAIResponseObjectStreamResponseOutputTextAnnotationAdded'
|
||||
- $ref: '#/components/schemas/OpenAIResponseObjectStreamResponseFileSearchCallInProgress'
|
||||
- $ref: '#/components/schemas/OpenAIResponseObjectStreamResponseFileSearchCallSearching'
|
||||
- $ref: '#/components/schemas/OpenAIResponseObjectStreamResponseFileSearchCallCompleted'
|
||||
- $ref: '#/components/schemas/OpenAIResponseObjectStreamResponseIncomplete'
|
||||
- $ref: '#/components/schemas/OpenAIResponseObjectStreamResponseFailed'
|
||||
- $ref: '#/components/schemas/OpenAIResponseObjectStreamResponseCompleted'
|
||||
|
|
@ -6312,6 +6349,18 @@ components:
|
|||
response.mcp_call.completed: '#/components/schemas/OpenAIResponseObjectStreamResponseMcpCallCompleted'
|
||||
response.content_part.added: '#/components/schemas/OpenAIResponseObjectStreamResponseContentPartAdded'
|
||||
response.content_part.done: '#/components/schemas/OpenAIResponseObjectStreamResponseContentPartDone'
|
||||
response.reasoning_text.delta: '#/components/schemas/OpenAIResponseObjectStreamResponseReasoningTextDelta'
|
||||
response.reasoning_text.done: '#/components/schemas/OpenAIResponseObjectStreamResponseReasoningTextDone'
|
||||
response.reasoning_summary_part.added: '#/components/schemas/OpenAIResponseObjectStreamResponseReasoningSummaryPartAdded'
|
||||
response.reasoning_summary_part.done: '#/components/schemas/OpenAIResponseObjectStreamResponseReasoningSummaryPartDone'
|
||||
response.reasoning_summary_text.delta: '#/components/schemas/OpenAIResponseObjectStreamResponseReasoningSummaryTextDelta'
|
||||
response.reasoning_summary_text.done: '#/components/schemas/OpenAIResponseObjectStreamResponseReasoningSummaryTextDone'
|
||||
response.refusal.delta: '#/components/schemas/OpenAIResponseObjectStreamResponseRefusalDelta'
|
||||
response.refusal.done: '#/components/schemas/OpenAIResponseObjectStreamResponseRefusalDone'
|
||||
response.output_text.annotation.added: '#/components/schemas/OpenAIResponseObjectStreamResponseOutputTextAnnotationAdded'
|
||||
response.file_search_call.in_progress: '#/components/schemas/OpenAIResponseObjectStreamResponseFileSearchCallInProgress'
|
||||
response.file_search_call.searching: '#/components/schemas/OpenAIResponseObjectStreamResponseFileSearchCallSearching'
|
||||
response.file_search_call.completed: '#/components/schemas/OpenAIResponseObjectStreamResponseFileSearchCallCompleted'
|
||||
response.incomplete: '#/components/schemas/OpenAIResponseObjectStreamResponseIncomplete'
|
||||
response.failed: '#/components/schemas/OpenAIResponseObjectStreamResponseFailed'
|
||||
response.completed: '#/components/schemas/OpenAIResponseObjectStreamResponseCompleted'
|
||||
|
|
@ -6487,6 +6536,99 @@ components:
|
|||
title: OpenAIResponseObjectStreamResponseFailed
|
||||
description: >-
|
||||
Streaming event emitted when a response fails.
|
||||
"OpenAIResponseObjectStreamResponseFileSearchCallCompleted":
|
||||
type: object
|
||||
properties:
|
||||
item_id:
|
||||
type: string
|
||||
description: >-
|
||||
Unique identifier of the completed file search call
|
||||
output_index:
|
||||
type: integer
|
||||
description: >-
|
||||
Index position of the item in the output list
|
||||
sequence_number:
|
||||
type: integer
|
||||
description: >-
|
||||
Sequential number for ordering streaming events
|
||||
type:
|
||||
type: string
|
||||
const: response.file_search_call.completed
|
||||
default: response.file_search_call.completed
|
||||
description: >-
|
||||
Event type identifier, always "response.file_search_call.completed"
|
||||
additionalProperties: false
|
||||
required:
|
||||
- item_id
|
||||
- output_index
|
||||
- sequence_number
|
||||
- type
|
||||
title: >-
|
||||
OpenAIResponseObjectStreamResponseFileSearchCallCompleted
|
||||
description: >-
|
||||
Streaming event for completed file search calls.
|
||||
"OpenAIResponseObjectStreamResponseFileSearchCallInProgress":
|
||||
type: object
|
||||
properties:
|
||||
item_id:
|
||||
type: string
|
||||
description: >-
|
||||
Unique identifier of the file search call
|
||||
output_index:
|
||||
type: integer
|
||||
description: >-
|
||||
Index position of the item in the output list
|
||||
sequence_number:
|
||||
type: integer
|
||||
description: >-
|
||||
Sequential number for ordering streaming events
|
||||
type:
|
||||
type: string
|
||||
const: response.file_search_call.in_progress
|
||||
default: response.file_search_call.in_progress
|
||||
description: >-
|
||||
Event type identifier, always "response.file_search_call.in_progress"
|
||||
additionalProperties: false
|
||||
required:
|
||||
- item_id
|
||||
- output_index
|
||||
- sequence_number
|
||||
- type
|
||||
title: >-
|
||||
OpenAIResponseObjectStreamResponseFileSearchCallInProgress
|
||||
description: >-
|
||||
Streaming event for file search calls in progress.
|
||||
"OpenAIResponseObjectStreamResponseFileSearchCallSearching":
|
||||
type: object
|
||||
properties:
|
||||
item_id:
|
||||
type: string
|
||||
description: >-
|
||||
Unique identifier of the file search call
|
||||
output_index:
|
||||
type: integer
|
||||
description: >-
|
||||
Index position of the item in the output list
|
||||
sequence_number:
|
||||
type: integer
|
||||
description: >-
|
||||
Sequential number for ordering streaming events
|
||||
type:
|
||||
type: string
|
||||
const: response.file_search_call.searching
|
||||
default: response.file_search_call.searching
|
||||
description: >-
|
||||
Event type identifier, always "response.file_search_call.searching"
|
||||
additionalProperties: false
|
||||
required:
|
||||
- item_id
|
||||
- output_index
|
||||
- sequence_number
|
||||
- type
|
||||
title: >-
|
||||
OpenAIResponseObjectStreamResponseFileSearchCallSearching
|
||||
description: >-
|
||||
Streaming event for file search currently searching.
|
||||
"OpenAIResponseObjectStreamResponseFunctionCallArgumentsDelta":
|
||||
type: object
|
||||
properties:
|
||||
|
|
@ -6879,6 +7021,62 @@ components:
|
|||
OpenAIResponseObjectStreamResponseOutputItemDone
|
||||
description: >-
|
||||
Streaming event for when an output item is completed.
|
||||
"OpenAIResponseObjectStreamResponseOutputTextAnnotationAdded":
|
||||
type: object
|
||||
properties:
|
||||
item_id:
|
||||
type: string
|
||||
description: >-
|
||||
Unique identifier of the item to which the annotation is being added
|
||||
output_index:
|
||||
type: integer
|
||||
description: >-
|
||||
Index position of the output item in the response's output array
|
||||
content_index:
|
||||
type: integer
|
||||
description: >-
|
||||
Index position of the content part within the output item
|
||||
annotation_index:
|
||||
type: integer
|
||||
description: >-
|
||||
Index of the annotation within the content part
|
||||
annotation:
|
||||
oneOf:
|
||||
- $ref: '#/components/schemas/OpenAIResponseAnnotationFileCitation'
|
||||
- $ref: '#/components/schemas/OpenAIResponseAnnotationCitation'
|
||||
- $ref: '#/components/schemas/OpenAIResponseAnnotationContainerFileCitation'
|
||||
- $ref: '#/components/schemas/OpenAIResponseAnnotationFilePath'
|
||||
discriminator:
|
||||
propertyName: type
|
||||
mapping:
|
||||
file_citation: '#/components/schemas/OpenAIResponseAnnotationFileCitation'
|
||||
url_citation: '#/components/schemas/OpenAIResponseAnnotationCitation'
|
||||
container_file_citation: '#/components/schemas/OpenAIResponseAnnotationContainerFileCitation'
|
||||
file_path: '#/components/schemas/OpenAIResponseAnnotationFilePath'
|
||||
description: The annotation object being added
|
||||
sequence_number:
|
||||
type: integer
|
||||
description: >-
|
||||
Sequential number for ordering streaming events
|
||||
type:
|
||||
type: string
|
||||
const: response.output_text.annotation.added
|
||||
default: response.output_text.annotation.added
|
||||
description: >-
|
||||
Event type identifier, always "response.output_text.annotation.added"
|
||||
additionalProperties: false
|
||||
required:
|
||||
- item_id
|
||||
- output_index
|
||||
- content_index
|
||||
- annotation_index
|
||||
- annotation
|
||||
- sequence_number
|
||||
- type
|
||||
title: >-
|
||||
OpenAIResponseObjectStreamResponseOutputTextAnnotationAdded
|
||||
description: >-
|
||||
Streaming event for when an annotation is added to output text.
|
||||
"OpenAIResponseObjectStreamResponseOutputTextDelta":
|
||||
type: object
|
||||
properties:
|
||||
|
|
@ -6958,6 +7156,314 @@ components:
|
|||
OpenAIResponseObjectStreamResponseOutputTextDone
|
||||
description: >-
|
||||
Streaming event for when text output is completed.
|
||||
"OpenAIResponseObjectStreamResponseReasoningSummaryPartAdded":
|
||||
type: object
|
||||
properties:
|
||||
item_id:
|
||||
type: string
|
||||
description: Unique identifier of the output item
|
||||
output_index:
|
||||
type: integer
|
||||
description: Index position of the output item
|
||||
part:
|
||||
$ref: '#/components/schemas/OpenAIResponseContentPartReasoningSummary'
|
||||
description: The summary part that was added
|
||||
sequence_number:
|
||||
type: integer
|
||||
description: >-
|
||||
Sequential number for ordering streaming events
|
||||
summary_index:
|
||||
type: integer
|
||||
description: >-
|
||||
Index of the summary part within the reasoning summary
|
||||
type:
|
||||
type: string
|
||||
const: response.reasoning_summary_part.added
|
||||
default: response.reasoning_summary_part.added
|
||||
description: >-
|
||||
Event type identifier, always "response.reasoning_summary_part.added"
|
||||
additionalProperties: false
|
||||
required:
|
||||
- item_id
|
||||
- output_index
|
||||
- part
|
||||
- sequence_number
|
||||
- summary_index
|
||||
- type
|
||||
title: >-
|
||||
OpenAIResponseObjectStreamResponseReasoningSummaryPartAdded
|
||||
description: >-
|
||||
Streaming event for when a new reasoning summary part is added.
|
||||
"OpenAIResponseObjectStreamResponseReasoningSummaryPartDone":
|
||||
type: object
|
||||
properties:
|
||||
item_id:
|
||||
type: string
|
||||
description: Unique identifier of the output item
|
||||
output_index:
|
||||
type: integer
|
||||
description: Index position of the output item
|
||||
part:
|
||||
$ref: '#/components/schemas/OpenAIResponseContentPartReasoningSummary'
|
||||
description: The completed summary part
|
||||
sequence_number:
|
||||
type: integer
|
||||
description: >-
|
||||
Sequential number for ordering streaming events
|
||||
summary_index:
|
||||
type: integer
|
||||
description: >-
|
||||
Index of the summary part within the reasoning summary
|
||||
type:
|
||||
type: string
|
||||
const: response.reasoning_summary_part.done
|
||||
default: response.reasoning_summary_part.done
|
||||
description: >-
|
||||
Event type identifier, always "response.reasoning_summary_part.done"
|
||||
additionalProperties: false
|
||||
required:
|
||||
- item_id
|
||||
- output_index
|
||||
- part
|
||||
- sequence_number
|
||||
- summary_index
|
||||
- type
|
||||
title: >-
|
||||
OpenAIResponseObjectStreamResponseReasoningSummaryPartDone
|
||||
description: >-
|
||||
Streaming event for when a reasoning summary part is completed.
|
||||
"OpenAIResponseObjectStreamResponseReasoningSummaryTextDelta":
|
||||
type: object
|
||||
properties:
|
||||
delta:
|
||||
type: string
|
||||
description: Incremental summary text being added
|
||||
item_id:
|
||||
type: string
|
||||
description: Unique identifier of the output item
|
||||
output_index:
|
||||
type: integer
|
||||
description: Index position of the output item
|
||||
sequence_number:
|
||||
type: integer
|
||||
description: >-
|
||||
Sequential number for ordering streaming events
|
||||
summary_index:
|
||||
type: integer
|
||||
description: >-
|
||||
Index of the summary part within the reasoning summary
|
||||
type:
|
||||
type: string
|
||||
const: response.reasoning_summary_text.delta
|
||||
default: response.reasoning_summary_text.delta
|
||||
description: >-
|
||||
Event type identifier, always "response.reasoning_summary_text.delta"
|
||||
additionalProperties: false
|
||||
required:
|
||||
- delta
|
||||
- item_id
|
||||
- output_index
|
||||
- sequence_number
|
||||
- summary_index
|
||||
- type
|
||||
title: >-
|
||||
OpenAIResponseObjectStreamResponseReasoningSummaryTextDelta
|
||||
description: >-
|
||||
Streaming event for incremental reasoning summary text updates.
|
||||
"OpenAIResponseObjectStreamResponseReasoningSummaryTextDone":
|
||||
type: object
|
||||
properties:
|
||||
text:
|
||||
type: string
|
||||
description: Final complete summary text
|
||||
item_id:
|
||||
type: string
|
||||
description: Unique identifier of the output item
|
||||
output_index:
|
||||
type: integer
|
||||
description: Index position of the output item
|
||||
sequence_number:
|
||||
type: integer
|
||||
description: >-
|
||||
Sequential number for ordering streaming events
|
||||
summary_index:
|
||||
type: integer
|
||||
description: >-
|
||||
Index of the summary part within the reasoning summary
|
||||
type:
|
||||
type: string
|
||||
const: response.reasoning_summary_text.done
|
||||
default: response.reasoning_summary_text.done
|
||||
description: >-
|
||||
Event type identifier, always "response.reasoning_summary_text.done"
|
||||
additionalProperties: false
|
||||
required:
|
||||
- text
|
||||
- item_id
|
||||
- output_index
|
||||
- sequence_number
|
||||
- summary_index
|
||||
- type
|
||||
title: >-
|
||||
OpenAIResponseObjectStreamResponseReasoningSummaryTextDone
|
||||
description: >-
|
||||
Streaming event for when reasoning summary text is completed.
|
||||
"OpenAIResponseObjectStreamResponseReasoningTextDelta":
|
||||
type: object
|
||||
properties:
|
||||
content_index:
|
||||
type: integer
|
||||
description: >-
|
||||
Index position of the reasoning content part
|
||||
delta:
|
||||
type: string
|
||||
description: Incremental reasoning text being added
|
||||
item_id:
|
||||
type: string
|
||||
description: >-
|
||||
Unique identifier of the output item being updated
|
||||
output_index:
|
||||
type: integer
|
||||
description: >-
|
||||
Index position of the item in the output list
|
||||
sequence_number:
|
||||
type: integer
|
||||
description: >-
|
||||
Sequential number for ordering streaming events
|
||||
type:
|
||||
type: string
|
||||
const: response.reasoning_text.delta
|
||||
default: response.reasoning_text.delta
|
||||
description: >-
|
||||
Event type identifier, always "response.reasoning_text.delta"
|
||||
additionalProperties: false
|
||||
required:
|
||||
- content_index
|
||||
- delta
|
||||
- item_id
|
||||
- output_index
|
||||
- sequence_number
|
||||
- type
|
||||
title: >-
|
||||
OpenAIResponseObjectStreamResponseReasoningTextDelta
|
||||
description: >-
|
||||
Streaming event for incremental reasoning text updates.
|
||||
"OpenAIResponseObjectStreamResponseReasoningTextDone":
|
||||
type: object
|
||||
properties:
|
||||
content_index:
|
||||
type: integer
|
||||
description: >-
|
||||
Index position of the reasoning content part
|
||||
text:
|
||||
type: string
|
||||
description: Final complete reasoning text
|
||||
item_id:
|
||||
type: string
|
||||
description: >-
|
||||
Unique identifier of the completed output item
|
||||
output_index:
|
||||
type: integer
|
||||
description: >-
|
||||
Index position of the item in the output list
|
||||
sequence_number:
|
||||
type: integer
|
||||
description: >-
|
||||
Sequential number for ordering streaming events
|
||||
type:
|
||||
type: string
|
||||
const: response.reasoning_text.done
|
||||
default: response.reasoning_text.done
|
||||
description: >-
|
||||
Event type identifier, always "response.reasoning_text.done"
|
||||
additionalProperties: false
|
||||
required:
|
||||
- content_index
|
||||
- text
|
||||
- item_id
|
||||
- output_index
|
||||
- sequence_number
|
||||
- type
|
||||
title: >-
|
||||
OpenAIResponseObjectStreamResponseReasoningTextDone
|
||||
description: >-
|
||||
Streaming event for when reasoning text is completed.
|
||||
"OpenAIResponseObjectStreamResponseRefusalDelta":
|
||||
type: object
|
||||
properties:
|
||||
content_index:
|
||||
type: integer
|
||||
description: Index position of the content part
|
||||
delta:
|
||||
type: string
|
||||
description: Incremental refusal text being added
|
||||
item_id:
|
||||
type: string
|
||||
description: Unique identifier of the output item
|
||||
output_index:
|
||||
type: integer
|
||||
description: >-
|
||||
Index position of the item in the output list
|
||||
sequence_number:
|
||||
type: integer
|
||||
description: >-
|
||||
Sequential number for ordering streaming events
|
||||
type:
|
||||
type: string
|
||||
const: response.refusal.delta
|
||||
default: response.refusal.delta
|
||||
description: >-
|
||||
Event type identifier, always "response.refusal.delta"
|
||||
additionalProperties: false
|
||||
required:
|
||||
- content_index
|
||||
- delta
|
||||
- item_id
|
||||
- output_index
|
||||
- sequence_number
|
||||
- type
|
||||
title: >-
|
||||
OpenAIResponseObjectStreamResponseRefusalDelta
|
||||
description: >-
|
||||
Streaming event for incremental refusal text updates.
|
||||
"OpenAIResponseObjectStreamResponseRefusalDone":
|
||||
type: object
|
||||
properties:
|
||||
content_index:
|
||||
type: integer
|
||||
description: Index position of the content part
|
||||
refusal:
|
||||
type: string
|
||||
description: Final complete refusal text
|
||||
item_id:
|
||||
type: string
|
||||
description: Unique identifier of the output item
|
||||
output_index:
|
||||
type: integer
|
||||
description: >-
|
||||
Index position of the item in the output list
|
||||
sequence_number:
|
||||
type: integer
|
||||
description: >-
|
||||
Sequential number for ordering streaming events
|
||||
type:
|
||||
type: string
|
||||
const: response.refusal.done
|
||||
default: response.refusal.done
|
||||
description: >-
|
||||
Event type identifier, always "response.refusal.done"
|
||||
additionalProperties: false
|
||||
required:
|
||||
- content_index
|
||||
- refusal
|
||||
- item_id
|
||||
- output_index
|
||||
- sequence_number
|
||||
- type
|
||||
title: >-
|
||||
OpenAIResponseObjectStreamResponseRefusalDone
|
||||
description: >-
|
||||
Streaming event for when refusal text is completed.
|
||||
"OpenAIResponseObjectStreamResponseWebSearchCallCompleted":
|
||||
type: object
|
||||
properties:
|
||||
|
|
|
|||
Loading…
Add table
Add a link
Reference in a new issue