phoenix-oss/llama-stack-mirror
0
0
Fork 1
mirror of https://github.com/meta-llama/llama-stack.git synced 2025-08-07 02:58:21 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

add MetricResponseMixin to chat completion response types

Browse source
This commit is contained in:
Dinesh Yeduguru 2025-02-11 12:39:10 -08:00
parent 71cae67d7b
commit 183e9a08cc
4 changed files with 142 additions and 116 deletions
Download patch file Download diff file Expand all files Collapse all files

154
docs/_static/llama-stack-spec.html vendored
View file

@ -3106,6 +3106,12 @@
"ChatCompletionResponse": {
"type": "object",
"properties": {
"metrics": {
"type": "array",
"items": {
"$ref": "#/components/schemas/MetricEvent"
}
},
"completion_message": {
"$ref": "#/components/schemas/CompletionMessage",
"description": "The complete response message"
@ -3124,6 +3130,77 @@
],
"description": "Response from a chat completion request."
},
"MetricEvent": {
"type": "object",
"properties": {
"trace_id": {
"type": "string"
},
"span_id": {
"type": "string"
},
"timestamp": {
"type": "string",
"format": "date-time"
},
"attributes": {
"type": "object",
"additionalProperties": {
"oneOf": [
{
"type": "null"
},
{
"type": "boolean"
},
{
"type": "number"
},
{
"type": "string"
},
{
"type": "array"
},
{
"type": "object"
}
]
}
},
"type": {
"type": "string",
"const": "metric",
"default": "metric"
},
"metric": {
"type": "string"
},
"value": {
"oneOf": [
{
"type": "integer"
},
{
"type": "number"
}
]
},
"unit": {
"type": "string"
}
},
"additionalProperties": false,
"required": [
"trace_id",
"span_id",
"timestamp",
"type",
"metric",
"value",
"unit"
]
},
"TokenLogProbs": {
"type": "object",
"properties": {
@ -3388,6 +3465,12 @@
"ChatCompletionResponseStreamChunk": {
"type": "object",
"properties": {
"metrics": {
"type": "array",
"items": {
"$ref": "#/components/schemas/MetricEvent"
}
},
"event": {
"$ref": "#/components/schemas/ChatCompletionResponseEvent",
"description": "The event containing the new content"
@ -6374,77 +6457,6 @@
"critical"
]
},
"MetricEvent": {
"type": "object",
"properties": {
"trace_id": {
"type": "string"
},
"span_id": {
"type": "string"
},
"timestamp": {
"type": "string",
"format": "date-time"
},
"attributes": {
"type": "object",
"additionalProperties": {
"oneOf": [
{
"type": "null"
},
{
"type": "boolean"
},
{
"type": "number"
},
{
"type": "string"
},
{
"type": "array"
},
{
"type": "object"
}
]
}
},
"type": {
"type": "string",
"const": "metric",
"default": "metric"
},
"metric": {
"type": "string"
},
"value": {
"oneOf": [
{
"type": "integer"
},
{
"type": "number"
}
]
},
"unit": {
"type": "string"
}
},
"additionalProperties": false,
"required": [
"trace_id",
"span_id",
"timestamp",
"type",
"metric",
"value",
"unit"
]
},
"SpanEndPayload": {
"type": "object",
"properties": {

90
docs/_static/llama-stack-spec.yaml vendored
View file

@ -1925,6 +1925,10 @@ components:
ChatCompletionResponse:
type: object
properties:
metrics:
type: array
items:
$ref: '#/components/schemas/MetricEvent'
completion_message:
$ref: '#/components/schemas/CompletionMessage'
description: The complete response message
@ -1938,6 +1942,47 @@ components:
required:
- completion_message
description: Response from a chat completion request.
MetricEvent:
type: object
properties:
trace_id:
type: string
span_id:
type: string
timestamp:
type: string
format: date-time
attributes:
type: object
additionalProperties:
oneOf:
- type: 'null'
- type: boolean
- type: number
- type: string
- type: array
- type: object
type:
type: string
const: metric
default: metric
metric:
type: string
value:
oneOf:
- type: integer
- type: number
unit:
type: string
additionalProperties: false
required:
- trace_id
- span_id
- timestamp
- type
- metric
- value
- unit
TokenLogProbs:
type: object
properties:
@ -2173,6 +2218,10 @@ components:
ChatCompletionResponseStreamChunk:
type: object
properties:
metrics:
type: array
items:
$ref: '#/components/schemas/MetricEvent'
event:
$ref: '#/components/schemas/ChatCompletionResponseEvent'
description: The event containing the new content
@ -4070,47 +4119,6 @@ components:
- warn
- error
- critical
MetricEvent:
type: object
properties:
trace_id:
type: string
span_id:
type: string
timestamp:
type: string
format: date-time
attributes:
type: object
additionalProperties:
oneOf:
- type: 'null'
- type: boolean
- type: number
- type: string
- type: array
- type: object
type:
type: string
const: metric
default: metric
metric:
type: string
value:
oneOf:
- type: integer
- type: number
unit:
type: string
additionalProperties: false
required:
- trace_id
- span_id
- timestamp
- type
- metric
- value
- unit
SpanEndPayload:
type: object
properties:

7
llama_stack/apis/inference/inference.py
View file

@ -13,8 +13,8 @@ from typing import (
Literal,
Optional,
Protocol,
runtime_checkable,
Union,
runtime_checkable,
)
from llama_models.llama3.api.datatypes import (
@ -31,6 +31,7 @@ from typing_extensions import Annotated
from llama_stack.apis.common.content_types import ContentDelta, InterleavedContent
from llama_stack.apis.models import Model
from llama_stack.apis.telemetry.telemetry import MetricResponseMixin
from llama_stack.providers.utils.telemetry.trace_protocol import trace_protocol
@ -357,7 +358,7 @@ class ChatCompletionRequest(BaseModel):
@json_schema_type
class ChatCompletionResponseStreamChunk(BaseModel):
class ChatCompletionResponseStreamChunk(MetricResponseMixin, BaseModel):
"""A chunk of a streamed chat completion response.
:param event: The event containing the new content
@ -367,7 +368,7 @@ class ChatCompletionResponseStreamChunk(BaseModel):
@json_schema_type
class ChatCompletionResponse(BaseModel):
class ChatCompletionResponse(MetricResponseMixin, BaseModel):
"""Response from a chat completion request.
:param completion_message: The complete response message

7
llama_stack/apis/telemetry/telemetry.py
View file

@ -13,8 +13,8 @@ from typing import (
Literal,
Optional,
Protocol,
runtime_checkable,
Union,
runtime_checkable,
)
from llama_models.schema_utils import json_schema_type, register_schema, webmethod
@ -94,6 +94,11 @@ class MetricEvent(EventCommon):
unit: str
@json_schema_type
class MetricResponseMixin:
metrics: Optional[List[MetricEvent]] = None
@json_schema_type
class StructuredLogType(Enum):
SPAN_START = "span_start"