Braintrust CrossObject API
The CrossObject API from Braintrust — 1 operation(s) for crossobject.
OpenAPI Specification
openapi: 3.1.1
info:
version: 1.0.0
title: Braintrust Acls CrossObject API
description: 'API specification for the backend data server. The API is hosted globally at
https://api.braintrust.dev or in your own environment.
You can access the OpenAPI spec for this API at https://github.com/braintrustdata/braintrust-openapi.'
license:
name: Apache 2.0
servers:
- url: https://api.braintrust.dev
security:
- bearerAuth: []
- {}
tags:
- name: CrossObject
paths:
/v1/insert:
post:
operationId: postCrossObjectInsert
tags:
- CrossObject
description: Insert events and feedback across object types
summary: Cross-object insert
security:
- bearerAuth: []
requestBody:
description: A mapping from event object type -> object id -> events to insert
required: false
content:
application/json:
schema:
$ref: '#/components/schemas/CrossObjectInsertRequest'
responses:
'200':
description: Returns the inserted row ids for the events on each individual object
content:
application/json:
schema:
$ref: '#/components/schemas/CrossObjectInsertResponse'
'400':
description: The request was unacceptable, often due to missing a required parameter
content:
text/plain:
schema:
type: string
application/json:
schema:
nullable: true
'401':
description: No valid API key provided
content:
text/plain:
schema:
type: string
application/json:
schema:
nullable: true
'403':
description: The API key doesn’t have permissions to perform the request
content:
text/plain:
schema:
type: string
application/json:
schema:
nullable: true
'429':
description: Too many requests hit the API too quickly. We recommend an exponential backoff of your requests
headers:
Retry-After:
schema:
type: string
content:
text/plain:
schema:
type: string
application/json:
schema:
nullable: true
'500':
description: Something went wrong on Braintrust's end. (These are rare.)
content:
text/plain:
schema:
type: string
application/json:
schema:
nullable: true
components:
schemas:
SpanAttributes:
type: object
nullable: true
properties:
name:
type: string
nullable: true
description: Name of the span, for display purposes only
type:
$ref: '#/components/schemas/SpanType'
purpose:
type: string
nullable: true
enum:
- scorer
- null
description: A special value that indicates the span was generated by a scoring automation
additionalProperties:
nullable: true
description: Human-identifying attributes of the span, such as name, type, etc.
InsertDatasetEvent:
type: object
properties:
input:
nullable: true
description: The argument that uniquely define an input case (an arbitrary, JSON serializable object)
expected:
nullable: true
description: The output of your application, including post-processing (an arbitrary, JSON serializable object)
metadata:
type: object
nullable: true
properties:
model:
type: string
nullable: true
description: The model used for this example
additionalProperties:
nullable: true
description: A dictionary with additional data about the test example, model outputs, or just about anything else that's relevant, that you can use to help find and analyze examples later. For example, you could log the `prompt`, example's `id`, or anything else that would be useful to slice/dice later. The values in `metadata` can be any JSON-serializable type, but its keys must be strings
tags:
type: array
nullable: true
items:
type: string
description: A list of tags to log
id:
type: string
nullable: true
description: A unique identifier for the dataset event. If you don't provide one, Braintrust will generate one for you
created:
type: string
nullable: true
format: date-time
description: The timestamp the dataset event was created
origin:
$ref: '#/components/schemas/ObjectReferenceNullish'
facets:
type: object
nullable: true
additionalProperties:
nullable: true
description: Facets for categorization (dictionary from facet id to value)
_object_delete:
type: boolean
nullable: true
description: Pass `_object_delete=true` to mark the dataset event deleted. Deleted events will not show up in subsequent fetches for this dataset
_is_merge:
type: boolean
nullable: true
description: 'The `_is_merge` field controls how the row is merged with any existing row with the same id in the DB. By default (or when set to `false`), the existing row is completely replaced by the new row. When set to `true`, the new row is deep-merged into the existing row, if one is found. If no existing row is found, the new row is inserted as is.
For example, say there is an existing row in the DB `{"id": "foo", "input": {"a": 5, "b": 10}}`. If we merge a new row as `{"_is_merge": true, "id": "foo", "input": {"b": 11, "c": 20}}`, the new row will be `{"id": "foo", "input": {"a": 5, "b": 11, "c": 20}}`. If we replace the new row as `{"id": "foo", "input": {"b": 11, "c": 20}}`, the new row will be `{"id": "foo", "input": {"b": 11, "c": 20}}`'
_merge_paths:
type: array
nullable: true
items:
type: array
items:
type: string
description: 'The `_merge_paths` field allows controlling the depth of the merge, when `_is_merge=true`. `_merge_paths` is a list of paths, where each path is a list of field names. The deep merge will not descend below any of the specified merge paths.
For example, say there is an existing row in the DB `{"id": "foo", "input": {"a": {"b": 10}, "c": {"d": 20}}, "output": {"a": 20}}`. If we merge a new row as `{"_is_merge": true, "_merge_paths": [["input", "a"], ["output"]], "input": {"a": {"q": 30}, "c": {"e": 30}, "bar": "baz"}, "output": {"d": 40}}`, the new row will be `{"id": "foo": "input": {"a": {"q": 30}, "c": {"d": 20, "e": 30}, "bar": "baz"}, "output": {"d": 40}}`. In this case, due to the merge paths, we have replaced `input.a` and `output`, but have still deep-merged `input` and `input.c`.'
_array_delete:
type: array
nullable: true
items:
type: object
properties:
path:
type: array
items:
type: string
delete:
type: array
items:
nullable: true
required:
- path
- delete
description: 'The `_array_delete` field allows removing specific values from array fields. It is an array of objects with `path` and `delete` properties.
For example, to remove tags "foo" and "bar" from an existing row: `{"_is_merge": true, "_array_delete": [{"path": ["tags"], "delete": ["foo", "bar"]}]}`. For nested fields like `metadata.categories`, use `[{"path": ["metadata", "categories"], "delete": ["value"]}]`. This will remove those specific values from the array while preserving others.'
_parent_id:
type: string
nullable: true
description: 'DEPRECATED: The `_parent_id` field is deprecated and should not be used. Support for `_parent_id` will be dropped in a future version of Braintrust. Log `span_id`, `root_span_id`, and `span_parents` explicitly instead.
Use the `_parent_id` field to create this row as a subspan of an existing row. Tracking hierarchical relationships are important for tracing (see the [guide](https://www.braintrust.dev/docs/instrument) for full details).
For example, say we have logged a row `{"id": "abc", "input": "foo", "output": "bar", "expected": "boo", "scores": {"correctness": 0.33}}`. We can create a sub-span of the parent row by logging `{"_parent_id": "abc", "id": "llm_call", "input": {"prompt": "What comes after foo?"}, "output": "bar", "metrics": {"tokens": 1}}`. In the webapp, only the root span row `"abc"` will show up in the summary view. You can view the full trace hierarchy (in this case, the `"llm_call"` row) by clicking on the "abc" row.
If the row is being merged into an existing row, this field will be ignored.'
span_id:
type: string
nullable: true
description: 'Use `span_id`, `root_span_id`, and `span_parents` instead of `_parent_id`, which is now deprecated. The span_id is a unique identifier describing the row''s place in the a trace, and the root_span_id is a unique identifier for the whole trace. See the [guide](https://www.braintrust.dev/docs/instrument) for full details.
For example, say we have logged a row `{"id": "abc", "span_id": "span0", "root_span_id": "root_span0", "input": "foo", "output": "bar", "expected": "boo", "scores": {"correctness": 0.33}}`. We can create a sub-span of the parent row by logging `{"id": "llm_call", "span_id": "span1", "root_span_id": "root_span0", "span_parents": ["span0"], "input": {"prompt": "What comes after foo?"}, "output": "bar", "metrics": {"tokens": 1}}`. In the webapp, only the root span row `"abc"` will show up in the summary view. You can view the full trace hierarchy (in this case, the `"llm_call"` row) by clicking on the "abc" row.
If the row is being merged into an existing row, this field will be ignored.'
root_span_id:
type: string
nullable: true
description: 'Use `span_id`, `root_span_id`, and `span_parents` instead of `_parent_id`, which is now deprecated. The span_id is a unique identifier describing the row''s place in the a trace, and the root_span_id is a unique identifier for the whole trace. See the [guide](https://www.braintrust.dev/docs/instrument) for full details.
For example, say we have logged a row `{"id": "abc", "span_id": "span0", "root_span_id": "root_span0", "input": "foo", "output": "bar", "expected": "boo", "scores": {"correctness": 0.33}}`. We can create a sub-span of the parent row by logging `{"id": "llm_call", "span_id": "span1", "root_span_id": "root_span0", "span_parents": ["span0"], "input": {"prompt": "What comes after foo?"}, "output": "bar", "metrics": {"tokens": 1}}`. In the webapp, only the root span row `"abc"` will show up in the summary view. You can view the full trace hierarchy (in this case, the `"llm_call"` row) by clicking on the "abc" row.
If the row is being merged into an existing row, this field will be ignored.'
span_parents:
type: array
nullable: true
items:
type: string
description: 'Use `span_id`, `root_span_id`, and `span_parents` instead of `_parent_id`, which is now deprecated. The span_id is a unique identifier describing the row''s place in the a trace, and the root_span_id is a unique identifier for the whole trace. See the [guide](https://www.braintrust.dev/docs/instrument) for full details.
For example, say we have logged a row `{"id": "abc", "span_id": "span0", "root_span_id": "root_span0", "input": "foo", "output": "bar", "expected": "boo", "scores": {"correctness": 0.33}}`. We can create a sub-span of the parent row by logging `{"id": "llm_call", "span_id": "span1", "root_span_id": "root_span0", "span_parents": ["span0"], "input": {"prompt": "What comes after foo?"}, "output": "bar", "metrics": {"tokens": 1}}`. In the webapp, only the root span row `"abc"` will show up in the summary view. You can view the full trace hierarchy (in this case, the `"llm_call"` row) by clicking on the "abc" row.
If the row is being merged into an existing row, this field will be ignored.'
description: A dataset event
InsertProjectLogsEvent:
type: object
properties:
input:
nullable: true
description: The arguments that uniquely define a user input (an arbitrary, JSON serializable object).
output:
nullable: true
description: The output of your application, including post-processing (an arbitrary, JSON serializable object), that allows you to determine whether the result is correct or not. For example, in an app that generates SQL queries, the `output` should be the _result_ of the SQL query generated by the model, not the query itself, because there may be multiple valid queries that answer a single question.
expected:
nullable: true
description: The ground truth value (an arbitrary, JSON serializable object) that you'd compare to `output` to determine if your `output` value is correct or not. Braintrust currently does not compare `output` to `expected` for you, since there are so many different ways to do that correctly. Instead, these values are just used to help you navigate while digging into analyses. However, we may later use these values to re-score outputs or fine-tune your models.
error:
nullable: true
description: The error that occurred, if any.
scores:
type: object
nullable: true
additionalProperties:
type: number
nullable: true
minimum: 0
maximum: 1
description: A dictionary of numeric values (between 0 and 1) to log. The scores should give you a variety of signals that help you determine how accurate the outputs are compared to what you expect and diagnose failures. For example, a summarization app might have one score that tells you how accurate the summary is, and another that measures the word similarity between the generated and grouth truth summary. The word similarity score could help you determine whether the summarization was covering similar concepts or not. You can use these scores to help you sort, filter, and compare logs.
metadata:
type: object
nullable: true
properties:
model:
type: string
nullable: true
description: The model used for this example
additionalProperties:
nullable: true
description: A dictionary with additional data about the test example, model outputs, or just about anything else that's relevant, that you can use to help find and analyze examples later. For example, you could log the `prompt`, example's `id`, or anything else that would be useful to slice/dice later. The values in `metadata` can be any JSON-serializable type, but its keys must be strings
tags:
type: array
nullable: true
items:
type: string
description: A list of tags to log
metrics:
type: object
nullable: true
properties:
start:
type: number
nullable: true
description: A unix timestamp recording when the section of code which produced the project logs event started
end:
type: number
nullable: true
description: A unix timestamp recording when the section of code which produced the project logs event finished
prompt_tokens:
type: integer
nullable: true
description: The number of tokens in the prompt used to generate the project logs event (only set if this is an LLM span)
completion_tokens:
type: integer
nullable: true
description: The number of tokens in the completion generated by the model (only set if this is an LLM span)
tokens:
type: integer
nullable: true
description: The total number of tokens in the input and output of the project logs event.
caller_functionname:
nullable: true
description: This metric is deprecated
caller_filename:
nullable: true
description: This metric is deprecated
caller_lineno:
nullable: true
description: This metric is deprecated
additionalProperties:
type: number
description: Metrics are numerical measurements tracking the execution of the code that produced the project logs event. Use "start" and "end" to track the time span over which the project logs event was produced
context:
type: object
nullable: true
properties:
caller_functionname:
type: string
nullable: true
description: The function in code which created the project logs event
caller_filename:
type: string
nullable: true
description: Name of the file in code where the project logs event was created
caller_lineno:
type: integer
nullable: true
description: Line of code where the project logs event was created
additionalProperties:
nullable: true
description: Context is additional information about the code that produced the project logs event. It is essentially the textual counterpart to `metrics`. Use the `caller_*` attributes to track the location in code which produced the project logs event
span_attributes:
$ref: '#/components/schemas/SpanAttributes'
id:
type: string
nullable: true
description: A unique identifier for the project logs event. If you don't provide one, Braintrust will generate one for you
created:
type: string
nullable: true
format: date-time
description: The timestamp the project logs event was created
origin:
$ref: '#/components/schemas/ObjectReferenceNullish'
facets:
type: object
nullable: true
additionalProperties:
nullable: true
description: Facets for categorization (dictionary from facet id to value)
_object_delete:
type: boolean
nullable: true
description: Pass `_object_delete=true` to mark the project logs event deleted. Deleted events will not show up in subsequent fetches for this project logs
_is_merge:
type: boolean
nullable: true
description: 'The `_is_merge` field controls how the row is merged with any existing row with the same id in the DB. By default (or when set to `false`), the existing row is completely replaced by the new row. When set to `true`, the new row is deep-merged into the existing row, if one is found. If no existing row is found, the new row is inserted as is.
For example, say there is an existing row in the DB `{"id": "foo", "input": {"a": 5, "b": 10}}`. If we merge a new row as `{"_is_merge": true, "id": "foo", "input": {"b": 11, "c": 20}}`, the new row will be `{"id": "foo", "input": {"a": 5, "b": 11, "c": 20}}`. If we replace the new row as `{"id": "foo", "input": {"b": 11, "c": 20}}`, the new row will be `{"id": "foo", "input": {"b": 11, "c": 20}}`'
_merge_paths:
type: array
nullable: true
items:
type: array
items:
type: string
description: 'The `_merge_paths` field allows controlling the depth of the merge, when `_is_merge=true`. `_merge_paths` is a list of paths, where each path is a list of field names. The deep merge will not descend below any of the specified merge paths.
For example, say there is an existing row in the DB `{"id": "foo", "input": {"a": {"b": 10}, "c": {"d": 20}}, "output": {"a": 20}}`. If we merge a new row as `{"_is_merge": true, "_merge_paths": [["input", "a"], ["output"]], "input": {"a": {"q": 30}, "c": {"e": 30}, "bar": "baz"}, "output": {"d": 40}}`, the new row will be `{"id": "foo": "input": {"a": {"q": 30}, "c": {"d": 20, "e": 30}, "bar": "baz"}, "output": {"d": 40}}`. In this case, due to the merge paths, we have replaced `input.a` and `output`, but have still deep-merged `input` and `input.c`.'
_array_delete:
type: array
nullable: true
items:
type: object
properties:
path:
type: array
items:
type: string
delete:
type: array
items:
nullable: true
required:
- path
- delete
description: 'The `_array_delete` field allows removing specific values from array fields. It is an array of objects with `path` and `delete` properties.
For example, to remove tags "foo" and "bar" from an existing row: `{"_is_merge": true, "_array_delete": [{"path": ["tags"], "delete": ["foo", "bar"]}]}`. For nested fields like `metadata.categories`, use `[{"path": ["metadata", "categories"], "delete": ["value"]}]`. This will remove those specific values from the array while preserving others.'
_parent_id:
type: string
nullable: true
description: 'DEPRECATED: The `_parent_id` field is deprecated and should not be used. Support for `_parent_id` will be dropped in a future version of Braintrust. Log `span_id`, `root_span_id`, and `span_parents` explicitly instead.
Use the `_parent_id` field to create this row as a subspan of an existing row. Tracking hierarchical relationships are important for tracing (see the [guide](https://www.braintrust.dev/docs/instrument) for full details).
For example, say we have logged a row `{"id": "abc", "input": "foo", "output": "bar", "expected": "boo", "scores": {"correctness": 0.33}}`. We can create a sub-span of the parent row by logging `{"_parent_id": "abc", "id": "llm_call", "input": {"prompt": "What comes after foo?"}, "output": "bar", "metrics": {"tokens": 1}}`. In the webapp, only the root span row `"abc"` will show up in the summary view. You can view the full trace hierarchy (in this case, the `"llm_call"` row) by clicking on the "abc" row.
If the row is being merged into an existing row, this field will be ignored.'
span_id:
type: string
nullable: true
description: 'Use `span_id`, `root_span_id`, and `span_parents` instead of `_parent_id`, which is now deprecated. The span_id is a unique identifier describing the row''s place in the a trace, and the root_span_id is a unique identifier for the whole trace. See the [guide](https://www.braintrust.dev/docs/instrument) for full details.
For example, say we have logged a row `{"id": "abc", "span_id": "span0", "root_span_id": "root_span0", "input": "foo", "output": "bar", "expected": "boo", "scores": {"correctness": 0.33}}`. We can create a sub-span of the parent row by logging `{"id": "llm_call", "span_id": "span1", "root_span_id": "root_span0", "span_parents": ["span0"], "input": {"prompt": "What comes after foo?"}, "output": "bar", "metrics": {"tokens": 1}}`. In the webapp, only the root span row `"abc"` will show up in the summary view. You can view the full trace hierarchy (in this case, the `"llm_call"` row) by clicking on the "abc" row.
If the row is being merged into an existing row, this field will be ignored.'
root_span_id:
type: string
nullable: true
description: 'Use `span_id`, `root_span_id`, and `span_parents` instead of `_parent_id`, which is now deprecated. The span_id is a unique identifier describing the row''s place in the a trace, and the root_span_id is a unique identifier for the whole trace. See the [guide](https://www.braintrust.dev/docs/instrument) for full details.
For example, say we have logged a row `{"id": "abc", "span_id": "span0", "root_span_id": "root_span0", "input": "foo", "output": "bar", "expected": "boo", "scores": {"correctness": 0.33}}`. We can create a sub-span of the parent row by logging `{"id": "llm_call", "span_id": "span1", "root_span_id": "root_span0", "span_parents": ["span0"], "input": {"prompt": "What comes after foo?"}, "output": "bar", "metrics": {"tokens": 1}}`. In the webapp, only the root span row `"abc"` will show up in the summary view. You can view the full trace hierarchy (in this case, the `"llm_call"` row) by clicking on the "abc" row.
If the row is being merged into an existing row, this field will be ignored.'
span_parents:
type: array
nullable: true
items:
type: string
description: 'Use `span_id`, `root_span_id`, and `span_parents` instead of `_parent_id`, which is now deprecated. The span_id is a unique identifier describing the row''s place in the a trace, and the root_span_id is a unique identifier for the whole trace. See the [guide](https://www.braintrust.dev/docs/instrument) for full details.
For example, say we have logged a row `{"id": "abc", "span_id": "span0", "root_span_id": "root_span0", "input": "foo", "output": "bar", "expected": "boo", "scores": {"correctness": 0.33}}`. We can create a sub-span of the parent row by logging `{"id": "llm_call", "span_id": "span1", "root_span_id": "root_span0", "span_parents": ["span0"], "input": {"prompt": "What comes after foo?"}, "output": "bar", "metrics": {"tokens": 1}}`. In the webapp, only the root span row `"abc"` will show up in the summary view. You can view the full trace hierarchy (in this case, the `"llm_call"` row) by clicking on the "abc" row.
If the row is being merged into an existing row, this field will be ignored.'
description: A project logs event
FeedbackExperimentItem:
type: object
properties:
id:
type: string
description: The id of the experiment event to log feedback for. This is the row `id` returned by `POST /v1/experiment/{experiment_id}/insert`
scores:
type: object
nullable: true
additionalProperties:
type: number
nullable: true
minimum: 0
maximum: 1
description: A dictionary of numeric values (between 0 and 1) to log. These scores will be merged into the existing scores for the experiment event
expected:
nullable: true
description: The ground truth value (an arbitrary, JSON serializable object) that you'd compare to `output` to determine if your `output` value is correct or not
comment:
type: string
nullable: true
description: An optional comment string to log about the experiment event
metadata:
type: object
nullable: true
additionalProperties:
nullable: true
description: A dictionary with additional data about the feedback. If you have a `user_id`, you can log it here and access it in the Braintrust UI. Note, this metadata does not correspond to the main event itself, but rather the audit log attached to the event.
source:
type: string
nullable: true
enum:
- app
- api
- external
- null
description: The source of the feedback. Must be one of "external" (default), "app", or "api"
tags:
type: array
nullable: true
items:
type: string
description: A list of tags to log
required:
- id
InsertExperimentEvent:
type: object
properties:
input:
nullable: true
description: The arguments that uniquely define a test case (an arbitrary, JSON serializable object). Later on, Braintrust will use the `input` to know whether two test cases are the same between experiments, so they should not contain experiment-specific state. A simple rule of thumb is that if you run the same experiment twice, the `input` should be identical
output:
nullable: true
description: The output of your application, including post-processing (an arbitrary, JSON serializable object), that allows you to determine whether the result is correct or not. For example, in an app that generates SQL queries, the `output` should be the _result_ of the SQL query generated by the model, not the query itself, because there may be multiple valid queries that answer a single question
expected:
nullable: true
description: The ground truth value (an arbitrary, JSON serializable object) that you'd compare to `output` to determine if your `output` value is correct or not. Braintrust currently does not compare `output` to `expected` for you, since there are so many different ways to do that correctly. Instead, these values are just used to help you navigate your experiments while digging into analyses. However, we may later use these values to re-score outputs or fine-tune your models
error:
nullable: true
description: The error that occurred, if any.
scores:
type: object
nullable: true
additionalProperties:
type: number
nullable: true
minimum: 0
maximum: 1
description: A dictionary of numeric values (between 0 and 1) to log. The scores should give you a variety of signals that help you determine how accurate the outputs are compared to what you expect and diagnose failures. For example, a summarization app might have one score that tells you how accurate the summary is, and another that measures the word similarity between the generated and grouth truth summary. The word similarity score could help you determine whether the summarization was covering similar concepts or not. You can use these scores to help you sort, filter, and compare experiments
metadata:
type: object
nullable: true
properties:
model:
type: string
nullable: true
description: The model used for this example
additionalProperties:
nullable: true
description: A dictionary with additional data about the test example, model outputs, or just about anything else that's relevant, that you can use to help find and analyze examples later. For example, you could log the `prompt`, example's `id`, or anything else that would be useful to slice/dice later. The values in `metadata` can be any JSON-serializable type, but its keys must be strings
tags:
type: array
nullable: true
items:
type: string
description: A list of tags to log
metrics:
type: object
nullable: true
properties:
start:
type: number
nullable: true
description: A unix timestamp recording when the section of code which produced the experiment event started
end:
type: number
nullable: true
description: A unix timestamp recording when the section of code which produced the experiment event finished
prompt_tokens:
type: integer
nullable: true
description: The number of tokens in the prompt used to generate the experiment event (only set if this is an LLM span)
completion_tokens:
type: integer
nullable: true
description: The number of tokens in the completion generated by the model (only set if this is an LLM span)
tokens
# --- truncated at 32 KB (49 KB total) ---
# Full source: https://raw.githubusercontent.com/api-evangelist/braintrust/refs/heads/main/openapi/braintrust-crossobject-api-openapi.yml