deeprails 1.8.0__tar.gz → 1.10.0__tar.gz
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Potentially problematic release.
This version of deeprails might be problematic. Click here for more details.
- deeprails-1.10.0/.release-please-manifest.json +3 -0
- {deeprails-1.8.0 → deeprails-1.10.0}/CHANGELOG.md +17 -0
- {deeprails-1.8.0 → deeprails-1.10.0}/PKG-INFO +1 -1
- {deeprails-1.8.0 → deeprails-1.10.0}/api.md +5 -18
- {deeprails-1.8.0 → deeprails-1.10.0}/pyproject.toml +1 -1
- {deeprails-1.8.0 → deeprails-1.10.0}/src/deeprails/_client.py +1 -9
- {deeprails-1.8.0 → deeprails-1.10.0}/src/deeprails/_version.py +1 -1
- {deeprails-1.8.0 → deeprails-1.10.0}/src/deeprails/resources/__init__.py +0 -14
- {deeprails-1.8.0 → deeprails-1.10.0}/src/deeprails/resources/defend.py +8 -8
- {deeprails-1.8.0 → deeprails-1.10.0}/src/deeprails/resources/monitor.py +21 -21
- {deeprails-1.8.0 → deeprails-1.10.0}/src/deeprails/types/__init__.py +3 -5
- {deeprails-1.8.0 → deeprails-1.10.0}/src/deeprails/types/defend_create_workflow_params.py +1 -1
- {deeprails-1.8.0 → deeprails-1.10.0}/src/deeprails/types/defend_response.py +1 -1
- {deeprails-1.8.0 → deeprails-1.10.0}/src/deeprails/types/defend_submit_event_params.py +4 -3
- deeprails-1.8.0/src/deeprails/types/evaluation.py → deeprails-1.10.0/src/deeprails/types/monitor_detail_response.py +63 -5
- deeprails-1.8.0/src/deeprails/types/monitor_submit_event_response.py → deeprails-1.10.0/src/deeprails/types/monitor_event_response.py +2 -15
- deeprails-1.8.0/src/deeprails/types/api_response.py → deeprails-1.10.0/src/deeprails/types/monitor_response.py +2 -15
- {deeprails-1.8.0 → deeprails-1.10.0}/src/deeprails/types/monitor_submit_event_params.py +3 -2
- {deeprails-1.8.0 → deeprails-1.10.0}/tests/api_resources/test_defend.py +2 -2
- {deeprails-1.8.0 → deeprails-1.10.0}/tests/api_resources/test_monitor.py +35 -35
- deeprails-1.8.0/.release-please-manifest.json +0 -3
- deeprails-1.8.0/src/deeprails/resources/evaluate.py +0 -334
- deeprails-1.8.0/src/deeprails/types/evaluate_create_params.py +0 -62
- deeprails-1.8.0/src/deeprails/types/monitor_retrieve_response.py +0 -80
- deeprails-1.8.0/tests/api_resources/test_evaluate.py +0 -222
- {deeprails-1.8.0 → deeprails-1.10.0}/.gitignore +0 -0
- {deeprails-1.8.0 → deeprails-1.10.0}/CONTRIBUTING.md +0 -0
- {deeprails-1.8.0 → deeprails-1.10.0}/LICENSE +0 -0
- {deeprails-1.8.0 → deeprails-1.10.0}/README.md +0 -0
- {deeprails-1.8.0 → deeprails-1.10.0}/bin/check-release-environment +0 -0
- {deeprails-1.8.0 → deeprails-1.10.0}/bin/publish-pypi +0 -0
- {deeprails-1.8.0 → deeprails-1.10.0}/examples/.keep +0 -0
- {deeprails-1.8.0 → deeprails-1.10.0}/noxfile.py +0 -0
- {deeprails-1.8.0 → deeprails-1.10.0}/release-please-config.json +0 -0
- {deeprails-1.8.0 → deeprails-1.10.0}/requirements-dev.lock +0 -0
- {deeprails-1.8.0 → deeprails-1.10.0}/requirements.lock +0 -0
- {deeprails-1.8.0 → deeprails-1.10.0}/src/deeprails/__init__.py +0 -0
- {deeprails-1.8.0 → deeprails-1.10.0}/src/deeprails/_base_client.py +0 -0
- {deeprails-1.8.0 → deeprails-1.10.0}/src/deeprails/_compat.py +0 -0
- {deeprails-1.8.0 → deeprails-1.10.0}/src/deeprails/_constants.py +0 -0
- {deeprails-1.8.0 → deeprails-1.10.0}/src/deeprails/_exceptions.py +0 -0
- {deeprails-1.8.0 → deeprails-1.10.0}/src/deeprails/_files.py +0 -0
- {deeprails-1.8.0 → deeprails-1.10.0}/src/deeprails/_models.py +0 -0
- {deeprails-1.8.0 → deeprails-1.10.0}/src/deeprails/_qs.py +0 -0
- {deeprails-1.8.0 → deeprails-1.10.0}/src/deeprails/_resource.py +0 -0
- {deeprails-1.8.0 → deeprails-1.10.0}/src/deeprails/_response.py +0 -0
- {deeprails-1.8.0 → deeprails-1.10.0}/src/deeprails/_streaming.py +0 -0
- {deeprails-1.8.0 → deeprails-1.10.0}/src/deeprails/_types.py +0 -0
- {deeprails-1.8.0 → deeprails-1.10.0}/src/deeprails/_utils/__init__.py +0 -0
- {deeprails-1.8.0 → deeprails-1.10.0}/src/deeprails/_utils/_compat.py +0 -0
- {deeprails-1.8.0 → deeprails-1.10.0}/src/deeprails/_utils/_datetime_parse.py +0 -0
- {deeprails-1.8.0 → deeprails-1.10.0}/src/deeprails/_utils/_logs.py +0 -0
- {deeprails-1.8.0 → deeprails-1.10.0}/src/deeprails/_utils/_proxy.py +0 -0
- {deeprails-1.8.0 → deeprails-1.10.0}/src/deeprails/_utils/_reflection.py +0 -0
- {deeprails-1.8.0 → deeprails-1.10.0}/src/deeprails/_utils/_resources_proxy.py +0 -0
- {deeprails-1.8.0 → deeprails-1.10.0}/src/deeprails/_utils/_streams.py +0 -0
- {deeprails-1.8.0 → deeprails-1.10.0}/src/deeprails/_utils/_sync.py +0 -0
- {deeprails-1.8.0 → deeprails-1.10.0}/src/deeprails/_utils/_transform.py +0 -0
- {deeprails-1.8.0 → deeprails-1.10.0}/src/deeprails/_utils/_typing.py +0 -0
- {deeprails-1.8.0 → deeprails-1.10.0}/src/deeprails/_utils/_utils.py +0 -0
- {deeprails-1.8.0 → deeprails-1.10.0}/src/deeprails/lib/.keep +0 -0
- {deeprails-1.8.0 → deeprails-1.10.0}/src/deeprails/py.typed +0 -0
- {deeprails-1.8.0 → deeprails-1.10.0}/src/deeprails/types/defend_update_workflow_params.py +0 -0
- {deeprails-1.8.0 → deeprails-1.10.0}/src/deeprails/types/monitor_create_params.py +0 -0
- {deeprails-1.8.0 → deeprails-1.10.0}/src/deeprails/types/monitor_retrieve_params.py +0 -0
- {deeprails-1.8.0 → deeprails-1.10.0}/src/deeprails/types/monitor_update_params.py +0 -0
- {deeprails-1.8.0 → deeprails-1.10.0}/src/deeprails/types/workflow_event_response.py +0 -0
- {deeprails-1.8.0 → deeprails-1.10.0}/tests/__init__.py +0 -0
- {deeprails-1.8.0 → deeprails-1.10.0}/tests/api_resources/__init__.py +0 -0
- {deeprails-1.8.0 → deeprails-1.10.0}/tests/conftest.py +0 -0
- {deeprails-1.8.0 → deeprails-1.10.0}/tests/sample_file.txt +0 -0
- {deeprails-1.8.0 → deeprails-1.10.0}/tests/test_client.py +0 -0
- {deeprails-1.8.0 → deeprails-1.10.0}/tests/test_deepcopy.py +0 -0
- {deeprails-1.8.0 → deeprails-1.10.0}/tests/test_extract_files.py +0 -0
- {deeprails-1.8.0 → deeprails-1.10.0}/tests/test_files.py +0 -0
- {deeprails-1.8.0 → deeprails-1.10.0}/tests/test_models.py +0 -0
- {deeprails-1.8.0 → deeprails-1.10.0}/tests/test_qs.py +0 -0
- {deeprails-1.8.0 → deeprails-1.10.0}/tests/test_required_args.py +0 -0
- {deeprails-1.8.0 → deeprails-1.10.0}/tests/test_response.py +0 -0
- {deeprails-1.8.0 → deeprails-1.10.0}/tests/test_streaming.py +0 -0
- {deeprails-1.8.0 → deeprails-1.10.0}/tests/test_transform.py +0 -0
- {deeprails-1.8.0 → deeprails-1.10.0}/tests/test_utils/test_datetime_parse.py +0 -0
- {deeprails-1.8.0 → deeprails-1.10.0}/tests/test_utils/test_proxy.py +0 -0
- {deeprails-1.8.0 → deeprails-1.10.0}/tests/test_utils/test_typing.py +0 -0
- {deeprails-1.8.0 → deeprails-1.10.0}/tests/utils.py +0 -0
|
@@ -1,5 +1,22 @@
|
|
|
1
1
|
# Changelog
|
|
2
2
|
|
|
3
|
+
## 1.10.0 (2025-10-29)
|
|
4
|
+
|
|
5
|
+
Full Changelog: [v1.9.0...v1.10.0](https://github.com/deeprails/deeprails-sdk-python/compare/v1.9.0...v1.10.0)
|
|
6
|
+
|
|
7
|
+
### Features
|
|
8
|
+
|
|
9
|
+
* **api:** remove evaluate api ([a3ddec1](https://github.com/deeprails/deeprails-sdk-python/commit/a3ddec1696eaa1247eea2a3bffd61e63d6537d30))
|
|
10
|
+
* **api:** remove evaluate references ([23519a3](https://github.com/deeprails/deeprails-sdk-python/commit/23519a3349a254fe2fe24a51aeb59545d0820b70))
|
|
11
|
+
|
|
12
|
+
## 1.9.0 (2025-10-24)
|
|
13
|
+
|
|
14
|
+
Full Changelog: [v1.8.0...v1.9.0](https://github.com/deeprails/deeprails-sdk-python/compare/v1.8.0...v1.9.0)
|
|
15
|
+
|
|
16
|
+
### Features
|
|
17
|
+
|
|
18
|
+
* **api:** remove apiresponse from monitor ([4971a99](https://github.com/deeprails/deeprails-sdk-python/commit/4971a99c7357bebbc5e86a2d76d2be55bb34f5ae))
|
|
19
|
+
|
|
3
20
|
## 1.8.0 (2025-10-22)
|
|
4
21
|
|
|
5
22
|
Full Changelog: [v1.7.0...v1.8.0](https://github.com/deeprails/deeprails-sdk-python/compare/v1.7.0...v1.8.0)
|
|
@@ -19,25 +19,12 @@ Methods:
|
|
|
19
19
|
Types:
|
|
20
20
|
|
|
21
21
|
```python
|
|
22
|
-
from deeprails.types import
|
|
22
|
+
from deeprails.types import MonitorDetailResponse, MonitorEventResponse, MonitorResponse
|
|
23
23
|
```
|
|
24
24
|
|
|
25
25
|
Methods:
|
|
26
26
|
|
|
27
|
-
- <code title="post /monitor">client.monitor.<a href="./src/deeprails/resources/monitor.py">create</a>(\*\*<a href="src/deeprails/types/monitor_create_params.py">params</a>) -> <a href="./src/deeprails/types/
|
|
28
|
-
- <code title="get /monitor/{monitor_id}">client.monitor.<a href="./src/deeprails/resources/monitor.py">retrieve</a>(monitor_id, \*\*<a href="src/deeprails/types/monitor_retrieve_params.py">params</a>) -> <a href="./src/deeprails/types/
|
|
29
|
-
- <code title="put /monitor/{monitor_id}">client.monitor.<a href="./src/deeprails/resources/monitor.py">update</a>(monitor_id, \*\*<a href="src/deeprails/types/monitor_update_params.py">params</a>) -> <a href="./src/deeprails/types/
|
|
30
|
-
- <code title="post /monitor/{monitor_id}/events">client.monitor.<a href="./src/deeprails/resources/monitor.py">submit_event</a>(monitor_id, \*\*<a href="src/deeprails/types/monitor_submit_event_params.py">params</a>) -> <a href="./src/deeprails/types/
|
|
31
|
-
|
|
32
|
-
# Evaluate
|
|
33
|
-
|
|
34
|
-
Types:
|
|
35
|
-
|
|
36
|
-
```python
|
|
37
|
-
from deeprails.types import Evaluation
|
|
38
|
-
```
|
|
39
|
-
|
|
40
|
-
Methods:
|
|
41
|
-
|
|
42
|
-
- <code title="post /evaluate">client.evaluate.<a href="./src/deeprails/resources/evaluate.py">create</a>(\*\*<a href="src/deeprails/types/evaluate_create_params.py">params</a>) -> <a href="./src/deeprails/types/evaluation.py">Evaluation</a></code>
|
|
43
|
-
- <code title="get /evaluate/{eval_id}">client.evaluate.<a href="./src/deeprails/resources/evaluate.py">retrieve</a>(eval_id) -> <a href="./src/deeprails/types/evaluation.py">Evaluation</a></code>
|
|
27
|
+
- <code title="post /monitor">client.monitor.<a href="./src/deeprails/resources/monitor.py">create</a>(\*\*<a href="src/deeprails/types/monitor_create_params.py">params</a>) -> <a href="./src/deeprails/types/monitor_response.py">MonitorResponse</a></code>
|
|
28
|
+
- <code title="get /monitor/{monitor_id}">client.monitor.<a href="./src/deeprails/resources/monitor.py">retrieve</a>(monitor_id, \*\*<a href="src/deeprails/types/monitor_retrieve_params.py">params</a>) -> <a href="./src/deeprails/types/monitor_detail_response.py">MonitorDetailResponse</a></code>
|
|
29
|
+
- <code title="put /monitor/{monitor_id}">client.monitor.<a href="./src/deeprails/resources/monitor.py">update</a>(monitor_id, \*\*<a href="src/deeprails/types/monitor_update_params.py">params</a>) -> <a href="./src/deeprails/types/monitor_response.py">MonitorResponse</a></code>
|
|
30
|
+
- <code title="post /monitor/{monitor_id}/events">client.monitor.<a href="./src/deeprails/resources/monitor.py">submit_event</a>(monitor_id, \*\*<a href="src/deeprails/types/monitor_submit_event_params.py">params</a>) -> <a href="./src/deeprails/types/monitor_event_response.py">MonitorEventResponse</a></code>
|
|
@@ -21,7 +21,7 @@ from ._types import (
|
|
|
21
21
|
)
|
|
22
22
|
from ._utils import is_given, get_async_library
|
|
23
23
|
from ._version import __version__
|
|
24
|
-
from .resources import defend, monitor
|
|
24
|
+
from .resources import defend, monitor
|
|
25
25
|
from ._streaming import Stream as Stream, AsyncStream as AsyncStream
|
|
26
26
|
from ._exceptions import APIStatusError, DeeprailsError
|
|
27
27
|
from ._base_client import (
|
|
@@ -45,7 +45,6 @@ __all__ = [
|
|
|
45
45
|
class Deeprails(SyncAPIClient):
|
|
46
46
|
defend: defend.DefendResource
|
|
47
47
|
monitor: monitor.MonitorResource
|
|
48
|
-
evaluate: evaluate.EvaluateResource
|
|
49
48
|
with_raw_response: DeeprailsWithRawResponse
|
|
50
49
|
with_streaming_response: DeeprailsWithStreamedResponse
|
|
51
50
|
|
|
@@ -105,7 +104,6 @@ class Deeprails(SyncAPIClient):
|
|
|
105
104
|
|
|
106
105
|
self.defend = defend.DefendResource(self)
|
|
107
106
|
self.monitor = monitor.MonitorResource(self)
|
|
108
|
-
self.evaluate = evaluate.EvaluateResource(self)
|
|
109
107
|
self.with_raw_response = DeeprailsWithRawResponse(self)
|
|
110
108
|
self.with_streaming_response = DeeprailsWithStreamedResponse(self)
|
|
111
109
|
|
|
@@ -217,7 +215,6 @@ class Deeprails(SyncAPIClient):
|
|
|
217
215
|
class AsyncDeeprails(AsyncAPIClient):
|
|
218
216
|
defend: defend.AsyncDefendResource
|
|
219
217
|
monitor: monitor.AsyncMonitorResource
|
|
220
|
-
evaluate: evaluate.AsyncEvaluateResource
|
|
221
218
|
with_raw_response: AsyncDeeprailsWithRawResponse
|
|
222
219
|
with_streaming_response: AsyncDeeprailsWithStreamedResponse
|
|
223
220
|
|
|
@@ -277,7 +274,6 @@ class AsyncDeeprails(AsyncAPIClient):
|
|
|
277
274
|
|
|
278
275
|
self.defend = defend.AsyncDefendResource(self)
|
|
279
276
|
self.monitor = monitor.AsyncMonitorResource(self)
|
|
280
|
-
self.evaluate = evaluate.AsyncEvaluateResource(self)
|
|
281
277
|
self.with_raw_response = AsyncDeeprailsWithRawResponse(self)
|
|
282
278
|
self.with_streaming_response = AsyncDeeprailsWithStreamedResponse(self)
|
|
283
279
|
|
|
@@ -390,28 +386,24 @@ class DeeprailsWithRawResponse:
|
|
|
390
386
|
def __init__(self, client: Deeprails) -> None:
|
|
391
387
|
self.defend = defend.DefendResourceWithRawResponse(client.defend)
|
|
392
388
|
self.monitor = monitor.MonitorResourceWithRawResponse(client.monitor)
|
|
393
|
-
self.evaluate = evaluate.EvaluateResourceWithRawResponse(client.evaluate)
|
|
394
389
|
|
|
395
390
|
|
|
396
391
|
class AsyncDeeprailsWithRawResponse:
|
|
397
392
|
def __init__(self, client: AsyncDeeprails) -> None:
|
|
398
393
|
self.defend = defend.AsyncDefendResourceWithRawResponse(client.defend)
|
|
399
394
|
self.monitor = monitor.AsyncMonitorResourceWithRawResponse(client.monitor)
|
|
400
|
-
self.evaluate = evaluate.AsyncEvaluateResourceWithRawResponse(client.evaluate)
|
|
401
395
|
|
|
402
396
|
|
|
403
397
|
class DeeprailsWithStreamedResponse:
|
|
404
398
|
def __init__(self, client: Deeprails) -> None:
|
|
405
399
|
self.defend = defend.DefendResourceWithStreamingResponse(client.defend)
|
|
406
400
|
self.monitor = monitor.MonitorResourceWithStreamingResponse(client.monitor)
|
|
407
|
-
self.evaluate = evaluate.EvaluateResourceWithStreamingResponse(client.evaluate)
|
|
408
401
|
|
|
409
402
|
|
|
410
403
|
class AsyncDeeprailsWithStreamedResponse:
|
|
411
404
|
def __init__(self, client: AsyncDeeprails) -> None:
|
|
412
405
|
self.defend = defend.AsyncDefendResourceWithStreamingResponse(client.defend)
|
|
413
406
|
self.monitor = monitor.AsyncMonitorResourceWithStreamingResponse(client.monitor)
|
|
414
|
-
self.evaluate = evaluate.AsyncEvaluateResourceWithStreamingResponse(client.evaluate)
|
|
415
407
|
|
|
416
408
|
|
|
417
409
|
Client = Deeprails
|
|
@@ -16,14 +16,6 @@ from .monitor import (
|
|
|
16
16
|
MonitorResourceWithStreamingResponse,
|
|
17
17
|
AsyncMonitorResourceWithStreamingResponse,
|
|
18
18
|
)
|
|
19
|
-
from .evaluate import (
|
|
20
|
-
EvaluateResource,
|
|
21
|
-
AsyncEvaluateResource,
|
|
22
|
-
EvaluateResourceWithRawResponse,
|
|
23
|
-
AsyncEvaluateResourceWithRawResponse,
|
|
24
|
-
EvaluateResourceWithStreamingResponse,
|
|
25
|
-
AsyncEvaluateResourceWithStreamingResponse,
|
|
26
|
-
)
|
|
27
19
|
|
|
28
20
|
__all__ = [
|
|
29
21
|
"DefendResource",
|
|
@@ -38,10 +30,4 @@ __all__ = [
|
|
|
38
30
|
"AsyncMonitorResourceWithRawResponse",
|
|
39
31
|
"MonitorResourceWithStreamingResponse",
|
|
40
32
|
"AsyncMonitorResourceWithStreamingResponse",
|
|
41
|
-
"EvaluateResource",
|
|
42
|
-
"AsyncEvaluateResource",
|
|
43
|
-
"EvaluateResourceWithRawResponse",
|
|
44
|
-
"AsyncEvaluateResourceWithRawResponse",
|
|
45
|
-
"EvaluateResourceWithStreamingResponse",
|
|
46
|
-
"AsyncEvaluateResourceWithStreamingResponse",
|
|
47
33
|
]
|
|
@@ -54,7 +54,7 @@ class DefendResource(SyncAPIResource):
|
|
|
54
54
|
automatic_hallucination_tolerance_levels: Dict[str, Literal["low", "medium", "high"]] | Omit = omit,
|
|
55
55
|
custom_hallucination_threshold_values: Dict[str, float] | Omit = omit,
|
|
56
56
|
description: str | Omit = omit,
|
|
57
|
-
|
|
57
|
+
max_improvement_attempts: int | Omit = omit,
|
|
58
58
|
# Use the following arguments if you need to pass additional parameters to the API that aren't available via kwargs.
|
|
59
59
|
# The extra values given here take precedence over values defined on the client or passed to this method.
|
|
60
60
|
extra_headers: Headers | None = None,
|
|
@@ -92,7 +92,7 @@ class DefendResource(SyncAPIResource):
|
|
|
92
92
|
|
|
93
93
|
description: Description for the workflow.
|
|
94
94
|
|
|
95
|
-
|
|
95
|
+
max_improvement_attempts: Max. number of improvement action retries until a given event passes the
|
|
96
96
|
guardrails. Defaults to 10.
|
|
97
97
|
|
|
98
98
|
extra_headers: Send extra headers
|
|
@@ -113,7 +113,7 @@ class DefendResource(SyncAPIResource):
|
|
|
113
113
|
"automatic_hallucination_tolerance_levels": automatic_hallucination_tolerance_levels,
|
|
114
114
|
"custom_hallucination_threshold_values": custom_hallucination_threshold_values,
|
|
115
115
|
"description": description,
|
|
116
|
-
"
|
|
116
|
+
"max_improvement_attempts": max_improvement_attempts,
|
|
117
117
|
},
|
|
118
118
|
defend_create_workflow_params.DefendCreateWorkflowParams,
|
|
119
119
|
),
|
|
@@ -214,7 +214,7 @@ class DefendResource(SyncAPIResource):
|
|
|
214
214
|
|
|
215
215
|
Args:
|
|
216
216
|
model_input: A dictionary of inputs sent to the LLM to generate output. The dictionary must
|
|
217
|
-
contain at least `user_prompt` or `system_prompt` field. For the
|
|
217
|
+
contain at least a `user_prompt` field or a `system_prompt` field. For the
|
|
218
218
|
ground_truth_adherence guardrail metric, `ground_truth` should be provided.
|
|
219
219
|
|
|
220
220
|
model_output: Output generated by the LLM to be evaluated.
|
|
@@ -332,7 +332,7 @@ class AsyncDefendResource(AsyncAPIResource):
|
|
|
332
332
|
automatic_hallucination_tolerance_levels: Dict[str, Literal["low", "medium", "high"]] | Omit = omit,
|
|
333
333
|
custom_hallucination_threshold_values: Dict[str, float] | Omit = omit,
|
|
334
334
|
description: str | Omit = omit,
|
|
335
|
-
|
|
335
|
+
max_improvement_attempts: int | Omit = omit,
|
|
336
336
|
# Use the following arguments if you need to pass additional parameters to the API that aren't available via kwargs.
|
|
337
337
|
# The extra values given here take precedence over values defined on the client or passed to this method.
|
|
338
338
|
extra_headers: Headers | None = None,
|
|
@@ -370,7 +370,7 @@ class AsyncDefendResource(AsyncAPIResource):
|
|
|
370
370
|
|
|
371
371
|
description: Description for the workflow.
|
|
372
372
|
|
|
373
|
-
|
|
373
|
+
max_improvement_attempts: Max. number of improvement action retries until a given event passes the
|
|
374
374
|
guardrails. Defaults to 10.
|
|
375
375
|
|
|
376
376
|
extra_headers: Send extra headers
|
|
@@ -391,7 +391,7 @@ class AsyncDefendResource(AsyncAPIResource):
|
|
|
391
391
|
"automatic_hallucination_tolerance_levels": automatic_hallucination_tolerance_levels,
|
|
392
392
|
"custom_hallucination_threshold_values": custom_hallucination_threshold_values,
|
|
393
393
|
"description": description,
|
|
394
|
-
"
|
|
394
|
+
"max_improvement_attempts": max_improvement_attempts,
|
|
395
395
|
},
|
|
396
396
|
defend_create_workflow_params.DefendCreateWorkflowParams,
|
|
397
397
|
),
|
|
@@ -492,7 +492,7 @@ class AsyncDefendResource(AsyncAPIResource):
|
|
|
492
492
|
|
|
493
493
|
Args:
|
|
494
494
|
model_input: A dictionary of inputs sent to the LLM to generate output. The dictionary must
|
|
495
|
-
contain at least `user_prompt` or `system_prompt` field. For the
|
|
495
|
+
contain at least a `user_prompt` field or a `system_prompt` field. For the
|
|
496
496
|
ground_truth_adherence guardrail metric, `ground_truth` should be provided.
|
|
497
497
|
|
|
498
498
|
model_output: Output generated by the LLM to be evaluated.
|
|
@@ -19,9 +19,9 @@ from .._response import (
|
|
|
19
19
|
async_to_streamed_response_wrapper,
|
|
20
20
|
)
|
|
21
21
|
from .._base_client import make_request_options
|
|
22
|
-
from ..types.
|
|
23
|
-
from ..types.
|
|
24
|
-
from ..types.
|
|
22
|
+
from ..types.monitor_response import MonitorResponse
|
|
23
|
+
from ..types.monitor_event_response import MonitorEventResponse
|
|
24
|
+
from ..types.monitor_detail_response import MonitorDetailResponse
|
|
25
25
|
|
|
26
26
|
__all__ = ["MonitorResource", "AsyncMonitorResource"]
|
|
27
27
|
|
|
@@ -57,7 +57,7 @@ class MonitorResource(SyncAPIResource):
|
|
|
57
57
|
extra_query: Query | None = None,
|
|
58
58
|
extra_body: Body | None = None,
|
|
59
59
|
timeout: float | httpx.Timeout | None | NotGiven = not_given,
|
|
60
|
-
) ->
|
|
60
|
+
) -> MonitorResponse:
|
|
61
61
|
"""
|
|
62
62
|
Use this endpoint to create a new monitor to evaluate model inputs and outputs
|
|
63
63
|
using guardrails
|
|
@@ -87,7 +87,7 @@ class MonitorResource(SyncAPIResource):
|
|
|
87
87
|
options=make_request_options(
|
|
88
88
|
extra_headers=extra_headers, extra_query=extra_query, extra_body=extra_body, timeout=timeout
|
|
89
89
|
),
|
|
90
|
-
cast_to=
|
|
90
|
+
cast_to=MonitorResponse,
|
|
91
91
|
)
|
|
92
92
|
|
|
93
93
|
def retrieve(
|
|
@@ -101,7 +101,7 @@ class MonitorResource(SyncAPIResource):
|
|
|
101
101
|
extra_query: Query | None = None,
|
|
102
102
|
extra_body: Body | None = None,
|
|
103
103
|
timeout: float | httpx.Timeout | None | NotGiven = not_given,
|
|
104
|
-
) ->
|
|
104
|
+
) -> MonitorDetailResponse:
|
|
105
105
|
"""
|
|
106
106
|
Use this endpoint to retrieve the details and evaluations associated with a
|
|
107
107
|
specific monitor
|
|
@@ -128,7 +128,7 @@ class MonitorResource(SyncAPIResource):
|
|
|
128
128
|
timeout=timeout,
|
|
129
129
|
query=maybe_transform({"limit": limit}, monitor_retrieve_params.MonitorRetrieveParams),
|
|
130
130
|
),
|
|
131
|
-
cast_to=
|
|
131
|
+
cast_to=MonitorDetailResponse,
|
|
132
132
|
)
|
|
133
133
|
|
|
134
134
|
def update(
|
|
@@ -144,7 +144,7 @@ class MonitorResource(SyncAPIResource):
|
|
|
144
144
|
extra_query: Query | None = None,
|
|
145
145
|
extra_body: Body | None = None,
|
|
146
146
|
timeout: float | httpx.Timeout | None | NotGiven = not_given,
|
|
147
|
-
) ->
|
|
147
|
+
) -> MonitorResponse:
|
|
148
148
|
"""
|
|
149
149
|
Use this endpoint to update the name, description, or status of an existing
|
|
150
150
|
monitor
|
|
@@ -180,7 +180,7 @@ class MonitorResource(SyncAPIResource):
|
|
|
180
180
|
options=make_request_options(
|
|
181
181
|
extra_headers=extra_headers, extra_query=extra_query, extra_body=extra_body, timeout=timeout
|
|
182
182
|
),
|
|
183
|
-
cast_to=
|
|
183
|
+
cast_to=MonitorResponse,
|
|
184
184
|
)
|
|
185
185
|
|
|
186
186
|
def submit_event(
|
|
@@ -208,7 +208,7 @@ class MonitorResource(SyncAPIResource):
|
|
|
208
208
|
extra_query: Query | None = None,
|
|
209
209
|
extra_body: Body | None = None,
|
|
210
210
|
timeout: float | httpx.Timeout | None | NotGiven = not_given,
|
|
211
|
-
) ->
|
|
211
|
+
) -> MonitorEventResponse:
|
|
212
212
|
"""
|
|
213
213
|
Use this endpoint to submit a model input and output pair to a monitor for
|
|
214
214
|
evaluation
|
|
@@ -220,7 +220,7 @@ class MonitorResource(SyncAPIResource):
|
|
|
220
220
|
`ground_truth_adherence`, and/or `comprehensive_safety`.
|
|
221
221
|
|
|
222
222
|
model_input: A dictionary of inputs sent to the LLM to generate output. The dictionary must
|
|
223
|
-
contain at least a `user_prompt` or `system_prompt` field. For
|
|
223
|
+
contain at least a `user_prompt` field or a `system_prompt` field. For
|
|
224
224
|
ground_truth_adherence guardrail metric, `ground_truth` should be provided.
|
|
225
225
|
|
|
226
226
|
model_output: Output generated by the LLM to be evaluated.
|
|
@@ -260,7 +260,7 @@ class MonitorResource(SyncAPIResource):
|
|
|
260
260
|
options=make_request_options(
|
|
261
261
|
extra_headers=extra_headers, extra_query=extra_query, extra_body=extra_body, timeout=timeout
|
|
262
262
|
),
|
|
263
|
-
cast_to=
|
|
263
|
+
cast_to=MonitorEventResponse,
|
|
264
264
|
)
|
|
265
265
|
|
|
266
266
|
|
|
@@ -295,7 +295,7 @@ class AsyncMonitorResource(AsyncAPIResource):
|
|
|
295
295
|
extra_query: Query | None = None,
|
|
296
296
|
extra_body: Body | None = None,
|
|
297
297
|
timeout: float | httpx.Timeout | None | NotGiven = not_given,
|
|
298
|
-
) ->
|
|
298
|
+
) -> MonitorResponse:
|
|
299
299
|
"""
|
|
300
300
|
Use this endpoint to create a new monitor to evaluate model inputs and outputs
|
|
301
301
|
using guardrails
|
|
@@ -325,7 +325,7 @@ class AsyncMonitorResource(AsyncAPIResource):
|
|
|
325
325
|
options=make_request_options(
|
|
326
326
|
extra_headers=extra_headers, extra_query=extra_query, extra_body=extra_body, timeout=timeout
|
|
327
327
|
),
|
|
328
|
-
cast_to=
|
|
328
|
+
cast_to=MonitorResponse,
|
|
329
329
|
)
|
|
330
330
|
|
|
331
331
|
async def retrieve(
|
|
@@ -339,7 +339,7 @@ class AsyncMonitorResource(AsyncAPIResource):
|
|
|
339
339
|
extra_query: Query | None = None,
|
|
340
340
|
extra_body: Body | None = None,
|
|
341
341
|
timeout: float | httpx.Timeout | None | NotGiven = not_given,
|
|
342
|
-
) ->
|
|
342
|
+
) -> MonitorDetailResponse:
|
|
343
343
|
"""
|
|
344
344
|
Use this endpoint to retrieve the details and evaluations associated with a
|
|
345
345
|
specific monitor
|
|
@@ -366,7 +366,7 @@ class AsyncMonitorResource(AsyncAPIResource):
|
|
|
366
366
|
timeout=timeout,
|
|
367
367
|
query=await async_maybe_transform({"limit": limit}, monitor_retrieve_params.MonitorRetrieveParams),
|
|
368
368
|
),
|
|
369
|
-
cast_to=
|
|
369
|
+
cast_to=MonitorDetailResponse,
|
|
370
370
|
)
|
|
371
371
|
|
|
372
372
|
async def update(
|
|
@@ -382,7 +382,7 @@ class AsyncMonitorResource(AsyncAPIResource):
|
|
|
382
382
|
extra_query: Query | None = None,
|
|
383
383
|
extra_body: Body | None = None,
|
|
384
384
|
timeout: float | httpx.Timeout | None | NotGiven = not_given,
|
|
385
|
-
) ->
|
|
385
|
+
) -> MonitorResponse:
|
|
386
386
|
"""
|
|
387
387
|
Use this endpoint to update the name, description, or status of an existing
|
|
388
388
|
monitor
|
|
@@ -418,7 +418,7 @@ class AsyncMonitorResource(AsyncAPIResource):
|
|
|
418
418
|
options=make_request_options(
|
|
419
419
|
extra_headers=extra_headers, extra_query=extra_query, extra_body=extra_body, timeout=timeout
|
|
420
420
|
),
|
|
421
|
-
cast_to=
|
|
421
|
+
cast_to=MonitorResponse,
|
|
422
422
|
)
|
|
423
423
|
|
|
424
424
|
async def submit_event(
|
|
@@ -446,7 +446,7 @@ class AsyncMonitorResource(AsyncAPIResource):
|
|
|
446
446
|
extra_query: Query | None = None,
|
|
447
447
|
extra_body: Body | None = None,
|
|
448
448
|
timeout: float | httpx.Timeout | None | NotGiven = not_given,
|
|
449
|
-
) ->
|
|
449
|
+
) -> MonitorEventResponse:
|
|
450
450
|
"""
|
|
451
451
|
Use this endpoint to submit a model input and output pair to a monitor for
|
|
452
452
|
evaluation
|
|
@@ -458,7 +458,7 @@ class AsyncMonitorResource(AsyncAPIResource):
|
|
|
458
458
|
`ground_truth_adherence`, and/or `comprehensive_safety`.
|
|
459
459
|
|
|
460
460
|
model_input: A dictionary of inputs sent to the LLM to generate output. The dictionary must
|
|
461
|
-
contain at least a `user_prompt` or `system_prompt` field. For
|
|
461
|
+
contain at least a `user_prompt` field or a `system_prompt` field. For
|
|
462
462
|
ground_truth_adherence guardrail metric, `ground_truth` should be provided.
|
|
463
463
|
|
|
464
464
|
model_output: Output generated by the LLM to be evaluated.
|
|
@@ -498,7 +498,7 @@ class AsyncMonitorResource(AsyncAPIResource):
|
|
|
498
498
|
options=make_request_options(
|
|
499
499
|
extra_headers=extra_headers, extra_query=extra_query, extra_body=extra_body, timeout=timeout
|
|
500
500
|
),
|
|
501
|
-
cast_to=
|
|
501
|
+
cast_to=MonitorEventResponse,
|
|
502
502
|
)
|
|
503
503
|
|
|
504
504
|
|
|
@@ -2,17 +2,15 @@
|
|
|
2
2
|
|
|
3
3
|
from __future__ import annotations
|
|
4
4
|
|
|
5
|
-
from .evaluation import Evaluation as Evaluation
|
|
6
|
-
from .api_response import APIResponse as APIResponse
|
|
7
5
|
from .defend_response import DefendResponse as DefendResponse
|
|
6
|
+
from .monitor_response import MonitorResponse as MonitorResponse
|
|
8
7
|
from .monitor_create_params import MonitorCreateParams as MonitorCreateParams
|
|
9
8
|
from .monitor_update_params import MonitorUpdateParams as MonitorUpdateParams
|
|
10
|
-
from .
|
|
9
|
+
from .monitor_event_response import MonitorEventResponse as MonitorEventResponse
|
|
10
|
+
from .monitor_detail_response import MonitorDetailResponse as MonitorDetailResponse
|
|
11
11
|
from .monitor_retrieve_params import MonitorRetrieveParams as MonitorRetrieveParams
|
|
12
12
|
from .workflow_event_response import WorkflowEventResponse as WorkflowEventResponse
|
|
13
|
-
from .monitor_retrieve_response import MonitorRetrieveResponse as MonitorRetrieveResponse
|
|
14
13
|
from .defend_submit_event_params import DefendSubmitEventParams as DefendSubmitEventParams
|
|
15
14
|
from .monitor_submit_event_params import MonitorSubmitEventParams as MonitorSubmitEventParams
|
|
16
15
|
from .defend_create_workflow_params import DefendCreateWorkflowParams as DefendCreateWorkflowParams
|
|
17
16
|
from .defend_update_workflow_params import DefendUpdateWorkflowParams as DefendUpdateWorkflowParams
|
|
18
|
-
from .monitor_submit_event_response import MonitorSubmitEventResponse as MonitorSubmitEventResponse
|
|
@@ -48,7 +48,7 @@ class DefendCreateWorkflowParams(TypedDict, total=False):
|
|
|
48
48
|
description: str
|
|
49
49
|
"""Description for the workflow."""
|
|
50
50
|
|
|
51
|
-
|
|
51
|
+
max_improvement_attempts: int
|
|
52
52
|
"""Max.
|
|
53
53
|
|
|
54
54
|
number of improvement action retries until a given event passes the guardrails.
|
|
@@ -31,7 +31,7 @@ class DefendResponse(BaseModel):
|
|
|
31
31
|
Nothing does not attempt any improvement.
|
|
32
32
|
"""
|
|
33
33
|
|
|
34
|
-
|
|
34
|
+
max_improvement_attempts: Optional[int] = None
|
|
35
35
|
"""Max.
|
|
36
36
|
|
|
37
37
|
number of improvement action retries until a given event passes the guardrails.
|
|
@@ -11,8 +11,9 @@ class DefendSubmitEventParams(TypedDict, total=False):
|
|
|
11
11
|
model_input: Required[ModelInput]
|
|
12
12
|
"""A dictionary of inputs sent to the LLM to generate output.
|
|
13
13
|
|
|
14
|
-
The dictionary must contain at least `user_prompt` or `system_prompt`
|
|
15
|
-
the ground_truth_adherence guardrail metric, `ground_truth` should be
|
|
14
|
+
The dictionary must contain at least a `user_prompt` field or a `system_prompt`
|
|
15
|
+
field. For the ground_truth_adherence guardrail metric, `ground_truth` should be
|
|
16
|
+
provided.
|
|
16
17
|
"""
|
|
17
18
|
|
|
18
19
|
model_output: Required[str]
|
|
@@ -36,7 +37,7 @@ class DefendSubmitEventParams(TypedDict, total=False):
|
|
|
36
37
|
|
|
37
38
|
class ModelInput(TypedDict, total=False):
|
|
38
39
|
ground_truth: str
|
|
39
|
-
"""The ground truth for evaluating Ground Truth Adherence guardrail."""
|
|
40
|
+
"""The ground truth for evaluating the Ground Truth Adherence guardrail."""
|
|
40
41
|
|
|
41
42
|
system_prompt: str
|
|
42
43
|
"""The system prompt used to generate the output."""
|
|
@@ -8,10 +8,10 @@ from pydantic import Field as FieldInfo
|
|
|
8
8
|
|
|
9
9
|
from .._models import BaseModel
|
|
10
10
|
|
|
11
|
-
__all__ = ["Evaluation", "
|
|
11
|
+
__all__ = ["MonitorDetailResponse", "Evaluation", "EvaluationModelInput", "Stats"]
|
|
12
12
|
|
|
13
13
|
|
|
14
|
-
class
|
|
14
|
+
class EvaluationModelInput(BaseModel):
|
|
15
15
|
ground_truth: Optional[str] = None
|
|
16
16
|
"""The ground truth for evaluating Ground Truth Adherence guardrail."""
|
|
17
17
|
|
|
@@ -29,11 +29,12 @@ class Evaluation(BaseModel):
|
|
|
29
29
|
evaluation_status: Literal["in_progress", "completed", "canceled", "queued", "failed"]
|
|
30
30
|
"""Status of the evaluation."""
|
|
31
31
|
|
|
32
|
-
api_model_input:
|
|
32
|
+
api_model_input: EvaluationModelInput = FieldInfo(alias="model_input")
|
|
33
33
|
"""A dictionary of inputs sent to the LLM to generate output.
|
|
34
34
|
|
|
35
|
-
The dictionary must contain at least `user_prompt` or `system_prompt`
|
|
36
|
-
ground_truth_adherence guardrail metric, `ground_truth` should be
|
|
35
|
+
The dictionary must contain at least a `user_prompt` field or a `system_prompt`
|
|
36
|
+
field. For ground_truth_adherence guardrail metric, `ground_truth` should be
|
|
37
|
+
provided.
|
|
37
38
|
"""
|
|
38
39
|
|
|
39
40
|
api_model_output: str = FieldInfo(alias="model_output")
|
|
@@ -102,3 +103,60 @@ class Evaluation(BaseModel):
|
|
|
102
103
|
|
|
103
104
|
start_timestamp: Optional[datetime] = None
|
|
104
105
|
"""The time the evaluation started in UTC."""
|
|
106
|
+
|
|
107
|
+
|
|
108
|
+
class Stats(BaseModel):
|
|
109
|
+
completed_evaluations: Optional[int] = None
|
|
110
|
+
"""Number of evaluations that completed successfully."""
|
|
111
|
+
|
|
112
|
+
failed_evaluations: Optional[int] = None
|
|
113
|
+
"""Number of evaluations that failed."""
|
|
114
|
+
|
|
115
|
+
in_progress_evaluations: Optional[int] = None
|
|
116
|
+
"""Number of evaluations currently in progress."""
|
|
117
|
+
|
|
118
|
+
queued_evaluations: Optional[int] = None
|
|
119
|
+
"""Number of evaluations currently queued."""
|
|
120
|
+
|
|
121
|
+
total_evaluations: Optional[int] = None
|
|
122
|
+
"""Total number of evaluations performed by this monitor."""
|
|
123
|
+
|
|
124
|
+
|
|
125
|
+
class MonitorDetailResponse(BaseModel):
|
|
126
|
+
monitor_id: str
|
|
127
|
+
"""A unique monitor ID."""
|
|
128
|
+
|
|
129
|
+
monitor_status: Literal["active", "inactive"]
|
|
130
|
+
"""Status of the monitor.
|
|
131
|
+
|
|
132
|
+
Can be `active` or `inactive`. Inactive monitors no longer record and evaluate
|
|
133
|
+
events.
|
|
134
|
+
"""
|
|
135
|
+
|
|
136
|
+
name: str
|
|
137
|
+
"""Name of this monitor."""
|
|
138
|
+
|
|
139
|
+
created_at: Optional[datetime] = None
|
|
140
|
+
"""The time the monitor was created in UTC."""
|
|
141
|
+
|
|
142
|
+
description: Optional[str] = None
|
|
143
|
+
"""Description of this monitor."""
|
|
144
|
+
|
|
145
|
+
evaluations: Optional[List[Evaluation]] = None
|
|
146
|
+
"""An array of all evaluations performed by this monitor.
|
|
147
|
+
|
|
148
|
+
Each one corresponds to a separate monitor event.
|
|
149
|
+
"""
|
|
150
|
+
|
|
151
|
+
stats: Optional[Stats] = None
|
|
152
|
+
"""
|
|
153
|
+
Contains five fields used for stats of this monitor: total evaluations,
|
|
154
|
+
completed evaluations, failed evaluations, queued evaluations, and in progress
|
|
155
|
+
evaluations.
|
|
156
|
+
"""
|
|
157
|
+
|
|
158
|
+
updated_at: Optional[datetime] = None
|
|
159
|
+
"""The most recent time the monitor was modified in UTC."""
|
|
160
|
+
|
|
161
|
+
user_id: Optional[str] = None
|
|
162
|
+
"""User ID of the user who created the monitor."""
|
|
@@ -5,10 +5,10 @@ from datetime import datetime
|
|
|
5
5
|
|
|
6
6
|
from .._models import BaseModel
|
|
7
7
|
|
|
8
|
-
__all__ = ["
|
|
8
|
+
__all__ = ["MonitorEventResponse"]
|
|
9
9
|
|
|
10
10
|
|
|
11
|
-
class
|
|
11
|
+
class MonitorEventResponse(BaseModel):
|
|
12
12
|
evaluation_id: str
|
|
13
13
|
"""A unique evaluation ID associated with this event."""
|
|
14
14
|
|
|
@@ -20,16 +20,3 @@ class Data(BaseModel):
|
|
|
20
20
|
|
|
21
21
|
created_at: Optional[datetime] = None
|
|
22
22
|
"""The time the monitor event was created in UTC."""
|
|
23
|
-
|
|
24
|
-
|
|
25
|
-
class MonitorSubmitEventResponse(BaseModel):
|
|
26
|
-
success: bool
|
|
27
|
-
"""Represents whether the request was completed successfully."""
|
|
28
|
-
|
|
29
|
-
data: Optional[Data] = None
|
|
30
|
-
|
|
31
|
-
message: Optional[str] = None
|
|
32
|
-
"""The accompanying message for the request.
|
|
33
|
-
|
|
34
|
-
Includes error details when applicable.
|
|
35
|
-
"""
|
|
@@ -6,10 +6,10 @@ from typing_extensions import Literal
|
|
|
6
6
|
|
|
7
7
|
from .._models import BaseModel
|
|
8
8
|
|
|
9
|
-
__all__ = ["
|
|
9
|
+
__all__ = ["MonitorResponse"]
|
|
10
10
|
|
|
11
11
|
|
|
12
|
-
class
|
|
12
|
+
class MonitorResponse(BaseModel):
|
|
13
13
|
monitor_id: str
|
|
14
14
|
"""A unique monitor ID."""
|
|
15
15
|
|
|
@@ -34,16 +34,3 @@ class Data(BaseModel):
|
|
|
34
34
|
|
|
35
35
|
user_id: Optional[str] = None
|
|
36
36
|
"""User ID of the user who created the monitor."""
|
|
37
|
-
|
|
38
|
-
|
|
39
|
-
class APIResponse(BaseModel):
|
|
40
|
-
success: bool
|
|
41
|
-
"""Represents whether the request was completed successfully."""
|
|
42
|
-
|
|
43
|
-
data: Optional[Data] = None
|
|
44
|
-
|
|
45
|
-
message: Optional[str] = None
|
|
46
|
-
"""The accompanying message for the request.
|
|
47
|
-
|
|
48
|
-
Includes error details when applicable.
|
|
49
|
-
"""
|
|
@@ -31,8 +31,9 @@ class MonitorSubmitEventParams(TypedDict, total=False):
|
|
|
31
31
|
model_input: Required[ModelInput]
|
|
32
32
|
"""A dictionary of inputs sent to the LLM to generate output.
|
|
33
33
|
|
|
34
|
-
The dictionary must contain at least a `user_prompt` or `system_prompt`
|
|
35
|
-
For ground_truth_adherence guardrail metric, `ground_truth` should be
|
|
34
|
+
The dictionary must contain at least a `user_prompt` field or a `system_prompt`
|
|
35
|
+
field. For ground_truth_adherence guardrail metric, `ground_truth` should be
|
|
36
|
+
provided.
|
|
36
37
|
"""
|
|
37
38
|
|
|
38
39
|
model_output: Required[str]
|