phoenix-oss/llama-stack-mirror
0
0
Fork 1
mirror of https://github.com/meta-llama/llama-stack.git synced 2025-12-03 09:53:45 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

fix(nvidia-safety): correct NeMo Guardrails API endpoint

Browse source
This commit is contained in:
r-bit-rry 2025-11-20 13:00:02 +02:00
parent 91f1b352b4
commit 1458e881e5
2 changed files with 9 additions and 45 deletions
Download patch file Download diff file Expand all files Collapse all files

15
src/llama_stack/providers/remote/safety/nvidia/nvidia.py
View file

@ -125,7 +125,7 @@ class NeMoGuardrails:
async def run(self, messages: list[OpenAIMessageParam]) -> RunShieldResponse: async def run(self, messages: list[OpenAIMessageParam]) -> RunShieldResponse:
""" """
Queries the /v1/guardrails/checks endpoint of the NeMo guardrails deployed API. Queries the /v1/chat/completions endpoint of the NeMo guardrails deployed API.
Args: Args:
messages (List[Message]): A list of Message objects to be checked for safety violations. messages (List[Message]): A list of Message objects to be checked for safety violations.
@ -138,19 +138,10 @@ class NeMoGuardrails:
requests.HTTPError: If the POST request fails. requests.HTTPError: If the POST request fails.
""" """
request_data = { request_data = {
"model": self.model,
"messages": [{"role": message.role, "content": message.content} for message in messages],
"temperature": self.temperature,
"top_p": 1,
"frequency_penalty": 0,
"presence_penalty": 0,
"max_tokens": 160,
"stream": False,
"guardrails": {
"config_id": self.config_id, "config_id": self.config_id,
}, "messages": [{"role": message.role, "content": message.content} for message in messages],
} }
response = await self._guardrails_post(path="/v1/guardrail/checks", data=request_data) response = await self._guardrails_post(path="/v1/chat/completions", data=request_data)
if response["status"] == "blocked": if response["status"] == "blocked":
user_message = "Sorry I cannot do this." user_message = "Sorry I cannot do this."

39
tests/unit/providers/nvidia/test_safety.py
View file

@ -152,22 +152,13 @@ async def test_run_shield_allowed(nvidia_adapter, mock_guardrails_post):
# Verify the Guardrails API was called correctly # Verify the Guardrails API was called correctly
mock_guardrails_post.assert_called_once_with( mock_guardrails_post.assert_called_once_with(
path="/v1/guardrail/checks", path="/v1/chat/completions",
data={ data={
"model": shield_id, "config_id": "self-check",
"messages": [ "messages": [
{"role": "user", "content": "Hello, how are you?"}, {"role": "user", "content": "Hello, how are you?"},
{"role": "assistant", "content": "I'm doing well, thank you for asking!"}, {"role": "assistant", "content": "I'm doing well, thank you for asking!"},
], ],
"temperature": 1.0,
"top_p": 1,
"frequency_penalty": 0,
"presence_penalty": 0,
"max_tokens": 160,
"stream": False,
"guardrails": {
"config_id": "self-check",
},
}, },
) )
@ -206,22 +197,13 @@ async def test_run_shield_blocked(nvidia_adapter, mock_guardrails_post):
# Verify the Guardrails API was called correctly # Verify the Guardrails API was called correctly
mock_guardrails_post.assert_called_once_with( mock_guardrails_post.assert_called_once_with(
path="/v1/guardrail/checks", path="/v1/chat/completions",
data={ data={
"model": shield_id, "config_id": "self-check",
"messages": [ "messages": [
{"role": "user", "content": "Hello, how are you?"}, {"role": "user", "content": "Hello, how are you?"},
{"role": "assistant", "content": "I'm doing well, thank you for asking!"}, {"role": "assistant", "content": "I'm doing well, thank you for asking!"},
], ],
"temperature": 1.0,
"top_p": 1,
"frequency_penalty": 0,
"presence_penalty": 0,
"max_tokens": 160,
"stream": False,
"guardrails": {
"config_id": "self-check",
},
}, },
) )
@ -286,22 +268,13 @@ async def test_run_shield_http_error(nvidia_adapter, mock_guardrails_post):
# Verify the Guardrails API was called correctly # Verify the Guardrails API was called correctly
mock_guardrails_post.assert_called_once_with( mock_guardrails_post.assert_called_once_with(
path="/v1/guardrail/checks", path="/v1/chat/completions",
data={ data={
"model": shield_id, "config_id": "self-check",
"messages": [ "messages": [
{"role": "user", "content": "Hello, how are you?"}, {"role": "user", "content": "Hello, how are you?"},
{"role": "assistant", "content": "I'm doing well, thank you for asking!"}, {"role": "assistant", "content": "I'm doing well, thank you for asking!"},
], ],
"temperature": 1.0,
"top_p": 1,
"frequency_penalty": 0,
"presence_penalty": 0,
"max_tokens": 160,
"stream": False,
"guardrails": {
"config_id": "self-check",
},
}, },
) )
# Verify the exception message # Verify the exception message