Merge pull request #1235 from samdoran/infer-max

Set maximum length for fields

Author: tisnik

src/models/rlsapi/requests.py

@@ -31,6 +31,7 @@ class RlsapiV1Attachment(ConfigurationBase):
 
     contents: str = Field(
         default="",
+        max_length=65_536,
         description="File contents read on client",
         examples=["# Configuration file\nkey=value"],
     )
@@ -50,6 +51,7 @@ class RlsapiV1Terminal(ConfigurationBase):
 
     output: str = Field(
         default="",
+        max_length=65_536,
         description="Terminal output from client",
         examples=["bash: command not found", "Permission denied"],
     )
@@ -129,6 +131,7 @@ class RlsapiV1Context(ConfigurationBase):
 
     stdin: str = Field(
         default="",
+        max_length=65_536,
         description="Redirect input from stdin",
         examples=["piped input from previous command"],
     )
@@ -173,6 +176,7 @@ class RlsapiV1InferRequest(ConfigurationBase):
     question: str = Field(
         ...,
         min_length=1,
+        max_length=10_240,
         description="User question",
         examples=["How do I list files?", "How do I configure SELinux?"],
     )

tests/integration/endpoints/test_rlsapi_v1_integration.py

@@ -13,6 +13,7 @@
 
 import pytest
 from fastapi import HTTPException, status
+from fastapi.testclient import TestClient
 from llama_stack_client import APIConnectionError
 from pytest_mock import MockerFixture
 
@@ -494,3 +495,22 @@ async def test_rlsapi_v1_infer_skip_rag(
         auth=test_auth,
     )
     assert isinstance(response, RlsapiV1InferResponse)
+
+
+@pytest.mark.parametrize(
+    "json",
+    (
+        ({"question": "?" * 10_241}),
+        ({"question": "Q", "context": {"stdin": "a" * 65_537}}),
+        ({"question": "Q", "context": {"attachments": {"contents": "A" * 65_537}}}),
+        ({"question": "Q", "context": {"terminal": {"output": "T" * 65_537}}}),
+    ),
+    ids=["question", "stdin", "attachment_contents", "terminal_output"],
+)
+def test_infer_size_limit(integration_http_client: TestClient, json) -> None:
+    """Test that a field exceeding limit is rejected."""
+    response = integration_http_client.post("/v1/infer", json=json)
+    detail = response.json()["detail"]
+
+    assert response.status_code == status.HTTP_422_UNPROCESSABLE_CONTENT
+    assert "string_too_long" in {item["type"] for item in detail}

tests/unit/app/endpoints/test_rlsapi_v1.py

@@ -309,7 +309,6 @@ def test_config_error_503_matches_llm_error_503_shape(
 # --- Test retrieve_simple_response ---
 
 
-@pytest.mark.asyncio
 async def test_retrieve_simple_response_success(
     mock_configuration: AppConfig, mock_llm_response: None
 ) -> None:
@@ -320,7 +319,6 @@ async def test_retrieve_simple_response_success(
     assert response == "This is a test LLM response."
 
 
-@pytest.mark.asyncio
 async def test_retrieve_simple_response_empty_output(
     mock_configuration: AppConfig, mock_empty_llm_response: None
 ) -> None:
@@ -331,7 +329,6 @@ async def test_retrieve_simple_response_empty_output(
     assert response == ""
 
 
-@pytest.mark.asyncio
 async def test_retrieve_simple_response_api_connection_error(
     mock_configuration: AppConfig, mock_api_connection_error: None
 ) -> None:
@@ -384,7 +381,6 @@ def test_get_rh_identity_context_with_empty_values(mocker: MockerFixture) -> Non
 # --- Test infer_endpoint ---
 
 
-@pytest.mark.asyncio
 async def test_infer_minimal_request(
     mocker: MockerFixture,
     mock_configuration: AppConfig,
@@ -409,7 +405,6 @@ async def test_infer_minimal_request(
     assert check_suid(response.data.request_id)
 
 
-@pytest.mark.asyncio
 async def test_infer_full_context_request(
     mocker: MockerFixture,
     mock_configuration: AppConfig,
@@ -441,7 +436,6 @@ async def test_infer_full_context_request(
     assert response.data.request_id
 
 
-@pytest.mark.asyncio
 async def test_infer_generates_unique_request_ids(
     mocker: MockerFixture,
     mock_configuration: AppConfig,
@@ -469,7 +463,6 @@ async def test_infer_generates_unique_request_ids(
     assert response1.data.request_id != response2.data.request_id
 
 
-@pytest.mark.asyncio
 async def test_infer_api_connection_error_returns_503(
     mocker: MockerFixture,
     mock_configuration: AppConfig,
@@ -492,7 +485,6 @@ async def test_infer_api_connection_error_returns_503(
     assert exc_info.value.status_code == status.HTTP_503_SERVICE_UNAVAILABLE
 
 
-@pytest.mark.asyncio
 async def test_infer_empty_llm_response_returns_fallback(
     mocker: MockerFixture,
     mock_configuration: AppConfig,
@@ -517,7 +509,6 @@ async def test_infer_empty_llm_response_returns_fallback(
 # --- Test Splunk integration ---
 
 
-@pytest.mark.asyncio
 async def test_infer_queues_splunk_event_on_success(
     mocker: MockerFixture,
     mock_configuration: AppConfig,
@@ -542,7 +533,6 @@ async def test_infer_queues_splunk_event_on_success(
     assert call_args[0][2] == "infer_with_llm"
 
 
-@pytest.mark.asyncio
 async def test_infer_queues_splunk_error_event_on_failure(
     mocker: MockerFixture,
     mock_configuration: AppConfig,
@@ -567,7 +557,6 @@ async def test_infer_queues_splunk_error_event_on_failure(
     assert call_args[0][2] == "infer_error"
 
 
-@pytest.mark.asyncio
 async def test_infer_splunk_event_includes_rh_identity_context(
     mocker: MockerFixture,
     mock_configuration: AppConfig,
@@ -638,7 +627,6 @@ def _setup_responses_mock_with_capture(
     return mock_create
 
 
-@pytest.mark.asyncio
 async def test_retrieve_simple_response_passes_tools(
     mocker: MockerFixture, mock_configuration: AppConfig
 ) -> None:
@@ -660,7 +648,6 @@ async def test_retrieve_simple_response_passes_tools(
     assert call_kwargs["tools"] == tools
 
 
-@pytest.mark.asyncio
 async def test_retrieve_simple_response_defaults_to_empty_tools(
     mocker: MockerFixture, mock_configuration: AppConfig
 ) -> None:
@@ -674,7 +661,6 @@ async def test_retrieve_simple_response_defaults_to_empty_tools(
     assert call_kwargs["tools"] == []
 
 
-@pytest.mark.asyncio
 async def test_infer_endpoint_calls_get_mcp_tools(
     mocker: MockerFixture,
     mock_configuration: AppConfig,
@@ -704,7 +690,6 @@ async def test_infer_endpoint_calls_get_mcp_tools(
     )
 
 
-@pytest.mark.asyncio
 async def test_infer_generic_runtime_error_reraises(
     mocker: MockerFixture,
     mock_configuration: AppConfig,
@@ -725,7 +710,6 @@ async def test_infer_generic_runtime_error_reraises(
         )
 
 
-@pytest.mark.asyncio
 async def test_infer_generic_runtime_error_records_failure(
     mocker: MockerFixture,
     mock_configuration: AppConfig,

tests/unit/models/rlsapi/test_requests.py

@@ -594,3 +594,33 @@ def test_priority_order(self, make_request: Any) -> None:
         )
         result = request.get_input_source()
         assert result == "Q\n\nS\n\nA\n\nT"
+
+
+@pytest.mark.parametrize(
+    ("model", "field", "max_length"),
+    [
+        (RlsapiV1Attachment, "contents", 65_536),
+        (RlsapiV1Terminal, "output", 65_536),
+        (RlsapiV1Context, "stdin", 65_536),
+        (RlsapiV1InferRequest, "question", 10_240),
+    ],
+    ids=[
+        "attachment-contents",
+        "terminal-output",
+        "context-stdin",
+        "infer-request-question",
+    ],
+)
+def test_value_max_length(model, field, max_length) -> None:
+    """Test that fields with longer than allowed data are not allowed"""
+    value = "a" * max_length
+    bad_value = value + "a"
+
+    instance = model(**{field: value})
+    with pytest.raises(
+        ValidationError,
+        match=f"should have at most {max_length} characters",
+    ):
+        model(**{field: bad_value})
+
+    assert getattr(instance, field) == value