Add basic support for passing error messages from API to UI while str…

…eaming

ragna/deploy/_api/core.py

@@ -296,14 +296,22 @@ async def answer(
             )
             core_chat = schema_to_core_chat(session, user=user, chat=chat)
 
-        core_answer = await core_chat.answer(prompt, stream=stream)
+        # smoke test to catch errors unrelated to streaming
+        # longer-term solution tracked by #409
+        try:
+            core_answer = await core_chat.answer(prompt, stream=stream)
+        except Exception as e:
+            raise HTTPException(
+                status_code=500,
+                detail=str(e),
+            ) from e
 
         if stream:
-
-            async def message_chunks() -> AsyncIterator[BaseModel]:
+            # smoke test to catch errors unrelated to streaming
+            # longer-term solution tracked by #409
+            try:
                 core_answer_stream = aiter(core_answer)
                 content_chunk = await anext(core_answer_stream)
-
                 answer = schemas.Message(
                     content=content_chunk,
                     role=core_answer.role,
@@ -312,6 +320,13 @@ async def message_chunks() -> AsyncIterator[BaseModel]:
                         for source in core_answer.sources
                     ],
                 )
+            except Exception as e:
+                raise HTTPException(
+                    status_code=500,
+                    detail=str(e),
+                ) from e
+
+            async def message_chunks() -> AsyncIterator[BaseModel]:
                 yield answer
 
                 # Avoid sending the sources multiple times

tests/deploy/api/test_streaming.py

@@ -0,0 +1,74 @@
+import io
+import json
+
+import pytest
+from fastapi.testclient import TestClient
+
+from ragna.core import Assistant, SourceStorage
+from ragna.deploy import Config
+from ragna.deploy._api import app
+
+from .utils import authenticate
+
+
+class FakeAssistant(Assistant):
+    def answer(self, prompt, sources):
+        if prompt == "Assistant Error":
+            raise Exception("Assistant Error")
+
+
+class FakeSourceStorage(SourceStorage):
+    async def store(self, documents):
+        pass
+
+    async def retrieve(self, documents, prompt):
+        if prompt == "SourceStorage Error":
+            raise Exception("SourceStorage Error")
+
+
+@pytest.mark.parametrize("prompt", ["Assistant Error", "SourceStorage Error"])
+def test_internal_server_error_response(tmp_local_root, prompt):
+    config = Config(
+        local_root=tmp_local_root,
+        assistants=[FakeAssistant],
+        source_storages=[FakeSourceStorage],
+    )
+
+    with TestClient(app(config=config, ignore_unavailable_components=False)) as client:
+        authenticate(client)
+
+        document_upload = (
+            client.post("/document", json={"name": "fake.txt"})
+            .raise_for_status()
+            .json()
+        )
+        document = document_upload["document"]
+        parameters = document_upload["parameters"]
+        client.request(
+            parameters["method"],
+            parameters["url"],
+            data=parameters["data"],
+            files={"file": io.BytesIO(b"!")},
+        )
+
+        chat_metadata = {
+            "name": "test-chat",
+            "source_storage": "FakeSourceStorage",
+            "assistant": "FakeAssistant",
+            "params": {},
+            "documents": [document],
+        }
+        chat = client.post("/chats", json=chat_metadata).raise_for_status().json()
+
+        _ = client.post(f"/chats/{chat['id']}/prepare").raise_for_status().json()
+
+        with client.stream(
+            "POST",
+            f"/chats/{chat['id']}/answer",
+            json={"prompt": prompt, "stream": True},
+        ) as response:
+            r = response.read()
+
+        assert response.status_code == 500
+
+        assert json.loads(r.decode("utf-8"))["detail"] == prompt