Skip to content

combine json chunks from requests #4317

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
wants to merge 3 commits into
base: main
Choose a base branch
from
+103 −15
Open

combine json chunks from requests #4317

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
76d0400
combine json chunks
jiapingzeng Oct 14, 2025
8f6c270
add unit tests
jiapingzeng Oct 21, 2025
2bc0ae1
enhance input error handling
jiapingzeng Oct 22, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
49 changes: 34 additions & 15 deletions plugin/src/main/java/org/opensearch/ml/rest/RestMLExecuteStreamAction.java
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -17,8 +17,8 @@
import static org.opensearch.ml.utils.RestActionUtils.isAsync;
import static org.opensearch.ml.utils.TenantAwareHelper.getTenantID;

import java.io.ByteArrayOutputStream;
import java.io.IOException;
import java.io.UncheckedIOException;
import java.nio.ByteBuffer;
import java.util.LinkedHashMap;
import java.util.List;
Expand Down Expand Up @@ -48,7 +48,6 @@
import org.opensearch.ml.common.MLModel;
import org.opensearch.ml.common.agent.MLAgent;
import org.opensearch.ml.common.dataset.remote.RemoteInferenceInputDataSet;
import org.opensearch.ml.common.exception.MLException;
import org.opensearch.ml.common.input.Input;
import org.opensearch.ml.common.input.MLInput;
import org.opensearch.ml.common.input.execute.agent.AgentMLInput;
Expand Down Expand Up @@ -158,10 +157,12 @@ public RestChannelConsumer prepareRequest(RestRequest request, NodeClient client
);
channel.prepareResponse(RestStatus.OK, headers);

Flux.from(channel).ofType(HttpChunk.class).concatMap(chunk -> {
final CompletableFuture<HttpChunk> future = new CompletableFuture<>();
Flux.from(channel).ofType(HttpChunk.class).collectList().flatMap(chunks -> {
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Is this possible to add some unit tests?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Not possible to create UT for prepareRequest as chunk collection happens async. However, I did try sanity test with chunked and non-chunked request, see my latest comment.

try {
MLExecuteTaskRequest mlExecuteTaskRequest = getRequest(agentId, request, chunk.content());
BytesReference completeContent = combineChunks(chunks);
MLExecuteTaskRequest mlExecuteTaskRequest = getRequest(agentId, request, completeContent);

final CompletableFuture<HttpChunk> future = new CompletableFuture<>();
StreamTransportResponseHandler<MLTaskResponse> handler = new StreamTransportResponseHandler<MLTaskResponse>() {
@Override
public void handleStreamResponse(StreamTransportResponse<MLTaskResponse> streamResponse) {
Expand Down Expand Up @@ -214,19 +215,23 @@ public MLTaskResponse read(StreamInput in) throws IOException {
handler
);

} catch (IOException e) {
throw new MLException("Got an exception in flux.", e);
return Mono.fromCompletionStage(future);
} catch (Exception e) {
log.error("Failed to parse or process request", e);
return Mono.error(e);
}

return Mono.fromCompletionStage(future);
}).doOnNext(channel::sendChunk).onErrorComplete(ex -> {
// Error handling
}).doOnNext(channel::sendChunk).onErrorResume(ex -> {
log.error("Error occurred", ex);
try {
channel.sendResponse(new BytesRestResponse(channel, (Exception) ex));
return true;
} catch (final IOException e) {
throw new UncheckedIOException(e);
String errorMessage = ex instanceof IOException
? "Failed to parse request: " + ex.getMessage()
: "Error processing request: " + ex.getMessage();
HttpChunk errorChunk = createHttpChunk("data: {\"error\": \"" + errorMessage.replace("\"", "\\\"") + "\"}\n\n", true);
channel.sendChunk(errorChunk);
} catch (Exception e) {
log.error("Failed to send error chunk", e);
}
return Mono.empty();
}).subscribe();
};

Expand Down Expand Up @@ -402,6 +407,20 @@ private String extractTensorResult(MLTaskResponse response, String tensorName) {
return Map.of();
}

@VisibleForTesting
BytesReference combineChunks(List<HttpChunk> chunks) {
try {
ByteArrayOutputStream buffer = new ByteArrayOutputStream();
for (HttpChunk chunk : chunks) {
chunk.content().writeTo(buffer);
}
return BytesReference.fromByteBuffer(ByteBuffer.wrap(buffer.toByteArray()));
} catch (IOException e) {
log.error("Failed to combine chunks", e);
throw new OpenSearchStatusException("Failed to combine request chunks", RestStatus.INTERNAL_SERVER_ERROR, e);
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

We spoke about not using this OpenSearchStatusException as it will throw 500 errors. Have you tested on your end?

Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

We can define HTTP error code in OpenSearchStatusException.
I see here we define it as 500 error. @mingshl Do you have any concern? Should we define as some other error code ?

Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I assume this is Server-side I/O failure while handling a request. If that's the case, then 500 should be the right status code.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I think 500 makes sense here as the chunks are failing to be combined which is I/O failure.

However when testing with invalid JSON once all the chunks are combined this 400 doesn't seem to be getting through. Will enhance error handling on this part instead.

return Mono.error(new OpenSearchStatusException("Failed to parse request", RestStatus.BAD_REQUEST, e));

}
}

private HttpChunk createHttpChunk(String sseData, boolean isLast) {
BytesReference bytesRef = BytesReference.fromByteBuffer(ByteBuffer.wrap(sseData.getBytes()));
return new HttpChunk() {
Expand Down
69 changes: 69 additions & 0 deletions plugin/src/test/java/org/opensearch/ml/rest/RestMLExecuteStreamActionTests.java
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -33,7 +33,9 @@
import org.opensearch.common.xcontent.XContentType;
import org.opensearch.core.action.ActionListener;
import org.opensearch.core.common.bytes.BytesArray;
import org.opensearch.core.common.bytes.BytesReference;
import org.opensearch.core.rest.RestStatus;
import org.opensearch.http.HttpChunk;
import org.opensearch.ml.common.FunctionName;
import org.opensearch.ml.common.MLModel;
import org.opensearch.ml.common.agent.LLMSpec;
Expand Down Expand Up @@ -302,4 +304,71 @@ public void testGetRequestAgentFrameworkDisabled() {
when(mlFeatureEnabledSetting.isAgentFrameworkEnabled()).thenReturn(false);
assertThrows(IllegalStateException.class, () -> restAction.handleRequest(request, channel, client));
}

@Test
public void testCombineChunksWithSingleChunk() {
String testContent = "{\"parameters\":{\"question\":\"test\"}}";
BytesArray bytesArray = new BytesArray(testContent);

HttpChunk mockChunk = mock(HttpChunk.class);
when(mockChunk.content()).thenReturn(bytesArray);

BytesReference result = restAction.combineChunks(List.of(mockChunk));

assertNotNull(result);
assertEquals(testContent, result.utf8ToString());
}

@Test
public void testCombineChunksWithMultipleChunks() {
String chunk1Content = "{\"parameters\":";
String chunk2Content = "{\"question\":";
String chunk3Content = "\"test\"}}";

BytesArray bytes1 = new BytesArray(chunk1Content);
BytesArray bytes2 = new BytesArray(chunk2Content);
BytesArray bytes3 = new BytesArray(chunk3Content);

HttpChunk mockChunk1 = mock(HttpChunk.class);
HttpChunk mockChunk2 = mock(HttpChunk.class);
HttpChunk mockChunk3 = mock(HttpChunk.class);

when(mockChunk1.content()).thenReturn(bytes1);
when(mockChunk2.content()).thenReturn(bytes2);
when(mockChunk3.content()).thenReturn(bytes3);

BytesReference result = restAction.combineChunks(List.of(mockChunk1, mockChunk2, mockChunk3));

assertNotNull(result);
String expectedContent = chunk1Content + chunk2Content + chunk3Content;
assertEquals(expectedContent, result.utf8ToString());
}

@Test
public void testCombineChunksWithEmptyList() {
BytesReference result = restAction.combineChunks(List.of());

assertNotNull(result);
assertEquals(0, result.length());
}

@Test
public void testCombineChunksWithLargeContent() {
StringBuilder largeContent = new StringBuilder();
for (int i = 0; i < 1000; i++) {
largeContent.append("chunk").append(i).append(",");
}
String content = largeContent.toString();

BytesArray bytesArray = new BytesArray(content);

HttpChunk mockChunk = mock(HttpChunk.class);
when(mockChunk.content()).thenReturn(bytesArray);

BytesReference result = restAction.combineChunks(List.of(mockChunk));

assertNotNull(result);
assertEquals(content.length(), result.length());
assertEquals(content, result.utf8ToString());
}
}