Skip to content

fix: propagate Reactor context to chunk event hooks in ReActAgent #1762

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
Vinke2 wants to merge 2 commits into
base: main
Choose a base branch
from
+87 −62
Open

fix: propagate Reactor context to chunk event hooks in ReActAgent #1762

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
7c03ee2
fix: propagate Reactor context to chunk event hooks in ReActAgent
Vinke2 Jun 15, 2026
47b3034
reformat
Vinke2 Jun 15, 2026
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
149 changes: 87 additions & 62 deletions agentscope-core/src/main/java/io/agentscope/core/ReActAgent.java
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2050,29 +2050,36 @@ private Flux<AgentEvent> modelCallStream(
.concatMap(
chunk -> {
List<Msg> chunkMsgs = context.processChunk(chunk);
for (Msg msg : chunkMsgs) {
hookDispatcher
.fireReasoningChunk(
msg,
context,
mci.model().getModelName())
.subscribe();
}
return Flux.deferContextual(

@AgentScopeJavaBot AgentScopeJavaBot Jun 15, 2026

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

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

[nit] Consider adding a brief inline comment explaining why Flux.deferContextual is needed here (e.g. // Propagate parent Reactor context so chunk hooks can access SubagentEventBus), similar to the existing comment at line 2402–2406 for executeToolCalls. This helps future readers understand the non-obvious context-propagation requirement without having to trace the full call chain.

@AgentScopeJavaBot AgentScopeJavaBot Jun 15, 2026

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

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

[nit] Consider adding a brief inline comment explaining why Flux.deferContextual is needed here (e.g. // Propagate parent Reactor context so chunk hooks can access SubagentEventBus), similar to the existing comment at line 2402–2406 for executeToolCalls. This helps future readers understand the non-obvious context-propagation requirement without having to trace the full call chain.

parentCtx -> {
for (Msg msg : chunkMsgs) {
hookDispatcher
.fireReasoningChunk(
msg,
context,
mci.model().getModelName())
.contextWrite(
ctx ->
ctx.putAll(
parentCtx))
.subscribe();

@AgentScopeJavaBot AgentScopeJavaBot Jun 15, 2026

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

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

[minor] The .subscribe() call remains bare — no error consumer is provided. If a hook's onEvent throws (or the Mono<Void> signals an error), Reactor will invoke onErrorDropped, silently swallowing the exception. This is a pre-existing issue, not introduced by this PR, but since you're already touching this line, consider adding a minimal error handler (e.g. .subscribe(v -> {}, e -> log.warn("Reasoning chunk hook failed", e))) to surface failures. Same applies to lines 2463 and 2975.

@AgentScopeJavaBot AgentScopeJavaBot Jun 15, 2026

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

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

[minor] The .subscribe() call remains bare — no error consumer is provided. If a hook's onEvent throws (or the Mono<Void> signals an error), Reactor will invoke onErrorDropped, silently swallowing the exception. This is a pre-existing issue, not introduced by this PR, but since you're already touching this line, consider adding a minimal error handler (e.g. .subscribe(v -> {}, e -> log.warn("Reasoning chunk hook failed", e))) to surface failures. Same applies to lines 2463 and 2975.

}

List<AgentEvent> events = new ArrayList<>();
for (ContentBlock block : chunk.getContent()) {
emitBlockEvents(
block,
replyId,
context,
textStarted,
thinkingStarted,
withToolEvents
? startedToolCalls
: new ConcurrentHashMap<>(),
events);
}
return Flux.fromIterable(events);
List<AgentEvent> events = new ArrayList<>();
for (ContentBlock block : chunk.getContent()) {
emitBlockEvents(
block,
replyId,
context,
textStarted,
thinkingStarted,
withToolEvents
? startedToolCalls
: new ConcurrentHashMap<>(),
events);
}
return Flux.fromIterable(events);
});
});

Flux<AgentEvent> endEvents =
Expand Down Expand Up @@ -2449,6 +2456,10 @@ private Flux<AgentEvent> runToolBatch(
hookDispatcher
.fireActingChunk(
toolUse, chunk, toolkit)
.contextWrite(
ctx ->
ctx.putAll(
parentCtx))
.subscribe();
});

Expand Down Expand Up @@ -2947,47 +2958,61 @@ private Flux<AgentEvent> summaryModelCallStream(
.concatMap(
chunk -> {
List<Msg> chunkMsgs = context.processChunk(chunk);
for (Msg msg : chunkMsgs) {
hookDispatcher
.fireSummaryChunk(
msg,
context,
hookOptions,
model.getModelName())
.subscribe();
}
return Flux.deferContextual(
parentCtx -> {
for (Msg msg : chunkMsgs) {
hookDispatcher
.fireSummaryChunk(
msg,
context,
hookOptions,
model.getModelName())
.contextWrite(
ctx ->
ctx.putAll(
parentCtx))
.subscribe();
}

List<AgentEvent> events = new ArrayList<>();
for (ContentBlock block : chunk.getContent()) {
if (block instanceof TextBlock tb) {
if (textStarted.compareAndSet(false, true)) {
events.add(
new TextBlockStartEvent(
replyId, "text"));
}
if (tb.getText() != null
&& !tb.getText().isEmpty()) {
events.add(
new TextBlockDeltaEvent(
replyId, "text", tb.getText()));
}
} else if (block instanceof ThinkingBlock tb) {
if (thinkingStarted.compareAndSet(false, true)) {
events.add(
new ThinkingBlockStartEvent(
replyId, "thinking"));
}
if (tb.getThinking() != null
&& !tb.getThinking().isEmpty()) {
events.add(
new ThinkingBlockDeltaEvent(
replyId,
"thinking",
tb.getThinking()));
}
}
}
return Flux.fromIterable(events);
List<AgentEvent> events = new ArrayList<>();
for (ContentBlock block : chunk.getContent()) {
if (block instanceof TextBlock tb) {
if (textStarted.compareAndSet(
false, true)) {
events.add(
new TextBlockStartEvent(
replyId, "text"));
}
if (tb.getText() != null
&& !tb.getText().isEmpty()) {
events.add(
new TextBlockDeltaEvent(
replyId,
"text",
tb.getText()));
}
} else if (block
instanceof ThinkingBlock tb) {
if (thinkingStarted.compareAndSet(
false, true)) {
events.add(
new ThinkingBlockStartEvent(
replyId,
"thinking"));
}
if (tb.getThinking() != null
&& !tb.getThinking()
.isEmpty()) {
events.add(
new ThinkingBlockDeltaEvent(
replyId,
"thinking",
tb.getThinking()));
}
}
}
return Flux.fromIterable(events);
});
});

Flux<AgentEvent> endEvents =
Expand Down
Loading