fix(websocket): narrow compact replay detection

This commit is contained in:
DragonFSKY
2026-03-23 23:27:20 +08:00
parent a0fe273081
commit d2d0e6f6a1
2 changed files with 40 additions and 19 deletions
@@ -703,29 +703,20 @@ func mergeJSONArrayRaw(existingRaw, appendRaw string) (string, error) {
return string(out), nil return string(out), nil
} }
// inputContainsFullTranscript returns true when the input array looks like a // inputContainsFullTranscript returns true when the input array carries compact
// complete conversation history rather than an incremental append. After a // replay markers that indicate the client already sent the full conversation
// client-side compact the input already carries the full (compacted) transcript // transcript. Merging that input with stale lastRequest/lastResponseOutput
// which may include assistant messages or compaction items. Merging that with // would duplicate or break function_call/function_call_output pairings, so the
// the stale lastRequest / lastResponseOutput would duplicate or break // caller should use the input as-is.
// function_call / function_call_output pairings, so the caller should use the
// input as-is.
// //
// Heuristic: the array is a full transcript when it contains either // Assistant messages alone are not enough to classify the payload as a replay:
// - a message with role="assistant", or // incremental websocket requests may legitimately append assistant items.
// - a compaction item (type="compaction" or "compaction_summary").
//
// Normal incremental turns only contain user messages or function_call_output
// items and never carry either of these signals.
func inputContainsFullTranscript(input gjson.Result) bool { func inputContainsFullTranscript(input gjson.Result) bool {
if !input.IsArray() { if !input.IsArray() {
return false return false
} }
for _, item := range input.Array() { for _, item := range input.Array() {
t := item.Get("type").String() t := item.Get("type").String()
if t == "message" && item.Get("role").String() == "assistant" {
return true
}
if t == "compaction" || t == "compaction_summary" { if t == "compaction" || t == "compaction_summary" {
return true return true
} }
@@ -1401,13 +1401,13 @@ func TestResponsesWebsocketCompactionResetsTurnStateOnTranscriptReplacement(t *t
} }
} }
func TestInputContainsFullTranscriptDetectsAssistantMessage(t *testing.T) { func TestInputContainsFullTranscriptFalseForAssistantMessageOnly(t *testing.T) {
input := gjson.Parse(`[ input := gjson.Parse(`[
{"type":"message","role":"user","content":"hello"}, {"type":"message","role":"user","content":"hello"},
{"type":"message","role":"assistant","content":"hi there"} {"type":"message","role":"assistant","content":"hi there"}
]`) ]`)
if !inputContainsFullTranscript(input) { if inputContainsFullTranscript(input) {
t.Fatal("expected full transcript when assistant message is present") t.Fatal("assistant message alone must not be treated as full transcript")
} }
} }
@@ -1501,3 +1501,33 @@ func TestNormalizeSubsequentRequestIncrementalInputStillMerges(t *testing.T) {
} }
} }
} }
func TestNormalizeSubsequentRequestAssistantIncrementalInputStillMerges(t *testing.T) {
lastRequest := []byte(`{"model":"gpt-5.4","stream":true,"input":[
{"type":"message","role":"user","id":"msg-1","content":"hello"}
]}`)
lastResponseOutput := []byte(`[
{"type":"message","role":"assistant","id":"msg-2","content":"prior assistant"},
{"type":"function_call","id":"fc-1","call_id":"call-1","name":"bash","arguments":"{}"}
]`)
raw := []byte(`{"type":"response.append","input":[
{"type":"message","role":"assistant","id":"msg-3","content":"patched assistant turn"}
]}`)
normalized, _, errMsg := normalizeResponsesWebsocketRequest(raw, lastRequest, lastResponseOutput)
if errMsg != nil {
t.Fatalf("unexpected error: %v", errMsg.Error)
}
input := gjson.GetBytes(normalized, "input").Array()
if len(input) != 4 {
t.Fatalf("input len = %d, want 4 (merged)", len(input))
}
wantIDs := []string{"msg-1", "msg-2", "fc-1", "msg-3"}
for i, want := range wantIDs {
got := input[i].Get("id").String()
if got != want {
t.Fatalf("input[%d].id = %q, want %q", i, got, want)
}
}
}