feat(websockets): add transcript state tracking and compaction trigger support

- Added methods for managing and tracking WebSocket transcript state, including recording, prepending, and replacing transcript inputs.
- Implemented `executeCompactionTriggerFromWebsocketContext` to support compaction triggers using recorded transcript context.
- Enhanced upstream-downstream ID mapping with additional utilities and state synchronization.
- Expanded test coverage to validate transcript state management, compaction payload generation, and WebSocket response handling.
This commit is contained in:
Luis Pater
2026-06-15 10:41:35 +08:00
parent ea90ab6f77
commit f33bc56bb9
2 changed files with 365 additions and 5 deletions
@@ -52,6 +52,7 @@ type xaiWebsocketIDState struct {
mu sync.Mutex
downstreamToUpstream map[string]string
sequence int
transcriptInput []json.RawMessage
}
type xaiWebsocketRequestIDMapper struct {
@@ -124,12 +125,124 @@ func (s *xaiWebsocketIDState) upstreamIDForDownstream(downstreamID string) strin
}
s.mu.Lock()
defer s.mu.Unlock()
if upstreamID := strings.TrimSpace(s.downstreamToUpstream[downstreamID]); upstreamID != "" {
return upstreamID
if upstreamID, ok := s.downstreamToUpstream[downstreamID]; ok {
return strings.TrimSpace(upstreamID)
}
return downstreamID
}
func (s *xaiWebsocketIDState) mapDownstreamToUpstream(downstreamID string, upstreamID string) {
downstreamID = strings.TrimSpace(downstreamID)
if s == nil || downstreamID == "" {
return
}
s.mu.Lock()
if s.downstreamToUpstream == nil {
s.downstreamToUpstream = make(map[string]string)
}
s.downstreamToUpstream[downstreamID] = strings.TrimSpace(upstreamID)
s.mu.Unlock()
}
func (s *xaiWebsocketIDState) snapshotTranscriptInput() []byte {
if s == nil {
return nil
}
s.mu.Lock()
defer s.mu.Unlock()
if len(s.transcriptInput) == 0 {
return nil
}
return xaiMarshalRawMessages(s.transcriptInput)
}
func (s *xaiWebsocketIDState) prependTranscriptInput(payload []byte) []byte {
if s == nil || len(payload) == 0 {
return payload
}
s.mu.Lock()
prefix := make([]json.RawMessage, 0, len(s.transcriptInput))
for _, item := range s.transcriptInput {
prefix = append(prefix, bytes.Clone(item))
}
s.mu.Unlock()
if len(prefix) == 0 {
return payload
}
current := xaiJSONRawMessages(gjson.GetBytes(payload, "input"))
merged := append(prefix, current...)
out, errSet := sjson.SetRawBytes(payload, "input", xaiMarshalRawMessages(merged))
if errSet != nil {
return payload
}
return out
}
func (s *xaiWebsocketIDState) recordTranscriptTurn(requestPayload []byte, completedPayload []byte) {
if s == nil || len(requestPayload) == 0 || len(completedPayload) == 0 {
return
}
inputItems := xaiJSONRawMessages(gjson.GetBytes(requestPayload, "input"))
outputItems := xaiJSONRawMessages(gjson.GetBytes(completedPayload, "response.output"))
if len(inputItems) == 0 && len(outputItems) == 0 {
return
}
s.mu.Lock()
defer s.mu.Unlock()
if strings.TrimSpace(gjson.GetBytes(requestPayload, "previous_response_id").String()) == "" {
s.transcriptInput = nil
}
s.transcriptInput = append(s.transcriptInput, inputItems...)
s.transcriptInput = append(s.transcriptInput, outputItems...)
}
func (s *xaiWebsocketIDState) replaceTranscriptWithItems(items ...[]byte) {
if s == nil {
return
}
next := make([]json.RawMessage, 0, len(items))
for _, item := range items {
item = bytes.TrimSpace(item)
if len(item) == 0 || !json.Valid(item) {
continue
}
next = append(next, bytes.Clone(item))
}
s.mu.Lock()
s.transcriptInput = next
s.mu.Unlock()
}
func xaiJSONRawMessages(result gjson.Result) []json.RawMessage {
if !result.Exists() || !result.IsArray() {
return nil
}
items := result.Array()
out := make([]json.RawMessage, 0, len(items))
for _, item := range items {
raw := bytes.TrimSpace([]byte(item.Raw))
if len(raw) == 0 || !json.Valid(raw) {
continue
}
out = append(out, bytes.Clone(raw))
}
return out
}
func xaiMarshalRawMessages(items []json.RawMessage) []byte {
var buf bytes.Buffer
buf.WriteByte('[')
for i, item := range items {
if i > 0 {
buf.WriteByte(',')
}
buf.Write(bytes.TrimSpace(item))
}
buf.WriteByte(']')
return buf.Bytes()
}
func (m *xaiWebsocketRequestIDMapper) upstreamRequestPayload(payload []byte) []byte {
if m == nil || len(payload) == 0 || m.downstreamPreviousID == m.upstreamPreviousID {
return payload
@@ -137,6 +250,9 @@ func (m *xaiWebsocketRequestIDMapper) upstreamRequestPayload(payload []byte) []b
if m.upstreamPreviousID == "" {
out, errDelete := sjson.DeleteBytes(payload, "previous_response_id")
if errDelete == nil {
if m.downstreamPreviousID != "" && m.state != nil {
out = m.state.prependTranscriptInput(out)
}
return out
}
return payload
@@ -275,12 +391,16 @@ func (e *XAIWebsocketsExecutor) ExecuteStream(ctx context.Context, auth *cliprox
if opts.Alt == "responses/compact" {
return nil, statusErr{code: http.StatusBadRequest, msg: "streaming not supported for /responses/compact"}
}
executionSessionID := executionSessionIDFromOptions(opts)
stateSessionID := xaiExecutionSessionID(req, opts)
if stateSessionID == "" {
stateSessionID = executionSessionID
}
idMapper := newXAIWebsocketRequestIDMapper(e.idStore, stateSessionID, req.Payload)
if xaiInputHasItemType(req.Payload, "compaction_trigger") {
return e.XAIExecutor.ExecuteStream(ctx, auth, req, opts)
return e.executeCompactionTriggerFromWebsocketContext(ctx, auth, req, opts, idMapper)
}
executionSessionID := executionSessionIDFromOptions(opts)
idMapper := newXAIWebsocketRequestIDMapper(e.idStore, executionSessionID, req.Payload)
token, baseURL := xaiCreds(auth)
if baseURL == "" {
baseURL = xaiauth.DefaultAPIBaseURL
@@ -450,6 +570,7 @@ func (e *XAIWebsocketsExecutor) ExecuteStream(ctx context.Context, auth *cliprox
var param any
outputItemsByIndex := make(map[int64][]byte)
var outputItemsFallback [][]byte
recordedTranscript := false
for {
if ctx != nil && ctx.Err() != nil {
terminateReason = "context_done"
@@ -524,11 +645,19 @@ func (e *XAIWebsocketsExecutor) ExecuteStream(ctx context.Context, auth *cliprox
reporter.Publish(ctx, detail)
}
payload = xaiPatchCompletedOutput(payload, outputItemsByIndex, outputItemsFallback)
if !warmupRequest && idMapper != nil && idMapper.state != nil && !recordedTranscript {
idMapper.state.recordTranscriptTurn(wsReqBody, payload)
recordedTranscript = true
}
case "response.done":
logXAIWebsocketTerminalResponse(executionSessionID, authID, wsURL, eventType, payload)
if detail, ok := helps.ParseCodexUsage(payload); ok {
reporter.Publish(ctx, detail)
}
if !warmupRequest && idMapper != nil && idMapper.state != nil && !recordedTranscript {
idMapper.state.recordTranscriptTurn(wsReqBody, payload)
recordedTranscript = true
}
}
if cliproxyexecutor.DownstreamWebsocket(ctx) {
@@ -589,6 +718,72 @@ func (e *XAIWebsocketsExecutor) ExecuteStream(ctx context.Context, auth *cliprox
return &cliproxyexecutor.StreamResult{Headers: upstreamHeaders, Chunks: out}, nil
}
func (e *XAIWebsocketsExecutor) executeCompactionTriggerFromWebsocketContext(ctx context.Context, auth *cliproxyauth.Auth, req cliproxyexecutor.Request, opts cliproxyexecutor.Options, idMapper *xaiWebsocketRequestIDMapper) (*cliproxyexecutor.StreamResult, error) {
if idMapper == nil || idMapper.state == nil {
return nil, statusErr{code: http.StatusBadRequest, msg: "xai websocket compaction context is unavailable"}
}
transcriptInput := idMapper.state.snapshotTranscriptInput()
if len(transcriptInput) == 0 {
return nil, statusErr{code: http.StatusBadRequest, msg: "xai websocket compaction context is empty"}
}
authID := ""
if auth != nil {
authID = auth.ID
}
log.Infof(
"xai websockets: compact fallback session=%s auth=%s input_items=%d",
xaiExecutionSessionID(req, opts),
strings.TrimSpace(authID),
len(gjson.ParseBytes(transcriptInput).Array()),
)
compactPayload, err := buildXAIWebsocketCompactionPayload(req.Payload, transcriptInput)
if err != nil {
return nil, err
}
compactReq := req
compactReq.Payload = compactPayload
prepared, data, headers, err := e.XAIExecutor.executeCompactRequest(ctx, auth, compactReq, opts)
if err != nil {
return nil, err
}
responseID := xaiCompactionResponseID(data)
idMapper.state.replaceTranscriptWithItems(xaiCompactionOutputItem(data, responseID))
idMapper.state.mapDownstreamToUpstream(responseID, "")
headers = headers.Clone()
if headers == nil {
headers = make(http.Header)
}
headers.Set("Content-Type", "text/event-stream")
chunks := xaiBuildCompactionTriggerStreamChunks(prepared, data)
out := make(chan cliproxyexecutor.StreamChunk, len(chunks))
for _, chunk := range chunks {
out <- cliproxyexecutor.StreamChunk{Payload: chunk}
}
close(out)
return &cliproxyexecutor.StreamResult{Headers: headers, Chunks: out}, nil
}
func buildXAIWebsocketCompactionPayload(payload []byte, transcriptInput []byte) ([]byte, error) {
if len(payload) == 0 {
payload = []byte(`{}`)
}
if len(transcriptInput) == 0 {
transcriptInput = []byte("[]")
}
out := bytes.Clone(payload)
var err error
out, err = sjson.SetRawBytes(out, "input", transcriptInput)
if err != nil {
return nil, err
}
out, _ = sjson.DeleteBytes(out, "previous_response_id")
return out, nil
}
func xaiWebsocketGenerateFalse(payload []byte) bool {
generate := gjson.GetBytes(payload, "generate")
return generate.Exists() && !generate.Bool()
@@ -4,6 +4,7 @@ import (
"bytes"
"context"
"fmt"
"io"
"net/http"
"net/http/httptest"
"strings"
@@ -245,6 +246,170 @@ func TestXAIWebsocketsExecuteStreamRewritesRepeatedResponseIDForDownstream(t *te
}
}
func TestXAIWebsocketsExecuteStreamCompactionTriggerUsesHTTPCompactWithRecordedContext(t *testing.T) {
upgrader := websocket.Upgrader{CheckOrigin: func(*http.Request) bool { return true }}
capturedWebsocketPayload := make(chan []byte, 1)
capturedCompactPayload := make(chan []byte, 1)
server := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
switch r.URL.Path {
case "/responses":
conn, err := upgrader.Upgrade(w, r, nil)
if err != nil {
t.Errorf("upgrade websocket: %v", err)
return
}
defer func() { _ = conn.Close() }()
for i := 0; i < 2; i++ {
_ = conn.SetReadDeadline(time.Now().Add(5 * time.Second))
_, payload, errRead := conn.ReadMessage()
if errRead != nil {
t.Errorf("read upstream websocket message: %v", errRead)
return
}
capturedWebsocketPayload <- bytes.Clone(payload)
completed := []byte(`{"type":"response.completed","response":{"id":"resp-real","output":[{"type":"message","id":"out-1","role":"assistant","content":"first answer"}],"usage":{"input_tokens":0,"output_tokens":0,"total_tokens":0}}}`)
if i == 1 {
completed = []byte(`{"type":"response.completed","response":{"id":"resp-after-compact","output":[{"type":"message","id":"out-2","role":"assistant","content":"second answer"}],"usage":{"input_tokens":0,"output_tokens":0,"total_tokens":0}}}`)
}
if errWrite := conn.WriteMessage(websocket.TextMessage, completed); errWrite != nil {
t.Errorf("write completed websocket message: %v", errWrite)
return
}
}
case "/responses/compact":
body, errRead := io.ReadAll(r.Body)
if errRead != nil {
t.Errorf("read compact body: %v", errRead)
return
}
capturedCompactPayload <- bytes.Clone(body)
w.Header().Set("Content-Type", "application/json")
_, _ = w.Write([]byte(`{"id":"resp_compact","model":"grok-4.3","output":[{"type":"compaction","encrypted_content":"opaque"}],"usage":{"input_tokens":1,"output_tokens":2,"total_tokens":3}}`))
default:
t.Errorf("path = %q, want /responses", r.URL.Path)
http.Error(w, "unexpected path", http.StatusNotFound)
}
}))
defer server.Close()
exec := NewXAIWebsocketsExecutor(&config.Config{})
exec.store = &codexWebsocketSessionStore{sessions: make(map[string]*codexWebsocketSession)}
exec.idStore = &xaiWebsocketIDStateStore{sessions: make(map[string]*xaiWebsocketIDState)}
auth := &cliproxyauth.Auth{
ID: "xai-auth-compaction",
Provider: "xai",
Attributes: map[string]string{
"base_url": server.URL,
"websockets": "true",
},
Metadata: map[string]any{"access_token": "xai-token"},
}
opts := cliproxyexecutor.Options{
SourceFormat: sdktranslator.FormatOpenAIResponse,
ResponseFormat: sdktranslator.FormatOpenAIResponse,
Stream: true,
Metadata: map[string]any{
cliproxyexecutor.ExecutionSessionMetadataKey: "xai-compaction-session",
},
}
result, err := exec.ExecuteStream(cliproxyexecutor.WithDownstreamWebsocket(context.Background()), auth, cliproxyexecutor.Request{
Model: "grok-4.3",
Payload: []byte(`{"model":"grok-4.3","stream":true,"input":[{"type":"message","id":"msg-1","role":"user","content":"first"}]}`),
}, opts)
if err != nil {
t.Fatalf("ExecuteStream first turn error: %v", err)
}
for chunk := range result.Chunks {
if chunk.Err != nil {
t.Fatalf("stream chunk error = %v", chunk.Err)
}
}
select {
case payload := <-capturedWebsocketPayload:
if got := gjson.GetBytes(payload, "type").String(); got != "response.create" {
t.Fatalf("type = %q, want response.create; payload=%s", got, payload)
}
input := gjson.GetBytes(payload, "input")
if !input.IsArray() || len(input.Array()) != 1 {
t.Fatalf("input = %s, want one first-turn item", input.Raw)
}
if gjson.GetBytes(payload, "stream").Exists() {
t.Fatalf("stream must be omitted for xAI websocket payload: %s", payload)
}
case <-time.After(5 * time.Second):
t.Fatal("timed out waiting for upstream websocket payload")
}
compactResult, err := exec.ExecuteStream(cliproxyexecutor.WithDownstreamWebsocket(context.Background()), auth, cliproxyexecutor.Request{
Model: "grok-4.3",
Payload: []byte(`{"model":"grok-4.3","stream":true,"previous_response_id":"resp-real-xai-1","input":[{"type":"compaction_trigger"}]}`),
}, opts)
if err != nil {
t.Fatalf("ExecuteStream compaction trigger error: %v", err)
}
for chunk := range compactResult.Chunks {
if chunk.Err != nil {
t.Fatalf("compact stream chunk error = %v", chunk.Err)
}
}
select {
case payload := <-capturedCompactPayload:
if xaiInputHasItemType(payload, "compaction_trigger") {
t.Fatalf("compaction_trigger reached xai compact body: %s", payload)
}
input := gjson.GetBytes(payload, "input")
if !input.IsArray() || len(input.Array()) != 2 {
t.Fatalf("compact input = %s, want first request input plus response output", input.Raw)
}
if got := input.Array()[0].Get("id").String(); got != "msg-1" {
t.Fatalf("compact input[0].id = %q, want msg-1; payload=%s", got, payload)
}
if got := input.Array()[1].Get("id").String(); got != "out-1" {
t.Fatalf("compact input[1].id = %q, want out-1; payload=%s", got, payload)
}
if got := gjson.GetBytes(payload, "previous_response_id").String(); got != "" {
t.Fatalf("compact previous_response_id = %q, want empty; payload=%s", got, payload)
}
case <-time.After(5 * time.Second):
t.Fatal("timed out waiting for compact HTTP payload")
}
nextResult, err := exec.ExecuteStream(cliproxyexecutor.WithDownstreamWebsocket(context.Background()), auth, cliproxyexecutor.Request{
Model: "grok-4.3",
Payload: []byte(`{"model":"grok-4.3","stream":true,"previous_response_id":"resp_compact","input":[{"type":"message","id":"msg-2","role":"user","content":"second"}]}`),
}, opts)
if err != nil {
t.Fatalf("ExecuteStream post-compaction turn error: %v", err)
}
for chunk := range nextResult.Chunks {
if chunk.Err != nil {
t.Fatalf("post-compaction stream chunk error = %v", chunk.Err)
}
}
select {
case payload := <-capturedWebsocketPayload:
if got := gjson.GetBytes(payload, "previous_response_id").String(); got != "" {
t.Fatalf("post-compaction previous_response_id = %q, want empty; payload=%s", got, payload)
}
input := gjson.GetBytes(payload, "input")
if !input.IsArray() || len(input.Array()) != 2 {
t.Fatalf("post-compaction input = %s, want compaction item plus new message", input.Raw)
}
if got := input.Array()[0].Get("type").String(); got != "compaction" {
t.Fatalf("post-compaction input[0].type = %q, want compaction; payload=%s", got, payload)
}
if got := input.Array()[1].Get("id").String(); got != "msg-2" {
t.Fatalf("post-compaction input[1].id = %q, want msg-2; payload=%s", got, payload)
}
case <-time.After(5 * time.Second):
t.Fatal("timed out waiting for post-compaction websocket payload")
}
}
func TestBuildXAIWebsocketRequestBodySetsStoreAndKeepsPromptCacheKey(t *testing.T) {
body := []byte(`{"model":"grok-4.3","stream":true,"stream_options":{"include_usage":true},"background":true,"prompt_cache_key":"cache-1","previous_response_id":"resp-prev","instructions":"system prompt","input":[{"type":"message","role":"user","content":"hello"}]}`)