Files
Agent-Native/runtime/llm/sse_test.go
T
weichen 6a1dcddb48
ci / docker-checkout-hostgateway (pull_request) Successful in 1s
ci / vet-and-test (pull_request) Successful in 52s
Record content-free stream terminal observations
Why: The r0.4 trial could prove that the parser missed [DONE] but could not distinguish clean EOF, malformed frames, read failure, or provider error without retaining raw SSE.

Evidence: RED receipts fast-slow-u3-r05-task1-terminal-red.json and task1-http-error-red.json; GREEN targeted receipt and full vet/test/race receipt fast-slow-u3-r05-task1-agent-native-full.json.

Impact: Add bounded scalar diagnostics and successful HTTP status to terminal events while preserving the existing incomplete-tool rejection, returned Usage, and retry behavior.

Refs: U3 r0.5 approved design and plan; Agent-Native PR #46
2026-10-14 07:50:39 +00:00

544 lines
18 KiB
Go

package llm
import (
"context"
"encoding/json"
"io"
"net/http"
"net/http/httptest"
"strings"
"testing"
"time"
)
func TestParseSSETextAndUsage(t *testing.T) {
body := strings.Join([]string{
`data: {"choices":[{"delta":{"content":"你"},"index":0}]}`,
`data: {"choices":[{"delta":{"content":"好"},"finish_reason":"stop"}],"usage":{"prompt_tokens":3,"completion_tokens":2,"total_tokens":5}}`,
`data: [DONE]`,
"",
}, "\n")
var deltas []string
var end Event
for ev := range ParseSSE(context.Background(), strings.NewReader(body)) {
switch ev.Kind {
case KindTextDelta:
deltas = append(deltas, ev.Text)
case KindMessageEnd:
end = ev
default:
t.Fatalf("unexpected event %#v", ev)
}
}
if strings.Join(deltas, "") != "你好" {
t.Fatalf("deltas %v", deltas)
}
if end.Message.StopReason != StopEnd || end.Message.Content != "你好" {
t.Fatalf("end %#v", end.Message)
}
if end.Usage.TotalTokens != 5 {
t.Fatalf("usage %#v", end.Usage)
}
if end.FinishReason != "stop" {
t.Fatalf("finish reason %q, want stop", end.FinishReason)
}
}
func TestParseSSEFinishReasonLengthKeepsStopEnd(t *testing.T) {
body := strings.Join([]string{
`data: {"choices":[{"delta":{"content":"半"},"finish_reason":"length"}]}`,
`data: [DONE]`,
"",
}, "\n")
var end Event
for ev := range ParseSSE(context.Background(), strings.NewReader(body)) {
if ev.Kind == KindMessageEnd {
end = ev
}
}
if end.FinishReason != "length" {
t.Fatalf("finish reason %q, want length", end.FinishReason)
}
if end.Message.StopReason != StopEnd {
t.Fatalf("stop reason %s, want end so existing callers stay unchanged", end.Message.StopReason)
}
}
func TestParseSSEToolCallIncrements(t *testing.T) {
body := strings.Join([]string{
`data: {"choices":[{"delta":{"tool_calls":[{"index":0,"id":"call_1","function":{"name":"read_file","arguments":""}}]}}]}`,
`data: {"choices":[{"delta":{"tool_calls":[{"index":0,"function":{"arguments":"{\"path\":"}}]}}]}`,
`data: {"choices":[{"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"a.md\"}"}}]},"finish_reason":"tool_calls"}]}`,
`data: [DONE]`,
"",
}, "\n")
var end Event
deltas := 0
for ev := range ParseSSE(context.Background(), strings.NewReader(body)) {
if ev.Kind == KindToolCallDelta {
deltas++
}
if ev.Kind == KindMessageEnd {
end = ev
}
}
if deltas != 3 {
t.Fatalf("tool deltas %d", deltas)
}
if end.Message.StopReason != StopToolCalls || len(end.Message.ToolCalls) != 1 {
t.Fatalf("end %#v", end.Message)
}
if end.TerminalObservation == nil || end.TerminalObservation.ReadOutcome != "done_marker" ||
!end.TerminalObservation.ToolArgumentsValid || end.TerminalObservation.ToolFragments != 3 {
t.Fatalf("valid multi-fragment terminal was not classified: %#v", end.TerminalObservation)
}
tc := end.Message.ToolCalls[0]
if tc.ID != "call_1" || tc.Name != "read_file" || tc.Arguments != `{"path":"a.md"}` {
t.Fatalf("tool call %#v", tc)
}
}
func TestParseSSEToolFinishWithoutCallFails(t *testing.T) {
for _, reason := range []string{"tool_calls", "function_call"} {
t.Run(reason, func(t *testing.T) {
body := strings.Join([]string{
`data: {"choices":[{"delta":{},"finish_reason":"` + reason + `"}],"usage":{"prompt_tokens":12,"completion_tokens":5,"total_tokens":17}}`,
`data: [DONE]`,
"",
}, "\n")
var end Event
for ev := range ParseSSE(context.Background(), strings.NewReader(body)) {
if ev.Kind == KindMessageEnd {
end = ev
}
}
if end.Err == nil || end.Message.StopReason != StopError || end.Usage.TotalTokens != 17 ||
!end.StreamComplete || len(end.Message.ToolCalls) != 0 {
t.Fatalf("empty tool finish must fail and retain Usage: %#v", end)
}
})
}
}
func TestParseSSELengthFailsUnfinishedToolCall(t *testing.T) {
body := strings.Join([]string{
`data: {"choices":[{"delta":{"tool_calls":[{"index":0,"id":"call_1","function":{"name":"read_file","arguments":"{\"path\":"}}]}}]}`,
`data: {"choices":[{"delta":{},"finish_reason":"length"}]}`,
`data: [DONE]`,
"",
}, "\n")
var end Event
for ev := range ParseSSE(context.Background(), strings.NewReader(body)) {
if ev.Kind == KindMessageEnd {
end = ev
}
}
if end.Message.StopReason != StopError || len(end.Message.ToolCalls) != 0 || end.Err == nil {
t.Fatalf("truncated tool call must fail without execution: %#v", end)
}
if end.FinishReason != "length" {
t.Fatalf("finish %q", end.FinishReason)
}
}
func TestParseSSEUncleanEndFailsToolCall(t *testing.T) {
body := `data: {"choices":[{"delta":{"tool_calls":[{"index":0,"id":"call_1","function":{"name":"read_file","arguments":"{\"path\":"}}]}}]}`
var end Event
for ev := range ParseSSE(context.Background(), strings.NewReader(body)) {
if ev.Kind == KindMessageEnd {
end = ev
}
}
if end.Kind != KindMessageEnd || end.Message.StopReason != StopError || len(end.Message.ToolCalls) != 0 || end.Err == nil {
t.Fatalf("unclean stream must fail without executing tools: %#v", end)
}
}
func TestParseSSEToolFinishWithoutDone(t *testing.T) {
body := strings.Join([]string{
`data: {"choices":[{"delta":{"tool_calls":[{"index":0,"id":"call_1","function":{"name":"echo","arguments":"{}"}}]}}]}`,
`data: {"choices":[{"delta":{},"finish_reason":"tool_calls"}],"usage":{"prompt_tokens":12,"completion_tokens":5,"total_tokens":17}}`,
"",
}, "\n\n")
var end Event
for ev := range ParseSSE(context.Background(), strings.NewReader(body)) {
if ev.Kind == KindMessageEnd {
end = ev
}
}
if end.Kind != KindMessageEnd || end.Message.StopReason != StopError ||
len(end.Message.ToolCalls) != 0 || end.Err == nil || end.Retryable ||
end.Usage.TotalTokens != 17 {
t.Fatalf("unsafe tool call from unfinished stream: %#v", end)
}
}
func TestTerminalObservationCleanEOFWithoutDone(t *testing.T) {
body := strings.Join([]string{
`data: {"choices":[{"delta":{"content":"PRIVATE_PROMPT_R05","tool_calls":[{"index":0,"id":"call_1","function":{"name":"PRIVATE_TOOL_R05","arguments":"{}"}}]}}]}`,
`data: {"choices":[{"delta":{},"finish_reason":"tool_calls"}],"usage":{"prompt_tokens":12,"completion_tokens":5,"total_tokens":17}}`,
"",
}, "\n\n")
var end Event
for ev := range ParseSSE(context.Background(), strings.NewReader(body)) {
if ev.Kind == KindMessageEnd {
end = ev
}
}
if end.Err == nil || end.StreamComplete || len(end.Message.ToolCalls) != 0 || end.Usage.TotalTokens != 17 {
t.Fatalf("unfinished tool stream was released or unmetered: %#v", end)
}
meta := end.TerminalObservation
if meta == nil || meta.ReadOutcome != "clean_eof" || !meta.LastFrameJSONComplete ||
!meta.UsagePresent || meta.ToolFragments != 1 || meta.ToolArgumentBytes != 2 || !meta.ToolArgumentsValid {
t.Fatalf("missing content-free clean EOF evidence: %#v", meta)
}
raw, err := json.Marshal(meta)
if err != nil {
t.Fatal(err)
}
if strings.Contains(string(raw), "PRIVATE_PROMPT_R05") || strings.Contains(string(raw), "PRIVATE_TOOL_R05") {
t.Fatalf("terminal observation retained model content: %s", raw)
}
}
func TestTerminalObservationMalformedAndReadError(t *testing.T) {
usageAndTool := `data: {"choices":[{"delta":{"tool_calls":[{"index":0,"id":"call_1","function":{"name":"echo","arguments":"{}"}}]},"finish_reason":"tool_calls"}],"usage":{"prompt_tokens":12,"completion_tokens":5,"total_tokens":17}}` + "\n\n"
tests := []struct {
name string
ctx context.Context
reader io.Reader
outcome string
usage int
frameValid bool
}{
{"invalid-frame", context.Background(), strings.NewReader(usageAndTool + "data: {not-json}\n\n"), "invalid_frame", 17, false},
{"read-error", context.Background(), &errorAfterDataReader{data: strings.NewReader(usageAndTool)}, "read_error", 17, true},
}
ctx, cancel := context.WithCancel(context.Background())
cancel()
tests = append(tests, struct {
name string
ctx context.Context
reader io.Reader
outcome string
usage int
frameValid bool
}{"cancelled", ctx, strings.NewReader(usageAndTool), "cancelled", 0, false})
for _, tc := range tests {
t.Run(tc.name, func(t *testing.T) {
var end Event
for ev := range ParseSSE(tc.ctx, tc.reader) {
if ev.Kind == KindMessageEnd || ev.Kind == KindError {
end = ev
}
}
if end.Err == nil || len(end.Message.ToolCalls) != 0 || end.Usage.TotalTokens != tc.usage {
t.Fatalf("failure lost Usage or released tools: %#v", end)
}
meta := end.TerminalObservation
if meta == nil || meta.ReadOutcome != tc.outcome || meta.LastFrameJSONComplete != tc.frameValid {
t.Fatalf("wrong terminal classification: %#v", meta)
}
})
}
}
func TestParseSSEInvalidToolArguments(t *testing.T) {
body := strings.Join([]string{
`data: {"choices":[{"delta":{"tool_calls":[{"index":0,"id":"call_1","function":{"name":"publish_work_checkpoint","arguments":"{\"findings\":"}}]}}]}`,
`data: {"choices":[{"delta":{},"finish_reason":"tool_calls"}],"usage":{"prompt_tokens":12,"completion_tokens":5,"total_tokens":17}}`,
`data: [DONE]`,
"",
}, "\n\n")
var end Event
for ev := range ParseSSE(context.Background(), strings.NewReader(body)) {
if ev.Kind == KindMessageEnd {
end = ev
}
}
if end.Kind != KindMessageEnd || end.Message.StopReason != StopError ||
len(end.Message.ToolCalls) != 0 || end.Err == nil || end.Retryable ||
end.Usage.TotalTokens != 17 {
t.Fatalf("unsafe tool call with invalid JSON arguments: %#v", end)
}
if end.TerminalObservation == nil || end.TerminalObservation.ToolArgumentsValid ||
end.TerminalObservation.ReadOutcome != "done_marker" {
t.Fatalf("invalid arguments incorrectly classified: %#v", end.TerminalObservation)
}
}
func TestParseSSERejectsMissingToolIdentity(t *testing.T) {
for _, tc := range []struct{ name, firstChunk string }{
{"missing-id", `data: {"choices":[{"delta":{"tool_calls":[{"index":0,"function":{"name":"echo","arguments":"{}"}}]}}]}`},
{"missing-name", `data: {"choices":[{"delta":{"tool_calls":[{"index":0,"id":"call_1","function":{"arguments":"{}"}}]}}]}`},
} {
t.Run(tc.name, func(t *testing.T) {
body := strings.Join([]string{
tc.firstChunk,
`data: {"choices":[{"delta":{},"finish_reason":"tool_calls"}],"usage":{"prompt_tokens":12,"completion_tokens":5,"total_tokens":17}}`,
`data: [DONE]`,
"",
}, "\n\n")
var end Event
for ev := range ParseSSE(context.Background(), strings.NewReader(body)) {
if ev.Kind == KindMessageEnd {
end = ev
}
}
if end.Message.StopReason != StopError || len(end.Message.ToolCalls) != 0 || end.Err == nil {
t.Fatalf("missing tool identity reached execution: %#v", end)
}
})
}
}
func TestParseSSERejectsNonObjectToolArguments(t *testing.T) {
for _, tc := range []struct{ name, arguments string }{
{"array", "[]"},
{"null", "null"},
} {
t.Run(tc.name, func(t *testing.T) {
body := strings.Join([]string{
`data: {"choices":[{"delta":{"tool_calls":[{"index":0,"id":"call_1","function":{"name":"echo","arguments":"` + tc.arguments + `"}}]}}]}`,
`data: {"choices":[{"delta":{},"finish_reason":"tool_calls"}],"usage":{"prompt_tokens":12,"completion_tokens":5,"total_tokens":17}}`,
`data: [DONE]`,
"",
}, "\n\n")
var end Event
for ev := range ParseSSE(context.Background(), strings.NewReader(body)) {
if ev.Kind == KindMessageEnd {
end = ev
}
}
if end.Message.StopReason != StopError || len(end.Message.ToolCalls) != 0 || end.Err == nil {
t.Fatalf("non-object tool arguments reached execution: %#v", end)
}
})
}
}
func TestParseSSECapacityErrorIsRetryable(t *testing.T) {
body := "data: {\"error\":{\"message\":\"server is at capacity\"}}\n\n"
var got Event
for ev := range ParseSSE(context.Background(), strings.NewReader(body)) {
got = ev
}
if got.Kind != KindError || !got.Retryable || !CapacityLimited(got.ErrorMessage) {
t.Fatalf("event %#v", got)
}
}
func TestParseSSEMalformedIsError(t *testing.T) {
body := "data: {not-json}\n\n"
var got Event
for ev := range ParseSSE(context.Background(), strings.NewReader(body)) {
got = ev
}
if got.Kind != KindError || got.Message.StopReason != StopError || got.Retryable {
t.Fatalf("event %#v", got)
}
}
func TestParseSSEMalformedChunkAfterUsageKeepsUsage(t *testing.T) {
body := strings.Join([]string{
`data: {"choices":[{"delta":{"tool_calls":[{"index":0,"id":"call_1","function":{"name":"echo","arguments":"{}"}}]},"finish_reason":"tool_calls"}],"usage":{"prompt_tokens":12,"completion_tokens":5,"total_tokens":17}}`,
`data: {not-json}`,
"",
}, "\n\n")
var end Event
for ev := range ParseSSE(context.Background(), strings.NewReader(body)) {
end = ev
}
if end.Kind != KindMessageEnd || end.Message.StopReason != StopError ||
len(end.Message.ToolCalls) != 0 || end.Err == nil || end.Retryable ||
end.Usage.TotalTokens != 17 {
t.Fatalf("malformed chunk dropped already returned usage: %#v", end)
}
}
func TestParseSSEProviderErrorAfterUsageKeepsUsage(t *testing.T) {
body := strings.Join([]string{
`data: {"choices":[{"delta":{"tool_calls":[{"index":0,"id":"call_1","function":{"name":"echo","arguments":"{}"}}]}}],"usage":{"prompt_tokens":12,"completion_tokens":5,"total_tokens":17}}`,
`data: {"error":{"message":"server is at capacity"}}`,
"",
}, "\n\n")
var end Event
for ev := range ParseSSE(context.Background(), strings.NewReader(body)) {
end = ev
}
if end.Kind != KindMessageEnd || end.Message.StopReason != StopError ||
len(end.Message.ToolCalls) != 0 || end.Err == nil || end.Retryable ||
end.Usage.TotalTokens != 17 {
t.Fatalf("provider error after usage retried or lost its charge: %#v", end)
}
}
type errorAfterDataReader struct{ data *strings.Reader }
func (r *errorAfterDataReader) Read(p []byte) (int, error) {
n, err := r.data.Read(p)
if err == io.EOF {
return 0, io.ErrUnexpectedEOF
}
return n, err
}
func TestParseSSEReaderErrorAfterUsageKeepsUsage(t *testing.T) {
body := strings.Join([]string{
`data: {"choices":[{"delta":{"tool_calls":[{"index":0,"id":"call_1","function":{"name":"echo","arguments":"{}"}}]}}],"usage":{"prompt_tokens":12,"completion_tokens":5,"total_tokens":17}}`,
"",
}, "\n\n")
var end Event
for ev := range ParseSSE(context.Background(), &errorAfterDataReader{data: strings.NewReader(body)}) {
end = ev
}
if end.Kind != KindMessageEnd || end.Message.StopReason != StopError ||
len(end.Message.ToolCalls) != 0 || end.Err == nil || end.Retryable ||
end.Usage.TotalTokens != 17 {
t.Fatalf("reader error after usage retried or lost its charge: %#v", end)
}
}
func TestParseSSECancel(t *testing.T) {
pr, pw := io.Pipe()
ctx, cancel := context.WithCancel(context.Background())
ch := ParseSSE(ctx, pr)
go func() {
_, _ = io.WriteString(pw, "data: {\"choices\":[{\"delta\":{\"content\":\"x\"}}]}\n\n")
cancel()
time.Sleep(20 * time.Millisecond)
_ = pw.Close()
}()
sawAbort := false
for ev := range ch {
if ev.Kind == KindError && ev.Message.StopReason == StopAborted {
sawAbort = true
}
}
if !sawAbort {
t.Fatal("expected aborted event")
}
}
func TestClientHTTPErrorIsRetryable(t *testing.T) {
srv := httpServer(t, func(w http.ResponseWriter, r *http.Request) {
if r.Header.Get("Authorization") != "Bearer secret" {
t.Errorf("auth header %q", r.Header.Get("Authorization"))
}
http.Error(w, "busy", http.StatusTooManyRequests)
})
c := &Client{BaseURL: srv, APIKey: "secret", HTTPClient: srvClient()}
var got Event
for ev := range c.Stream(context.Background(), "MiniMax-M2.7-highspeed", []Message{{Role: RoleUser, Content: "hi"}}, nil, StreamOptions{}) {
got = ev
}
if got.Kind != KindError || !got.Retryable || got.Message.StopReason != StopError {
t.Fatalf("event %#v", got)
}
}
func TestClientStreamsCompletion(t *testing.T) {
srv := httpServer(t, func(w http.ResponseWriter, r *http.Request) {
w.Header().Set("Content-Type", "text/event-stream")
_, _ = io.WriteString(w, "data: {\"choices\":[{\"delta\":{\"content\":\"ok\"},\"finish_reason\":\"stop\"}]}\n\ndata: [DONE]\n")
})
c := &Client{BaseURL: srv, APIKey: "k", HTTPClient: srvClient()}
var end Event
for ev := range c.Stream(context.Background(), "m", []Message{{Role: RoleUser, Content: "hi"}}, []ToolDecl{{Name: "read_file", Parameters: []byte(`{"type":"object"}`)}}, StreamOptions{}) {
if ev.Kind == KindMessageEnd {
end = ev
}
}
if end.Message.Content != "ok" || end.Message.StopReason != StopEnd {
t.Fatalf("end %#v", end.Message)
}
}
func TestClientCancel(t *testing.T) {
started := make(chan struct{})
srv := httpServer(t, func(w http.ResponseWriter, r *http.Request) {
close(started)
w.Header().Set("Content-Type", "text/event-stream")
fl, _ := w.(http.Flusher)
_, _ = io.WriteString(w, "data: {\"choices\":[{\"delta\":{\"content\":\"a\"}}]}\n\n")
if fl != nil {
fl.Flush()
}
<-r.Context().Done()
})
c := &Client{BaseURL: srv, APIKey: "k", HTTPClient: &http.Client{}}
ctx, cancel := context.WithCancel(context.Background())
ch := c.Stream(ctx, "m", []Message{{Role: RoleUser, Content: "hi"}}, nil, StreamOptions{})
<-started
cancel()
saw := false
deadline := time.After(2 * time.Second)
for {
select {
case ev, ok := <-ch:
if !ok {
if !saw {
t.Fatal("channel closed without abort")
}
return
}
if ev.Kind == KindError && ev.Message.StopReason == StopAborted {
saw = true
}
case <-deadline:
t.Fatal("timed out waiting for cancel")
}
}
}
func httpServer(t *testing.T, h http.HandlerFunc) string {
t.Helper()
srv := httptest.NewServer(h)
t.Cleanup(srv.Close)
return srv.URL
}
func srvClient() *http.Client {
return &http.Client{Timeout: 2 * time.Second}
}
func TestStreamErrorString(t *testing.T) {
if (&StreamError{Message: "boom"}).Error() != "boom" {
t.Fatal("error string")
}
}
func TestParseSSEReasoningDelta(t *testing.T) {
body := strings.Join([]string{
`data: {"choices":[{"delta":{"reasoning_content":"先分析需求"},"index":0}]}`,
`data: {"choices":[{"delta":{"reasoning_content":"再检索标准"},"index":0}]}`,
`data: {"choices":[{"delta":{"content":"结论"},"finish_reason":"stop"}],"usage":{"prompt_tokens":1,"completion_tokens":2,"total_tokens":3}}`,
`data: [DONE]`,
"",
}, "\n")
var reasoning []string
var text []string
var end Event
for ev := range ParseSSE(context.Background(), strings.NewReader(body)) {
switch ev.Kind {
case KindReasoningDelta:
reasoning = append(reasoning, ev.ReasoningDelta)
case KindTextDelta:
text = append(text, ev.Text)
case KindMessageEnd:
end = ev
}
}
if strings.Join(reasoning, "") != "先分析需求再检索标准" {
t.Fatalf("reasoning deltas %v", reasoning)
}
// 正文与思考链分流:Content 只含 text,不含 reasoning
if strings.Join(text, "") != "结论" {
t.Fatalf("text deltas %v", text)
}
if end.Message.Content != "结论" {
t.Fatalf("end content %q must not include reasoning", end.Message.Content)
}
}