Stream brokered OpenCode chat replies over /connect

Add the server-side half of the brokered chat path (route B): the
opencode.Chat adapter forwards chat_prompt to POST
/api/session/{id}/prompt through the owned Connection.Call translation
and derives chat_event frames from the authenticated /api/event SSE
firehose (text and reasoning deltas, tool parts with their status
transitions, step and usage). Every pushed frame carries a stable
session and message id, with a last-message-id fallback so step/tool
events that omit the assistant message id stay attributable.

The /connect handler keeps per-session retained state (chatHub) capped at
1 MiB with least-recently-updated eviction at 128 turns, so a
reconnecting client re-syncs through chat_messages instead of replaying
deltas the OpenCode per-session log never resends. Prompt dispatch
subscribes before posting and buffers pre-ack frames to preserve order,
and cancels the stream on disconnect so no goroutine leaks. chat_sessions
and chat_create are declared but not yet brokered: they answer with an
explicit protocol error rather than the generic Pong.

Reuse the existing transport, validation and normalization; no second
OpenCode transport or credential path is introduced.

Tests: protocol golden round-trip, adapter normalization + v1/v2 Call
translation, and handler ack/stream/validation/resume/disconnect, plus a
real-adapter no-tool evidence harness.
This commit is contained in:
2026-10-16 04:09:14 +01:00
parent 9fa09e6ce8
commit b8b545accd
9 changed files with 1786 additions and 10 deletions
+458
View File
@@ -0,0 +1,458 @@
package server
import (
"context"
"log"
"strings"
"sync"
"time"
"git.a-wan.net/alan/pipelines/internal/opencode"
"git.a-wan.net/alan/pipelines/internal/protocol"
)
// chatBroker is the server-side OpenCode chat seam. *opencode.Chat implements
// it; tests substitute a scripted broker. Keeping it an interface means the
// /connect handler never builds a second OpenCode transport.
type chatBroker interface {
Prompt(ctx context.Context, sessionID, text string) error
// Stream subscribes to the session's OpenCode event stream. ready is called
// once the subscription is attached, before any frame is read, so the caller
// can order the prompt after the subscription.
Stream(ctx context.Context, sessionID string, ready func(), emit func(protocol.ChatEventFrame) error) error
Interrupt(ctx context.Context, sessionID string) error
Messages(ctx context.Context, sessionID string, limit int) ([]protocol.ChatMessage, error)
Reply(ctx context.Context, request protocol.ChatReplyRequest) error
}
// chatRetainedLimit bounds the state one session may retain so a stalled reader
// or a runaway turn cannot grow memory without limit. It matches the OpenCode
// response bound (1 MiB).
const chatRetainedLimit = opencode.MaxResponse
// chatMaxTurns bounds how many sessions retain turn state. A long-lived server
// cannot accumulate state for every session ever prompted, so the
// least-recently-updated turn is evicted once the cap is exceeded.
const chatMaxTurns = 128
// chatHub retains the accumulated turn state per session, independently of any
// single connection, so a reconnecting client can re-sync what it missed. State
// is keyed by session and then by message id (a session holds many turns), and
// is capped at chatRetainedLimit.
type chatHub struct {
mu sync.Mutex
turns map[string]*chatTurn
now func() time.Time
}
// chatPart is the retained form of one part. Text and reasoning deltas append to
// text (amortized O(1)); a tool frame replaces part outright.
type chatPart struct {
part protocol.ChatPart
text []byte
}
// chatMessage is one retained assistant turn. Its parts keep the exact ids the
// supporting chat_event frames carried, so a resume preserves attribution.
type chatMessage struct {
id string
order []string
parts map[string]*chatPart
}
type chatTurn struct {
sessionID string
order []string
messages map[string]*chatMessage
step *protocol.ChatStep
usage *protocol.ChatUsage
size int
overflow bool
updated time.Time
}
func newChatHub() *chatHub {
return &chatHub{turns: map[string]*chatTurn{}, now: time.Now}
}
func (h *chatHub) clock() time.Time {
if h.now != nil {
return h.now()
}
return time.Now()
}
// apply folds one pushed frame into the retained state. Text and reasoning
// deltas append to their message's part; a tool frame upserts the full part
// (accounting for its whole size) so the latest lifecycle status, input,
// content, metadata, error and time are retained under the frame's message id.
// Step and usage frames update the session's latest progress.
func (h *chatHub) apply(frame protocol.ChatEventFrame) {
if frame.SessionID == "" {
return
}
h.mu.Lock()
defer h.mu.Unlock()
turn := h.turns[frame.SessionID]
if turn == nil {
h.evictLocked()
turn = &chatTurn{sessionID: frame.SessionID, messages: map[string]*chatMessage{}}
h.turns[frame.SessionID] = turn
}
turn.updated = h.clock()
switch frame.Kind {
case protocol.ChatEventStep:
turn.step = frame.Step
return
case protocol.ChatEventUsage:
turn.usage = frame.Usage
return
}
if frame.MessageID == "" {
return
}
message := turn.messages[frame.MessageID]
if message == nil {
message = &chatMessage{id: frame.MessageID, parts: map[string]*chatPart{}}
turn.order = append(turn.order, frame.MessageID)
turn.messages[frame.MessageID] = message
}
switch frame.Kind {
case protocol.ChatEventText, protocol.ChatEventReasoning:
id := ""
if frame.Part != nil {
id = frame.Part.ID
}
if id == "" || frame.Delta == "" {
return
}
if !turn.retain(len(frame.Delta)) {
return
}
part, ok := message.parts[id]
if !ok {
part = &chatPart{part: protocol.ChatPart{Type: frame.Kind, ID: id}}
message.order = append(message.order, id)
message.parts[id] = part
}
part.text = append(part.text, frame.Delta...)
case protocol.ChatEventTool:
if frame.Part == nil || frame.Part.ID == "" || turn.overflow {
return
}
// The tool part is retained whole, so size the complete part (input,
// content, metadata, error and time) and charge the difference against
// the cap on every upsert, not just the first-seen name.
incoming := chatPartBytes(frame.Part)
part, ok := message.parts[frame.Part.ID]
if !ok {
if !turn.retain(incoming) {
return
}
part = &chatPart{}
message.order = append(message.order, frame.Part.ID)
message.parts[frame.Part.ID] = part
} else if previous := chatPartBytes(&part.part); incoming > previous {
if !turn.retain(incoming - previous) {
return
}
} else {
turn.shrink(previous - incoming)
}
part.part = *frame.Part
}
}
// chatPartBytes estimates the retained size of a part so the per-session cap
// accounts for its whole state instead of only the first-seen name. It walks the
// fields rather than re-marshalling, so the cost matches the bytes it stores.
func chatPartBytes(part *protocol.ChatPart) int {
if part == nil {
return 0
}
n := len(part.Type) + len(part.ID) + len(part.Name)
if part.State == nil {
return n
}
n += len(part.State.Status) + len(part.State.Input)
for _, content := range part.State.Content {
n += len(content.Type) + len(content.Text) + len(content.URI) + len(content.Mime) + len(content.Name)
}
if part.State.Error != nil {
n += len(part.State.Error.Type) + len(part.State.Error.Message)
}
for key, value := range part.State.Metadata {
n += len(key)
if text, ok := value.(string); ok {
n += len(text)
} else {
n += 8
}
}
return n
}
// evictLocked drops the least-recently-updated turn when the session cap is
// reached, keeping retained memory bounded across session churn.
func (h *chatHub) evictLocked() {
if len(h.turns) < chatMaxTurns {
return
}
oldestID := ""
var oldest time.Time
for id, turn := range h.turns {
if oldestID == "" || turn.updated.Before(oldest) {
oldestID, oldest = id, turn.updated
}
}
if oldestID != "" {
delete(h.turns, oldestID)
}
}
// retain records n more bytes and reports whether the turn is still within the
// bound. The first overflow records one diagnostic and then drops further state
// instead of growing without limit.
func (t *chatTurn) retain(n int) bool {
if t.overflow {
return false
}
if t.size+n > chatRetainedLimit {
t.overflow = true
log.Printf("chat session %s retained state exceeds %d bytes; dropping further state", t.sessionID, chatRetainedLimit)
return false
}
t.size += n
return true
}
// shrink releases n bytes when a tool upsert replaces a larger retained part.
func (t *chatTurn) shrink(n int) {
t.size -= n
if t.size < 0 {
t.size = 0
}
}
// snapshot returns the accumulated session as one assistant message per
// retained message id, with the same stable part ids, or nil when nothing is
// retained. It is the resume source: a reconnecting client re-syncs the whole
// session once instead of replaying deltas, and each turn keeps its own id.
func (h *chatHub) snapshot(sessionID string) []protocol.ChatMessage {
h.mu.Lock()
defer h.mu.Unlock()
turn := h.turns[sessionID]
if turn == nil {
return nil
}
turn.updated = h.clock()
var messages []protocol.ChatMessage
for _, id := range turn.order {
message := turn.messages[id]
if message == nil || len(message.order) == 0 {
continue
}
out := protocol.ChatMessage{ID: id, SessionID: sessionID, Role: protocol.ChatRoleAssistant}
for _, partID := range message.order {
if part, ok := message.parts[partID]; ok {
piece := part.part
piece.Text = string(part.text)
out.Parts = append(out.Parts, piece)
}
}
messages = append(messages, out)
}
return messages
}
// handleChat dispatches one chat_* request. It returns true when the request was
// a chat type and has been fully answered (the caller must not fall through to
// the generic response). Frames are pushed after the ack, each through the
// connection-scoped write function so the read loop and the stream goroutine
// never interleave partial websocket writes.
func handleChat(ctx context.Context, request protocol.Message, broker chatBroker, hub *chatHub, write func(protocol.Message) error) bool {
switch request.Type {
case protocol.ChatPrompt:
handleChatPrompt(ctx, request, broker, hub, write)
return true
case protocol.ChatMessages:
handleChatMessages(ctx, request, broker, hub, write)
return true
case protocol.ChatInterrupt:
handleChatSimple(ctx, request, broker, hub, write)
return true
case protocol.ChatReply:
handleChatSimple(ctx, request, broker, hub, write)
return true
case protocol.ChatSessions, protocol.ChatCreate:
// Declared in the protocol for the client but not yet brokered. Answer
// explicitly so the client never mistakes the generic Pong for a chat
// reply, and the boundary stays visible instead of silently dropping.
_ = write(protocol.Message{Type: protocol.Error, ID: request.ID, Error: "chat_sessions and chat_create are not yet brokered"})
return true
}
return false
}
func handleChatPrompt(ctx context.Context, request protocol.Message, broker chatBroker, hub *chatHub, write func(protocol.Message) error) {
if broker == nil {
_ = write(protocol.Message{Type: protocol.Error, ID: request.ID, Error: "chat is not configured on this server"})
return
}
prompt := request.ChatPrompt
if prompt == nil {
_ = write(protocol.Message{Type: protocol.Error, ID: request.ID, Error: "a chat prompt payload is required"})
return
}
if !opencode.ValidSessionID(prompt.SessionID) {
_ = write(protocol.Message{Type: protocol.Error, ID: request.ID, Error: "a valid OpenCode session id is required"})
return
}
if strings.TrimSpace(prompt.Text) == "" {
_ = write(protocol.Message{Type: protocol.Error, ID: request.ID, Error: "a nonempty chat prompt is required"})
return
}
// Reject an oversized prompt before any OpenCode call; the bound matches the
// OpenCode request limit.
if len(prompt.Text) > opencode.MaxRequest {
_ = write(protocol.Message{Type: protocol.Error, ID: request.ID, Error: "chat prompt exceeds 64 KiB"})
return
}
// Subscribe to the OpenCode firehose before posting the prompt: the stream
// has no replay cursor (research 2.4), so a delta emitted between the POST
// and the subscription would otherwise be lost. Frames that arrive before the
// ack is written are buffered and flushed in order after it.
sessionID := prompt.SessionID
streamCtx, streamCancel := context.WithCancel(ctx)
attached := make(chan struct{})
var attachOnce sync.Once
streamResult := make(chan error, 1)
var bufferMu sync.Mutex
buffering := true
var buffered []protocol.ChatEventFrame
go func() {
streamResult <- broker.Stream(streamCtx, sessionID, func() {
attachOnce.Do(func() { close(attached) })
}, func(frame protocol.ChatEventFrame) error {
bufferMu.Lock()
if buffering {
buffered = append(buffered, frame)
bufferMu.Unlock()
return nil
}
bufferMu.Unlock()
hub.apply(frame)
return write(protocol.Message{Type: protocol.ChatEvent, ChatEvent: &frame})
})
}()
select {
case <-attached:
case err := <-streamResult:
streamCancel()
if err != nil && ctx.Err() == nil {
_ = write(protocol.Message{Type: protocol.Error, ID: request.ID, Error: "chat stream: " + err.Error()})
}
return
case <-ctx.Done():
streamCancel()
return
}
if err := broker.Prompt(ctx, sessionID, prompt.Text); err != nil {
streamCancel()
_ = write(protocol.Message{Type: protocol.Error, ID: request.ID, Error: err.Error()})
return
}
// Acknowledge the accepted prompt with the echoed request id, then stream the
// reply. The prompt is sent exactly once and never re-sent on resume.
if err := write(protocol.Message{Type: protocol.ChatPrompt, ID: request.ID}); err != nil {
streamCancel()
return
}
// Flush frames buffered while the prompt was accepted, then resume live
// delivery. Holding bufferMu preserves order: the stream goroutine either
// buffered before this flush or waits and writes after it.
bufferMu.Lock()
buffering = false
pending := buffered
buffered = nil
for _, frame := range pending {
hub.apply(frame)
if err := write(protocol.Message{Type: protocol.ChatEvent, ChatEvent: &frame}); err != nil {
bufferMu.Unlock()
streamCancel()
return
}
}
bufferMu.Unlock()
go drainChatStream(ctx, streamResult, streamCancel, sessionID, write)
}
// drainChatStream reports the terminal error of a chat subscription and releases
// its context once the stream has ended.
func drainChatStream(ctx context.Context, result <-chan error, cancel context.CancelFunc, sessionID string, write func(protocol.Message) error) {
defer cancel()
if err := <-result; err != nil && ctx.Err() == nil {
log.Printf("chat stream session=%s: %v", sessionID, err)
_ = write(protocol.Message{Type: protocol.Error, Error: "chat stream: " + err.Error()})
}
}
func handleChatMessages(ctx context.Context, request protocol.Message, broker chatBroker, hub *chatHub, write func(protocol.Message) error) {
query := request.ChatQuery
if query == nil || !opencode.ValidSessionID(query.SessionID) {
_ = write(protocol.Message{Type: protocol.Error, ID: request.ID, Error: "a valid OpenCode session id is required"})
return
}
response := protocol.Message{Type: protocol.ChatMessages, ID: request.ID}
// Resume from retained server state first: a reconnecting client gets the
// accumulated turns with the same stable ids without re-reading OpenCode.
if messages := hub.snapshot(query.SessionID); len(messages) > 0 {
response.ChatMessages = messages
_ = write(response)
return
}
if broker == nil {
_ = write(response)
return
}
messages, err := broker.Messages(ctx, query.SessionID, query.Limit)
if err != nil {
_ = write(protocol.Message{Type: protocol.Error, ID: request.ID, Error: err.Error()})
return
}
response.ChatMessages = messages
_ = write(response)
}
func handleChatSimple(ctx context.Context, request protocol.Message, broker chatBroker, hub *chatHub, write func(protocol.Message) error) {
if broker == nil {
_ = write(protocol.Message{Type: protocol.Error, ID: request.ID, Error: "chat is not configured on this server"})
return
}
if request.Type == protocol.ChatInterrupt {
query := request.ChatQuery
if query == nil || !opencode.ValidSessionID(query.SessionID) {
_ = write(protocol.Message{Type: protocol.Error, ID: request.ID, Error: "a valid OpenCode session id is required"})
return
}
if err := broker.Interrupt(ctx, query.SessionID); err != nil {
_ = write(protocol.Message{Type: protocol.Error, ID: request.ID, Error: err.Error()})
return
}
_ = write(protocol.Message{Type: protocol.ChatInterrupt, ID: request.ID})
return
}
reply := request.ChatReply
if reply == nil || !opencode.ValidSessionID(reply.SessionID) {
_ = write(protocol.Message{Type: protocol.Error, ID: request.ID, Error: "a valid OpenCode session id is required"})
return
}
if err := broker.Reply(ctx, *reply); err != nil {
_ = write(protocol.Message{Type: protocol.Error, ID: request.ID, Error: err.Error()})
return
}
_ = write(protocol.Message{Type: protocol.ChatReply, ID: request.ID})
}
+563
View File
@@ -0,0 +1,563 @@
package server
import (
"context"
"encoding/json"
"fmt"
"net/http"
"net/http/httptest"
"strings"
"sync"
"sync/atomic"
"testing"
"time"
"github.com/coder/websocket"
"github.com/coder/websocket/wsjson"
"git.a-wan.net/alan/pipelines/internal/opencode"
"git.a-wan.net/alan/pipelines/internal/protocol"
)
// fakeChat is a scripted chatBroker. It counts calls and either emits a fixed
// frame sequence or blocks until the caller's context is canceled.
type fakeChat struct {
mu sync.Mutex
promptCalls int
interrupts int
replies int
messages []protocol.ChatMessage
script []protocol.ChatEventFrame
started chan struct{}
done chan struct{}
startOnce sync.Once
block bool
subscribed bool
promptBeforeSubscribe bool
}
func (f *fakeChat) Prompt(context.Context, string, string) error {
f.mu.Lock()
defer f.mu.Unlock()
f.promptCalls++
if !f.subscribed {
f.promptBeforeSubscribe = true
}
return nil
}
func (f *fakeChat) Stream(ctx context.Context, _ string, ready func(), emit func(protocol.ChatEventFrame) error) error {
f.mu.Lock()
f.subscribed = true
f.mu.Unlock()
if ready != nil {
ready()
}
if f.started != nil {
f.startOnce.Do(func() { close(f.started) })
}
if f.block {
<-ctx.Done()
if f.done != nil {
close(f.done)
}
return ctx.Err()
}
for _, frame := range f.script {
if err := ctx.Err(); err != nil {
return err
}
if err := emit(frame); err != nil {
return err
}
}
return nil
}
func (f *fakeChat) Interrupt(context.Context, string) error {
f.mu.Lock()
defer f.mu.Unlock()
f.interrupts++
return nil
}
func (f *fakeChat) Messages(context.Context, string, int) ([]protocol.ChatMessage, error) {
return f.messages, nil
}
func (f *fakeChat) Reply(context.Context, protocol.ChatReplyRequest) error {
f.mu.Lock()
defer f.mu.Unlock()
f.replies++
return nil
}
func (f *fakeChat) counts() (int, int, int) {
f.mu.Lock()
defer f.mu.Unlock()
return f.promptCalls, f.interrupts, f.replies
}
func chatFrame(kind, delta string, part *protocol.ChatPart, step *protocol.ChatStep) protocol.ChatEventFrame {
return protocol.ChatEventFrame{SessionID: "ses_1", MessageID: "msg_1", Kind: kind, Delta: delta, Part: part, Step: step}
}
func scriptedChat() *fakeChat {
return &fakeChat{script: []protocol.ChatEventFrame{
chatFrame(protocol.ChatEventText, "hello ", &protocol.ChatPart{Type: protocol.ChatPartText, ID: "msg_1#0"}, nil),
chatFrame(protocol.ChatEventText, "world", &protocol.ChatPart{Type: protocol.ChatPartText, ID: "msg_1#0"}, nil),
chatFrame(protocol.ChatEventTool, "", &protocol.ChatPart{Type: protocol.ChatPartTool, ID: "call_1", Name: "shell", State: &protocol.ChatToolState{Status: protocol.ChatToolCompleted, Input: []byte(`{"command":"ls"}`)}}, nil),
chatFrame(protocol.ChatEventStep, "", nil, &protocol.ChatStep{Status: protocol.ChatStepEnded, Index: 1}),
}}
}
func dialChat(t *testing.T, broker chatBroker) (context.Context, context.CancelFunc, *websocket.Conn) {
t.Helper()
ctx, cancel := context.WithTimeout(context.Background(), 10*time.Second)
srv := httptest.NewServer(HandlerWithOptions(ctx, Options{Chat: broker}))
t.Cleanup(func() { srv.Close(); cancel() })
conn, _, err := websocket.Dial(ctx, "ws"+strings.TrimPrefix(srv.URL, "http")+"/connect", nil)
if err != nil {
t.Fatal(err)
}
t.Cleanup(func() { conn.CloseNow() })
return ctx, cancel, conn
}
func readChatTurn(t *testing.T, ctx context.Context, conn *websocket.Conn) []protocol.ChatEventFrame {
t.Helper()
var frames []protocol.ChatEventFrame
for {
var message protocol.Message
if err := wsjson.Read(ctx, conn, &message); err != nil {
t.Fatal(err)
}
if message.Type == protocol.Error {
t.Fatalf("unexpected error frame: %s", message.Error)
}
if message.ChatEvent == nil {
continue
}
frames = append(frames, *message.ChatEvent)
if message.ChatEvent.Kind == protocol.ChatEventStep && message.ChatEvent.Step != nil && message.ChatEvent.Step.Status == protocol.ChatStepEnded {
return frames
}
}
}
func TestChatPromptStreamsEvents(t *testing.T) {
broker := scriptedChat()
ctx, _, conn := dialChat(t, broker)
if err := wsjson.Write(ctx, conn, protocol.Message{Type: protocol.ChatPrompt, ID: "r1", ChatPrompt: &protocol.ChatPromptRequest{SessionID: "ses_1", Text: "hi"}}); err != nil {
t.Fatal(err)
}
var ack protocol.Message
if err := wsjson.Read(ctx, conn, &ack); err != nil {
t.Fatal(err)
}
if ack.Type != protocol.ChatPrompt || ack.ID != "r1" {
t.Fatalf("ack = %+v", ack)
}
frames := readChatTurn(t, ctx, conn)
if len(frames) != 4 {
t.Fatalf("got %d frames, want 4", len(frames))
}
sawText, sawStep := false, false
for _, frame := range frames {
if frame.SessionID != "ses_1" || frame.MessageID != "msg_1" {
t.Fatalf("frame missing stable ids: %+v", frame)
}
switch frame.Kind {
case protocol.ChatEventText:
sawText = true
case protocol.ChatEventStep:
sawStep = true
if frame.Step == nil || frame.Step.Status != protocol.ChatStepEnded {
t.Fatalf("step frame = %+v", frame)
}
}
}
if !sawText || !sawStep {
t.Fatalf("missing text or step frame: %+v", frames)
}
if prompts, _, _ := broker.counts(); prompts != 1 {
t.Fatalf("prompt forwarded %d times, want 1", prompts)
}
}
func TestChatPromptRejectsInvalidSession(t *testing.T) {
broker := scriptedChat()
ctx, _, conn := dialChat(t, broker)
for _, id := range []string{"", "not-a-session", "msg_1"} {
if err := wsjson.Write(ctx, conn, protocol.Message{Type: protocol.ChatPrompt, ID: "bad", ChatPrompt: &protocol.ChatPromptRequest{SessionID: id, Text: "hi"}}); err != nil {
t.Fatal(err)
}
var response protocol.Message
if err := wsjson.Read(ctx, conn, &response); err != nil {
t.Fatal(err)
}
if response.Type != protocol.Error || response.ID != "bad" {
t.Fatalf("session %q not rejected: %+v", id, response)
}
}
if prompts, _, _ := broker.counts(); prompts != 0 {
t.Fatalf("invalid sessions reached OpenCode: %d", prompts)
}
}
func TestChatPromptRejectsOversizedText(t *testing.T) {
broker := scriptedChat()
ctx, _, conn := dialChat(t, broker)
oversized := strings.Repeat("x", opencode.MaxRequest+1)
if err := wsjson.Write(ctx, conn, protocol.Message{Type: protocol.ChatPrompt, ID: "big", ChatPrompt: &protocol.ChatPromptRequest{SessionID: "ses_1", Text: oversized}}); err != nil {
t.Fatal(err)
}
var response protocol.Message
if err := wsjson.Read(ctx, conn, &response); err != nil {
t.Fatal(err)
}
if response.Type != protocol.Error || !strings.Contains(response.Error, "64 KiB") {
t.Fatalf("oversized prompt not rejected: %+v", response)
}
if prompts, _, _ := broker.counts(); prompts != 0 {
t.Fatalf("oversized prompt reached OpenCode: %d", prompts)
}
}
func TestChatPromptWithoutBroker(t *testing.T) {
ctx, _, conn := dialChat(t, nil)
if err := wsjson.Write(ctx, conn, protocol.Message{Type: protocol.ChatPrompt, ID: "r1", ChatPrompt: &protocol.ChatPromptRequest{SessionID: "ses_1", Text: "hi"}}); err != nil {
t.Fatal(err)
}
var response protocol.Message
if err := wsjson.Read(ctx, conn, &response); err != nil {
t.Fatal(err)
}
if response.Type != protocol.Error || !strings.Contains(response.Error, "not configured") {
t.Fatalf("unconfigured chat not rejected: %+v", response)
}
}
func TestChatMessagesResumesRetainedState(t *testing.T) {
broker := scriptedChat()
ctx, _, conn := dialChat(t, broker)
if err := wsjson.Write(ctx, conn, protocol.Message{Type: protocol.ChatPrompt, ID: "r1", ChatPrompt: &protocol.ChatPromptRequest{SessionID: "ses_1", Text: "hi"}}); err != nil {
t.Fatal(err)
}
var ack protocol.Message
if err := wsjson.Read(ctx, conn, &ack); err != nil {
t.Fatal(err)
}
readChatTurn(t, ctx, conn)
if err := wsjson.Write(ctx, conn, protocol.Message{Type: protocol.ChatMessages, ID: "r2", ChatQuery: &protocol.ChatQuery{SessionID: "ses_1"}}); err != nil {
t.Fatal(err)
}
var response protocol.Message
if err := wsjson.Read(ctx, conn, &response); err != nil {
t.Fatal(err)
}
if response.Type != protocol.ChatMessages || response.ID != "r2" || len(response.ChatMessages) != 1 {
t.Fatalf("resume response = %+v", response)
}
message := response.ChatMessages[0]
if message.ID != "msg_1" || message.Role != protocol.ChatRoleAssistant || len(message.Parts) != 2 {
t.Fatalf("resumed message = %+v", message)
}
if message.Parts[0].Type != protocol.ChatPartText || message.Parts[0].ID != "msg_1#0" || message.Parts[0].Text != "hello world" {
t.Fatalf("resumed text part = %+v", message.Parts[0])
}
if message.Parts[1].Type != protocol.ChatPartTool || message.Parts[1].State == nil || message.Parts[1].State.Status != protocol.ChatToolCompleted {
t.Fatalf("resumed tool part = %+v", message.Parts[1])
}
}
func TestChatDisconnectCancelsStream(t *testing.T) {
broker := scriptedChat()
broker.block = true
broker.started = make(chan struct{})
broker.done = make(chan struct{})
ctx, _, conn := dialChat(t, broker)
if err := wsjson.Write(ctx, conn, protocol.Message{Type: protocol.ChatPrompt, ID: "r1", ChatPrompt: &protocol.ChatPromptRequest{SessionID: "ses_1", Text: "hi"}}); err != nil {
t.Fatal(err)
}
var ack protocol.Message
if err := wsjson.Read(ctx, conn, &ack); err != nil {
t.Fatal(err)
}
select {
case <-broker.started:
case <-time.After(2 * time.Second):
t.Fatal("stream never started")
}
conn.CloseNow()
// The server's read loop returns on disconnect, canceling the connection
// context the subscription observes; no goroutine may keep running.
select {
case <-broker.done:
case <-time.After(3 * time.Second):
t.Fatal("chat stream was not canceled on disconnect")
}
if prompts, _, _ := broker.counts(); prompts != 1 {
t.Fatalf("prompt forwarded %d times, want 1", prompts)
}
}
func TestChatInterruptAndReply(t *testing.T) {
broker := scriptedChat()
ctx, _, conn := dialChat(t, broker)
if err := wsjson.Write(ctx, conn, protocol.Message{Type: protocol.ChatInterrupt, ID: "i1", ChatQuery: &protocol.ChatQuery{SessionID: "ses_1"}}); err != nil {
t.Fatal(err)
}
var response protocol.Message
if err := wsjson.Read(ctx, conn, &response); err != nil {
t.Fatal(err)
}
if response.Type != protocol.ChatInterrupt || response.ID != "i1" {
t.Fatalf("interrupt response = %+v", response)
}
if err := wsjson.Write(ctx, conn, protocol.Message{Type: protocol.ChatReply, ID: "p1", ChatReply: &protocol.ChatReplyRequest{SessionID: "ses_1", RequestID: "req_1", Decision: protocol.ChatDecisionOnce}}); err != nil {
t.Fatal(err)
}
if err := wsjson.Read(ctx, conn, &response); err != nil {
t.Fatal(err)
}
if response.Type != protocol.ChatReply || response.ID != "p1" {
t.Fatalf("reply response = %+v", response)
}
if _, interrupts, replies := broker.counts(); interrupts != 1 || replies != 1 {
t.Fatalf("interrupts=%d replies=%d, want 1/1", interrupts, replies)
}
}
func TestChatHubAccumulatesAndEvicts(t *testing.T) {
tick := 0
hub := newChatHub()
hub.now = func() time.Time { tick++; return time.Unix(int64(tick), 0) }
for _, delta := range []string{"a", "b", "c"} {
hub.apply(protocol.ChatEventFrame{SessionID: "ses_1", MessageID: "msg_1", Kind: protocol.ChatEventText, Delta: delta, Part: &protocol.ChatPart{Type: protocol.ChatPartText, ID: "msg_1#0"}})
}
messages := hub.snapshot("ses_1")
if len(messages) != 1 || len(messages[0].Parts) != 1 || messages[0].Parts[0].Text != "abc" {
t.Fatalf("snapshot = %+v", messages)
}
for i := 0; i < chatMaxTurns+5; i++ {
id := fmt.Sprintf("ses_%d", i)
hub.apply(protocol.ChatEventFrame{SessionID: id, MessageID: "m", Kind: protocol.ChatEventText, Delta: "x", Part: &protocol.ChatPart{Type: protocol.ChatPartText, ID: id + "#0"}})
}
if len(hub.turns) != chatMaxTurns {
t.Fatalf("retained %d turns, want %d", len(hub.turns), chatMaxTurns)
}
}
// TestChatHubKeepsMessagesSeparate pins per-message retention: a second prompt
// in one session must not fold its parts onto the first turn's message id.
func TestChatHubKeepsMessagesSeparate(t *testing.T) {
hub := newChatHub()
hub.apply(protocol.ChatEventFrame{SessionID: "ses_1", MessageID: "msg_1", Kind: protocol.ChatEventText, Delta: "one", Part: &protocol.ChatPart{Type: protocol.ChatPartText, ID: "msg_1#0"}})
hub.apply(protocol.ChatEventFrame{SessionID: "ses_1", MessageID: "msg_2", Kind: protocol.ChatEventText, Delta: "two", Part: &protocol.ChatPart{Type: protocol.ChatPartText, ID: "msg_2#0"}})
messages := hub.snapshot("ses_1")
if len(messages) != 2 {
t.Fatalf("messages = %+v", messages)
}
if messages[0].ID != "msg_1" || messages[0].Parts[0].ID != "msg_1#0" || messages[0].Parts[0].Text != "one" {
t.Fatalf("first message = %+v", messages[0])
}
if messages[1].ID != "msg_2" || messages[1].Parts[0].ID != "msg_2#0" || messages[1].Parts[0].Text != "two" {
t.Fatalf("second message = %+v", messages[1])
}
}
// TestChatHubChargesToolStateBytes pins the retained-state cap: a tool part's
// input/content bytes count against it, and an upsert that would exceed the cap
// is dropped with the overflow diagnostic rather than growing memory.
func TestChatHubChargesToolStateBytes(t *testing.T) {
hub := newChatHub()
big := strings.Repeat("x", chatRetainedLimit/2)
hub.apply(protocol.ChatEventFrame{SessionID: "ses_1", MessageID: "msg_1", Kind: protocol.ChatEventTool, Part: &protocol.ChatPart{Type: protocol.ChatPartTool, ID: "call_1", Name: "shell", State: &protocol.ChatToolState{Status: protocol.ChatToolStreaming, Input: json.RawMessage(`"` + big + `"`)}}})
hub.mu.Lock()
size := hub.turns["ses_1"].size
hub.mu.Unlock()
if size < len(big) {
t.Fatalf("tool input not charged: size=%d want >= %d", size, len(big))
}
huge := strings.Repeat("y", chatRetainedLimit)
hub.apply(protocol.ChatEventFrame{SessionID: "ses_1", MessageID: "msg_1", Kind: protocol.ChatEventTool, Part: &protocol.ChatPart{Type: protocol.ChatPartTool, ID: "call_1", Name: "shell", State: &protocol.ChatToolState{Status: protocol.ChatToolCompleted, Input: json.RawMessage(`"` + huge + `"`)}}})
messages := hub.snapshot("ses_1")
if len(messages) != 1 || len(messages[0].Parts) != 1 || messages[0].Parts[0].State == nil || messages[0].Parts[0].State.Status != protocol.ChatToolStreaming {
t.Fatalf("overflowing tool state was retained: %+v", messages)
}
}
// TestChatPromptSubscribesBeforePrompt pins the subscribe-before-prompt order:
// the firehose has no replay cursor, so the subscription must be attached before
// the prompt is posted or deltas emitted in between are lost.
func TestChatPromptSubscribesBeforePrompt(t *testing.T) {
broker := scriptedChat()
ctx, _, conn := dialChat(t, broker)
if err := wsjson.Write(ctx, conn, protocol.Message{Type: protocol.ChatPrompt, ID: "r1", ChatPrompt: &protocol.ChatPromptRequest{SessionID: "ses_1", Text: "hi"}}); err != nil {
t.Fatal(err)
}
var ack protocol.Message
if err := wsjson.Read(ctx, conn, &ack); err != nil {
t.Fatal(err)
}
readChatTurn(t, ctx, conn)
broker.mu.Lock()
before := broker.promptBeforeSubscribe
broker.mu.Unlock()
if before {
t.Fatal("prompt was posted before the event subscription was attached")
}
}
// realAdapterChat posts a chat_prompt through the real opencode.Chat broker and
// HandlerWithOptions against a v2 HTTP fixture that replays frames, so the
// acceptance path forwards to an actual OpenCode-shaped stream (not a scripted
// fake) and returns the chat_event frames the server pushed. It pins exactly one
// prompt POST and one event subscription per chat_prompt, then asserts the ack
// echoes the request id; callers assert the frames themselves.
func realAdapterChat(t *testing.T, frames []string) []protocol.ChatEventFrame {
t.Helper()
var prompts, streams atomic.Int32
fixture := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
switch {
case r.Method == http.MethodPost && r.URL.Path == "/api/session/ses_1/prompt":
prompts.Add(1)
w.Header().Set("Content-Type", "application/json")
_, _ = w.Write([]byte(`{"data":{"id":"msg_1","sessionID":"ses_1","type":"user"}}`))
case r.Method == http.MethodGet && r.URL.Path == "/api/event":
streams.Add(1)
w.Header().Set("Content-Type", "text/event-stream")
for _, frame := range frames {
_, _ = w.Write([]byte("data: " + frame + "\n\n"))
}
default:
t.Errorf("unexpected request %s %s", r.Method, r.URL.Path)
}
}))
defer fixture.Close()
broker := opencode.NewChat(&opencode.Connection{Backend: "v2", URL: fixture.URL})
ctx, _, conn := dialChat(t, broker)
if err := wsjson.Write(ctx, conn, protocol.Message{Type: protocol.ChatPrompt, ID: "r1", ChatPrompt: &protocol.ChatPromptRequest{SessionID: "ses_1", Text: "ping"}}); err != nil {
t.Fatal(err)
}
var ack protocol.Message
if err := wsjson.Read(ctx, conn, &ack); err != nil {
t.Fatal(err)
}
if ack.Type != protocol.ChatPrompt || ack.ID != "r1" {
t.Fatalf("ack = %+v", ack)
}
framesIn := readChatTurn(t, ctx, conn)
if prompts.Load() != 1 || streams.Load() != 1 {
t.Fatalf("prompts=%d streams=%d, want 1/1", prompts.Load(), streams.Load())
}
return framesIn
}
// assertStableFrameIDs fails unless every frame carries the fixture's session
// and message ids: a client cannot attribute a frame without them.
func assertStableFrameIDs(t *testing.T, frames []protocol.ChatEventFrame) {
t.Helper()
for _, frame := range frames {
if frame.SessionID != "ses_1" || frame.MessageID != "msg_1" {
t.Fatalf("frame missing stable ids: %+v", frame)
}
}
}
// TestChatPromptThroughRealAdapter drives a text, tool and step turn end to end.
func TestChatPromptThroughRealAdapter(t *testing.T) {
framesIn := realAdapterChat(t, []string{
`{"id":"evt_1","type":"session.text.delta","data":{"sessionID":"ses_1","assistantMessageID":"msg_1","ordinal":0,"delta":"pong"}}`,
`{"id":"evt_2","type":"session.tool.called","data":{"sessionID":"ses_1","assistantMessageID":"msg_1","id":"call_1","name":"shell","input":{"command":"ls"}}}`,
`{"id":"evt_3","type":"session.step.ended","data":{"sessionID":"ses_1","assistantMessageID":"msg_1","step":1}}`,
})
assertStableFrameIDs(t, framesIn)
sawText, sawTool := false, false
for _, frame := range framesIn {
if frame.Kind == protocol.ChatEventText && frame.Delta == "pong" {
sawText = true
}
if frame.Kind == protocol.ChatEventTool && frame.Part != nil && frame.Part.Name == "shell" {
sawTool = true
}
}
if !sawText || !sawTool {
t.Fatalf("missing text or tool frame: %+v", framesIn)
}
}
// TestChatPromptNoToolThroughRealAdapter is the task's evidence harness: it posts
// a trivial no-tool prompt to the real opencode.Chat adapter over a v2 SSE
// fixture that emits only a text delta and a step, and asserts both arrive as
// chat_event frames with stable session/message ids. It reuses dialChat,
// readChatTurn and realAdapterChat so it pins the same ack/stream contract as the
// other chat tests, and it pins exactly one prompt POST and one event subscription.
func TestChatPromptNoToolThroughRealAdapter(t *testing.T) {
framesIn := realAdapterChat(t, []string{
`{"id":"evt_1","type":"session.text.delta","data":{"sessionID":"ses_1","assistantMessageID":"msg_1","ordinal":0,"delta":"pong"}}`,
`{"id":"evt_2","type":"session.step.ended","data":{"sessionID":"ses_1","assistantMessageID":"msg_1","step":1}}`,
})
assertStableFrameIDs(t, framesIn)
sawText, sawStep, sawTool := false, false, false
for _, frame := range framesIn {
switch frame.Kind {
case protocol.ChatEventText:
sawText = frame.Delta == "pong"
case protocol.ChatEventStep:
sawStep = frame.Step != nil && frame.Step.Status == protocol.ChatStepEnded
case protocol.ChatEventTool:
sawTool = true
}
}
if !sawText || !sawStep {
t.Fatalf("missing text or step frame: %+v", framesIn)
}
if sawTool {
t.Fatalf("a trivial no-tool prompt produced a tool frame: %+v", framesIn)
}
}
// TestChatPromptRejectsEmptyText pins the whitespace guard: an empty or
// whitespace-only prompt is rejected before any OpenCode call.
func TestChatPromptRejectsEmptyText(t *testing.T) {
broker := scriptedChat()
ctx, _, conn := dialChat(t, broker)
for _, text := range []string{"", " ", "\t\n"} {
if err := wsjson.Write(ctx, conn, protocol.Message{Type: protocol.ChatPrompt, ID: "empty", ChatPrompt: &protocol.ChatPromptRequest{SessionID: "ses_1", Text: text}}); err != nil {
t.Fatal(err)
}
var response protocol.Message
if err := wsjson.Read(ctx, conn, &response); err != nil {
t.Fatal(err)
}
if response.Type != protocol.Error || response.ID != "empty" {
t.Fatalf("empty text %q not rejected: %+v", text, response)
}
}
if prompts, _, _ := broker.counts(); prompts != 0 {
t.Fatalf("empty prompts reached OpenCode: %d", prompts)
}
}
// TestChatSessionsAndCreateAnswerExplicitly pins the declared-but-unbrokered
// boundary: chat_sessions and chat_create are answered with a protocol error
// rather than falling through to the generic Pong.
func TestChatSessionsAndCreateAnswerExplicitly(t *testing.T) {
broker := scriptedChat()
ctx, _, conn := dialChat(t, broker)
for _, typ := range []string{protocol.ChatSessions, protocol.ChatCreate} {
if err := wsjson.Write(ctx, conn, protocol.Message{Type: typ, ID: "c1"}); err != nil {
t.Fatal(err)
}
var response protocol.Message
if err := wsjson.Read(ctx, conn, &response); err != nil {
t.Fatal(err)
}
if response.Type != protocol.Error || response.ID != "c1" {
t.Fatalf("%s did not answer with an explicit error: %+v", typ, response)
}
}
}
+25 -8
View File
@@ -5,6 +5,7 @@ import (
"log"
"net/http"
"strings"
"sync"
"time"
"github.com/coder/websocket"
@@ -21,6 +22,10 @@ type Options struct {
// ArtifactsRoot overrides where stage artifacts live. Empty uses the store's
// directory (the database's parent), matching the data-directory default.
ArtifactsRoot string
// Chat brokers OpenCode chat traffic over /connect. Nil answers every
// chat_* request with a protocol error; a configured server passes an
// *opencode.Chat bound to the owned endpoint.
Chat chatBroker
}
// Handler serves the persistent connection with default options. Canceling ctx
@@ -34,6 +39,7 @@ func Handler(ctx context.Context, stores ...*pipeline.Store) http.Handler {
func HandlerWithOptions(ctx context.Context, options Options, stores ...*pipeline.Store) http.Handler {
build := buildinfo.Current()
started := time.Now()
hub := newChatHub()
mux := http.NewServeMux()
var runs *runner
var runErr error
@@ -70,16 +76,27 @@ func HandlerWithOptions(ctx context.Context, options Options, stores ...*pipelin
}
defer conn.CloseNow()
conn.SetReadLimit(2 << 20)
// connCtx is canceled when this connection's read loop returns, so a
// chat subscription started for the client stops on disconnect. A single
// write mutex serializes the read loop's responses and the chat stream
// goroutine, which must never interleave partial websocket writes.
connCtx, connCancel := context.WithCancel(ctx)
defer connCancel()
var writeMu sync.Mutex
write := func(message protocol.Message) error {
writeMu.Lock()
defer writeMu.Unlock()
writeCtx, cancel := context.WithTimeout(connCtx, 5*time.Second)
defer cancel()
return wsjson.Write(writeCtx, conn, message)
}
if len(stores) > 0 {
items, err := stores[0].List(ctx)
if err != nil {
log.Printf("list pipelines: %v", err)
return
}
writeCtx, cancel := context.WithTimeout(ctx, 5*time.Second)
err = wsjson.Write(writeCtx, conn, protocol.Message{Type: protocol.PipelineList, Pipelines: items, NotificationEndpoints: notificationEndpoints()})
cancel()
if err != nil {
if err := write(protocol.Message{Type: protocol.PipelineList, Pipelines: items, NotificationEndpoints: notificationEndpoints()}); err != nil {
return
}
}
@@ -89,6 +106,9 @@ func HandlerWithOptions(ctx context.Context, options Options, stores ...*pipelin
return
}
log.Printf("command received: client=%s type=%q id=%q pipeline=%d version=%d", r.RemoteAddr, request.Type, request.ID, request.PipelineID, request.Version)
if handleChat(connCtx, request, options.Chat, hub, write) {
continue
}
response := protocol.Message{Type: protocol.Pong, ID: request.ID}
if request.Type == protocol.List || request.Type == protocol.Submit {
response.NotificationEndpoints = notificationEndpoints()
@@ -196,10 +216,7 @@ func HandlerWithOptions(ctx context.Context, options Options, stores ...*pipelin
response.Error = "execution unavailable: " + runErr.Error()
}
}
writeCtx, cancel := context.WithTimeout(ctx, 5*time.Second)
err := wsjson.Write(writeCtx, conn, response)
cancel()
if err != nil {
if err := write(response); err != nil {
log.Printf("write response: %v", err)
return
}