fix(sync): remove stale-delta skip and add parts-gap recovery (#889)

The pipeline's stale-delta mechanism incorrectly marked all
message.part.delta events as stale when a message.part.updated
coalesced, regardless of queue position. This caused valid streaming
deltas to be silently dropped, resulting in blank or incomplete
assistant messages.

Additionally, when part events were dropped by the reducer (missing
parts array or partID not found), there was no recovery path — the
state stayed permanently out of sync until the next SSE reconnect
or manual refresh.

Also discovered: message.updated that successfully writes an assistant
message but has empty parts would render a blank bubble, with no
repair triggered since repair only ran on reducer return false.

Changes:
- Remove staleDeltas Set and deltaKey from event-pipeline.ts
- Coalesce still replaces same-key events, but deltas are never skipped
- Add enqueuePartsRepair + repairSessionParts to sync-context.tsx
  (5s cooldown, deduped, async SDK re-fetch)
- Trigger repair on reducer return false for part events
- Trigger repair on message.updated return true with empty parts
- Add sync debug.ts with gated diagnostic logging
- Add pipeline coalescing tests
This commit is contained in:
jwcrystal
2026-04-11 23:40:47 +03:00
committed by GitHub
parent 1c7a0b8ddd
commit 964c209ff6
5 changed files with 351 additions and 18 deletions
@@ -35,6 +35,22 @@ function createSdkWithSingleEvent(event, hold) {
};
}
// Helper to create an SDK that yields multiple events in sequence, then holds.
function createSdkWithEvents(events, hold) {
return {
global: {
event: async () => ({
stream: (async function* () {
for (const event of events) {
yield event;
}
await hold;
})(),
}),
},
};
}
describe('createEventPipeline', () => {
it('falls back to payload.properties.directory when the SDK event omits top-level directory', async () => {
installDomStubs();
@@ -192,4 +208,136 @@ describe('createEventPipeline', () => {
expect(received[0].directory).toBe('global');
expect(received[0].payload.type).toBe('server.connected');
});
it('delivers message.part.delta events after a coalesced message.part.updated (no stale-delta skip)', async () => {
installDomStubs();
let releaseStream;
const hold = new Promise((resolve) => {
releaseStream = resolve;
});
const received = [];
// Simulate: part.updated arrives first, then delta, then part.updated again (coalesces with first).
// After coalescing, the delta should still be delivered — NOT skipped.
const directory = '/test/dir';
const sdk = createSdkWithEvents([
// T0: message.part.updated for part-A
{
payload: {
type: 'message.part.updated',
properties: {
directory,
part: { id: 'part-A', type: 'text', messageID: 'msg-1' },
},
},
},
// T1: message.part.delta for part-A (should flow through even after coalesce)
{
payload: {
type: 'message.part.delta',
properties: {
directory,
messageID: 'msg-1',
partID: 'part-A',
field: 'text',
delta: ' world',
},
},
},
// T2: message.part.updated for part-A — coalesces with T0
{
payload: {
type: 'message.part.updated',
properties: {
directory,
part: { id: 'part-A', type: 'text', messageID: 'msg-1' },
},
},
},
], hold);
const delivered = new Promise((resolve) => {
const { cleanup } = createEventPipeline({
sdk,
onEvent: (dir, payload) => {
received.push({ directory: dir, payload });
if (received.length === 2) {
cleanup();
releaseStream();
resolve();
}
},
});
});
await delivered;
// Coalescing means T0 and T2 merge into one event at T0's queue position.
// The delta is a different event type with no coalesce key, so it gets
// its own queue slot. After coalesce:
// - queue[0] = coalesced part.updated (from T2, replacing T0)
// - queue[1] = part.delta (from T1)
// Total: 2 events delivered
expect(received.length).toBe(2);
// The first event should be the coalesced message.part.updated
expect(received[0].payload.type).toBe('message.part.updated');
// The delta MUST be delivered — it should NOT be skipped
expect(received[1].payload.type).toBe('message.part.delta');
expect(received[1].payload.properties.delta).toBe(' world');
});
it('coalesces message.part.updated events for the same part', async () => {
installDomStubs();
let releaseStream;
const hold = new Promise((resolve) => {
releaseStream = resolve;
});
const received = [];
const directory = '/test/dir';
const sdk = createSdkWithEvents([
{
payload: {
type: 'message.part.updated',
properties: {
directory,
part: { id: 'part-A', type: 'text', messageID: 'msg-1' },
},
},
},
{
payload: {
type: 'message.part.updated',
properties: {
directory,
part: { id: 'part-A', type: 'text', messageID: 'msg-1' },
},
},
},
], hold);
const delivered = new Promise((resolve) => {
const { cleanup } = createEventPipeline({
sdk,
onEvent: (dir, payload) => {
received.push({ directory: dir, payload });
cleanup();
releaseStream();
resolve();
},
});
});
await delivered;
// Only 1 event should be delivered (coalesced)
expect(received.length).toBe(1);
expect(received[0].payload.type).toBe('message.part.updated');
});
});
+83
View File
@@ -0,0 +1,83 @@
/**
* Sync debug logging — gated behind localStorage flag.
*
* Enable in browser console:
* localStorage.setItem("openchamber:sync:debug", "1")
*
* Disable:
* localStorage.removeItem("openchamber:sync:debug")
*
* All checks are early-returns on the hot path — zero cost when disabled.
*/
const FLAG_KEY = "openchamber:sync:debug"
let _enabled: boolean | undefined
export function isSyncDebugEnabled(): boolean {
if (_enabled !== undefined) return _enabled
try {
_enabled = typeof localStorage !== "undefined" && localStorage.getItem(FLAG_KEY) === "1"
} catch {
_enabled = false
}
return _enabled
}
/** Force-refresh the flag (call after user toggles localStorage). */
export function refreshSyncDebugFlag(): void {
_enabled = undefined
}
type SyncDebugCategory = "pipeline" | "reducer" | "dispatch"
function log(cat: SyncDebugCategory, ...args: unknown[]): void {
if (!isSyncDebugEnabled()) return
const tag = `%c[sync:${cat}]`
const style = "color: #888"
console.log(tag, style, ...args)
}
export const syncDebug = {
pipeline: {
/** Event coalesced (replaced an earlier event in the queue). */
coalesced: (eventType: string, coalesceKey: string) =>
log("pipeline", "coalesced", eventType, coalesceKey),
/** Flush batch dispatched. */
flush: (count: number) =>
log("pipeline", "flush", `${count} events`),
},
reducer: {
/** message.updated skipped because role/finish/completed matched existing. */
messageUpdatedUnchanged: (sessionID: string, messageID: string, role: string, finish: unknown, completed: unknown) =>
log("reducer", "message.updated UNCHANGED (skipped)", { sessionID, messageID, role, finish, completed }),
/** message.part.updated arrived but no parts array exists for this messageID. */
partUpdatedNoExistingParts: (messageID: string, partID: string, partType: string) =>
log("reducer", "message.part.updated NO EXISTING PARTS", { messageID, partID, partType }),
/** message.part.delta arrived but parts array missing — silently dropped. */
partDeltaNoParts: (messageID: string, partID: string) =>
log("reducer", "message.part.delta DROPPED (no parts array)", { messageID, partID }),
/** message.part.delta arrived but partID not found in parts array. */
partDeltaNotFound: (messageID: string, partID: string) =>
log("reducer", "message.part.delta DROPPED (partID not found)", { messageID, partID }),
/** SKIP_PARTS filtered out a part. */
partSkipped: (messageID: string, partID: string, partType: string) =>
log("reducer", "message.part.updated SKIPPED (type filtered)", { messageID, partID, partType }),
},
dispatch: {
/** Event dispatched to store but reducer returned false (no state change). */
eventNoChange: (eventType: string, sessionID?: string, messageID?: string) =>
log("dispatch", "event → no change", { eventType, sessionID, messageID }),
/** Event applied to store successfully. */
eventApplied: (eventType: string, sessionID?: string, messageID?: string) =>
log("dispatch", "event → applied", { eventType, sessionID, messageID }),
},
} as const
+4 -15
View File
@@ -9,6 +9,7 @@
*/
import type { Event, OpencodeClient } from "@opencode-ai/sdk/v2/client"
import { syncDebug } from "./debug"
// ---------------------------------------------------------------------------
// Types
@@ -86,13 +87,9 @@ export function createEventPipeline(input: EventPipelineInput) {
let queue: QueuedEvent[] = []
let buffer: QueuedEvent[] = []
const coalesced = new Map<string, number>()
const staleDeltas = new Set<string>()
let timer: ReturnType<typeof setTimeout> | undefined
let last = 0
const deltaKey = (directory: string, messageID: string, partID: string) =>
`${directory}:${messageID}:${partID}`
// Coalesce key — same-type events for the same entity replace earlier ones
const key = (directory: string, payload: Event): string | undefined => {
if (payload.type === "session.status") {
@@ -109,7 +106,7 @@ export function createEventPipeline(input: EventPipelineInput) {
return undefined
}
// Flush — swap queue, dispatch events, skip stale deltas
// Flush — swap queue, dispatch events
const flush = () => {
if (timer) clearTimeout(timer)
timer = undefined
@@ -117,21 +114,16 @@ export function createEventPipeline(input: EventPipelineInput) {
if (queue.length === 0) return
const events = queue
const skip = staleDeltas.size > 0 ? new Set(staleDeltas) : undefined
queue = buffer
buffer = events
queue.length = 0
coalesced.clear()
staleDeltas.clear()
last = Date.now()
syncDebug.pipeline.flush(events.length)
// React 18 batches synchronous setState calls automatically,
// equivalent to SolidJS batch()
for (const event of events) {
if (skip && event.payload.type === "message.part.delta") {
const props = event.payload.properties as { messageID: string; partID: string }
if (skip.has(deltaKey(event.directory, props.messageID, props.partID))) continue
}
onEvent(event.directory, event.payload)
}
@@ -214,10 +206,7 @@ export function createEventPipeline(input: EventPipelineInput) {
const i = coalesced.get(k)
if (i !== undefined) {
queue[i] = { directory, payload: normalizedPayload }
if (normalizedPayload.type === "message.part.updated") {
const part = (normalizedPayload.properties as { part: { messageID: string; id: string } }).part
staleDeltas.add(deltaKey(directory, part.messageID, part.id))
}
syncDebug.pipeline.coalesced(normalizedPayload.type, k)
continue
}
coalesced.set(k, queue.length)
+15 -3
View File
@@ -14,6 +14,7 @@ import { Binary } from "./binary"
import type { GlobalState, State } from "./types"
import { dropSessionCaches } from "./session-cache"
import { stripSessionDiffSnapshots } from "./sanitize"
import { syncDebug } from "./debug"
const SKIP_PARTS = new Set(["patch", "step-start", "step-finish"])
@@ -160,6 +161,7 @@ export function applyDirectoryEvent(
&& (existing as { finish?: unknown }).finish === (info as { finish?: unknown }).finish
&& (existing.time as { completed?: number })?.completed === (info.time as { completed?: number })?.completed
if (unchanged) {
syncDebug.reducer.messageUpdatedUnchanged(info.sessionID, info.id, info.role, (info as { finish?: unknown }).finish, (info.time as { completed?: number })?.completed)
return false
}
const next = [...messages]
@@ -190,10 +192,14 @@ export function applyDirectoryEvent(
case "message.part.updated": {
const part = (event.properties as { part: Part }).part
if (SKIP_PARTS.has(part.type)) return false
if (SKIP_PARTS.has(part.type)) {
syncDebug.reducer.partSkipped((part as { messageID: string }).messageID, part.id, part.type)
return false
}
const messageID = (part as { messageID: string }).messageID
const parts = draft.part[messageID]
if (!parts) {
syncDebug.reducer.partUpdatedNoExistingParts(messageID, part.id, part.type)
draft.part[messageID] = [part]
return true
}
@@ -246,9 +252,15 @@ export function applyDirectoryEvent(
delta: string
}
const parts = draft.part[props.messageID]
if (!parts) return false
if (!parts) {
syncDebug.reducer.partDeltaNoParts(props.messageID, props.partID)
return false
}
const result = Binary.search(parts, props.partID, (p) => p.id)
if (!result.found) return false
if (!result.found) {
syncDebug.reducer.partDeltaNotFound(props.messageID, props.partID)
return false
}
const existing = parts[result.index] as Record<string, unknown>
const existingValue = existing[props.field] as string | undefined
// Create new Part object + new array so React detects the change
+101
View File
@@ -15,6 +15,7 @@ import { updateStreamingState } from "./streaming"
import { setActionRefs } from "./session-actions"
import { setSyncRefs } from "./sync-refs"
import { stripMessageDiffSnapshots, stripSessionDiffSnapshots } from "./sanitize"
import { syncDebug } from "./debug"
import { opencodeClient } from "@/lib/opencode/client"
import { usePermissionStore } from "@/stores/permissionStore"
import { autoRespondsPermission, normalizeDirectory } from "@/stores/utils/permissionAutoAccept"
@@ -100,6 +101,79 @@ const RECONNECT_SKIP_PARTS = new Set(["patch", "step-start", "step-finish"])
const cmp = (a: string, b: string) => (a < b ? -1 : a > b ? 1 : 0)
// ---------------------------------------------------------------------------
// Parts-gap recovery — when SSE events arrive but parts are missing,
// trigger a targeted re-fetch for the affected sessions.
// Tracked per-directory, deduplicated, and auto-expiring.
// ---------------------------------------------------------------------------
type PendingRepair = {
sessionID: string
directory: string
enqueuedAt: number
}
const REPAIR_COOLDOWN_MS = 5_000
const pendingRepairs = new Map<string, PendingRepair>() // key: directory:sessionID
const repairKey = (directory: string, sessionID: string) => `${directory}:${sessionID}`
function enqueuePartsRepair(directory: string, sessionID: string, childStores: ChildStoreManager) {
if (!directory || directory === "global" || !sessionID) return
const k = repairKey(directory, sessionID)
const existing = pendingRepairs.get(k)
if (existing && Date.now() - existing.enqueuedAt < REPAIR_COOLDOWN_MS) return
pendingRepairs.set(k, { sessionID, directory, enqueuedAt: Date.now() })
// Defer to next microtask so we don't hold up the current event batch
void Promise.resolve().then(async () => {
const store = childStores.getChild(directory)
if (!store) {
pendingRepairs.delete(k)
return
}
try {
await repairSessionParts(directory, sessionID, store)
} catch {
// Transient failure — next SSE event or reconnect will catch up.
} finally {
pendingRepairs.delete(k)
}
})
}
async function repairSessionParts(
directory: string,
sessionID: string,
store: StoreApi<DirectoryStore>,
) {
const scopedClient = opencodeClient.getScopedSdkClient(directory)
const result = await retry(() =>
scopedClient.session.messages({ sessionID, limit: RECONNECT_MESSAGE_LIMIT }),
)
const records = (result.data ?? []).filter((record: { info?: { id?: string } }) => !!record?.info?.id)
if (records.length === 0) return
store.setState((state: DirectoryStore) => {
const nextPartState = { ...state.part }
for (const record of records) {
const messageId = record?.info?.id
if (!messageId) continue
const newParts = (record.parts ?? [])
.filter((part: Part) => !!part?.id && !RECONNECT_SKIP_PARTS.has(part.type))
.sort((a: Part, b: Part) => cmp(a.id, b.id))
const existing = nextPartState[messageId]
// Only patch if parts were missing or fewer than server has
if (!existing || existing.length < newParts.length) {
nextPartState[messageId] = newParts
}
}
return { part: nextPartState }
})
}
// Module-level refs for notification viewed check.
// Used to determine if user is currently viewing the session when a notification arrives.
let _activeDirectory = ""
@@ -804,6 +878,33 @@ function handleEvent(
if (applyDirectoryEvent(draft, payload)) {
store.setState(draft)
const sessionID = getSessionIdFromPayload(payload) ?? undefined
const messageID = getMessageIdFromPayload(payload) ?? undefined
syncDebug.dispatch.eventApplied(payload.type, sessionID, messageID)
// Parts-gap recovery on message.updated: if the message was inserted or
// replaced but draft.part[messageID] is empty, the parts were lost or
// never arrived. Trigger repair so the UI doesn't render a blank bubble.
if (sessionID && messageID && payload.type === "message.updated") {
const after = store.getState()
const info = (payload.properties as { info: Message }).info
if (info.role === "assistant" && (!after.part[messageID] || after.part[messageID].length === 0)) {
enqueuePartsRepair(resolvedDirectory, sessionID, childStores)
}
}
} else {
const sessionID = getSessionIdFromPayload(payload) ?? undefined
const messageID = getMessageIdFromPayload(payload) ?? undefined
syncDebug.dispatch.eventNoChange(payload.type, sessionID, messageID)
// Parts-gap recovery: if a part event was dropped because the parts array
// was missing (message not yet inserted or parts lost), trigger a repair
// fetch for the session.
if (sessionID && messageID && (
payload.type === "message.part.delta" || payload.type === "message.part.updated"
)) {
enqueuePartsRepair(resolvedDirectory, sessionID, childStores)
}
}
updateRoutingIndexFromEvent(routingIndex, resolvedDirectory, payload)