perf(think,sessions): stop re-reading the transcript during a turn by mattzcarey · Pull Request #2219 · cloudflare/agents

mattzcarey · 2026-09-08T17:46:49Z

What

Think stops reading the transcript during a turn.

A long Think turn used to read the whole persisted history 2 + T times, where T is the number of tool updates in the turn (client results, approvals, cross-message results from approved server tools, execution outcomes). Each read walks the sized path query (several billed rows per message for the byte subqueries) and then hydrates every row. With a 400-message transcript and 20 tool updates that is on the order of 10k billed row reads for a turn whose stream log writes about 30 rows.

Four changes, all on the read side. Nothing about what is durable, or when, moves.

A tool update resolves its one target row from memory. _applyToolUpdateToMessages no longer calls session.getHistory(). _resolveToolCallOwner asks the in-flight accumulator first (a row under its id exists only when stall recovery persisted a partial), then the live cache for the owner's id, and reads that row alone. The apply is still a first-write-wins read-modify-write of the stored form, so Tool part state regresses during client tool round-trip; continuation starts but emits no output #1404 replays still write nothing and _scheduleAutoContinuation 50ms timer fires before all parallel client-tool results arrive → MissingToolResultsError #1649 mid-stream results still reach both the accumulator and the row. Storage is scanned only when the cache does not cover the active path (a windowed hydration or a failed boot hydration), and then newest first, stopping at the owner.
A chat request reconciles against the cache. _reconcileAndPersistIncoming uses this.messages as the server transcript when the cache covers the path, skips echoed messages whose stored form is already what the server holds (the client posts its whole transcript every request; each echoed message used to cost Sessions an existence read plus a full-row compare), and no longer re-reads the path after the writes. The change feed already patched the cache for every write. A windowed cache keeps the old reads.
Media eviction is gated from memory and armed by appends. Every cache refresh used to arm a pass whose first act is a content-free scan of the whole stored path. The gate now runs on each linear append (the event that ages older messages) and schedules a pass only when an aged cached message still carries an inline payload (hasEvictableMedia), or, when aged rows are not all in memory, once per distinct stored size.
Two per-call costs become per-isolate. think_config writes of an unchanged request body or client-tool schemas are skipped, and the agent-tool child-run DDL (two CREATE IF NOT EXISTS plus three ALTER attempts that throw and are swallowed) runs once per isolate instead of before every milestone, progress snapshot and child-run read.

Sessions gains the read shape Think needs: session.history({ newestFirst: true }) streams the active path leaf → root by following parent pointers, paying one row per message the consumer takes. Compaction overlays are honored without planning them up front: an overlay that applies ends at a row the walk reaches before anything it covers, so the walk stays row-by-row until it lands on a compaction's end, then plans the remaining prefix by id and streams it overlay-collapsed in eight-row windows. A lookup that stops in the messages after the last compaction never pays for that.

Sessions also reports two writes the change feed used to hide: import (one per row importMessage() actually writes) and compaction (an overlay stored through addCompaction()). Think marks its cache stale on import and refreshes on compaction, so a host cache can no longer claim to cover a path that changed underneath it. compact() still reports compact.

Two guards keep the cache honest over a long-lived isolate. Appends are charged against the hydration budget measured at the last refresh, and once they would carry the cache past it the cache stops claiming to cover the path, so the next boundary re-reads and re-windows (#1710). And the eviction gate treats compaction overlays like a windowed hydration: rows collapsed under an overlay are not in the cache, so a compacted session gets one content-free pass per distinct stored size rather than none.

Measured

Billed rows, real Durable Object SQLite, pinned in tests.

Operation	Before	After
Find a tool call's owner three messages from the leaf, 200-message path (Sessions bench)	1801 rows read (`getHistory`)	< 12 rows read (`history({ newestFirst })`)
One client tool result applied, 12-message vs 160-message transcript (Think)	grows with the transcript	identical, < 20 rows read
Turn start with the client echoing 12 vs 160 messages plus one new user message (Think)	grows with the transcript	identical, < 10 rows read, 1 row written

The Think numbers come from a rowsRead/rowsWritten counter on the test agent's ctx.storage.sql.exec, the same accounting the streams and sessions benches use.

Not in this PR

The recovery progress counter still does a KV get + put per credited chunk (two per tool call, one per text/reasoning segment start, one per 5 s of deltas). Deriving it from the stream log's tail, or crediting on flush, changes a persisted contract shared with @cloudflare/ai-chat and belongs in its own change.
compactAfter re-derives the token estimate over the whole content-free path on every inserted append (Sessions). A running total kept alongside the leaf cache would remove it.
On a wake recovery the orphaned stream's chunks are read in full up to three times across getPartialStreamText, persistOrphanedStream and the give-up path.
The _annotateActionApprovalChunk per-chunk SELECT only fires for paused durable actions and is left alone.

Tests

agents sessions suites (69): newest-first ordering, overlays, early exit, branch leaf; import and compaction change events; the storage-ops bench pins the owner lookup under 12 rows against a full read over 1000, with and without a compacted prefix, and bounds a lookup that has to enter the overlay.
@cloudflare/think suites: client-tools (new rows-read tests for tool update and turn start), hydration-budget (tool result for a row outside the window still lands via the storage fallback), media-eviction (hasEvictableMedia; a linear append schedules the pass with no cache refresh), plus the full Think workers project and the @cloudflare/ai-chat workers project.
npm run typecheck, oxfmt --check ., oxlint clean.

A long Think turn read the whole persisted history 2 + T times: once to reconcile the client's transcript, once to refresh the cache after persisting it, and once per tool update (client result, approval, cross-message result, execution outcome) to find the one message that owns the call. Each read pays the sized path query and hydrates every row, so the billed reads grew with transcript length times tool count. Think now resolves a tool update's owner from the in-flight accumulator and the live cache and reads that row alone; a chat request reconciles against the cache, skips echoed messages whose stored form is unchanged, and does not re-read the path after its writes. Storage is walked only when the cache does not cover the active path, and then newest first, stopping at the owner. The cache re-windows itself once appends carry it past the hydration budget and marks itself stale on Sessions `import` and `compaction` events. Media eviction is gated from memory on each linear append instead of a stored-path scan after every cache refresh; unchanged think_config rewrites and the per-call agent-tool DDL are skipped. Sessions gains `history({ newestFirst: true })`, a leaf-to-root walk over parent pointers that pays one row per message the consumer takes and plans compaction overlays only once it reaches a compacted span, and reports `import` and `compaction` writes on the change feed. Pinned on real DO SQLite: the owner lookup on a 200-message path reads under 12 rows against 1801 for a full read; a Think tool update and a turn start bill the same rows on a 12-message and a 160-message transcript.

changeset-bot · 2026-09-08T17:46:55Z

🦋 Changeset detected

Latest commit: 160c147

The changes in this PR will be included in the next version bump.

This PR includes changesets to release 3 packages

Name	Type
agents	Patch
@cloudflare/think	Patch
@cloudflare/agent-think	Patch

Not sure what this means? Click here to learn what changesets are.

Click here if you're a maintainer who wants to add another changeset to this PR

agent-think · 2026-09-08T17:48:51Z

🟡 agents import sizes

Measured 267 runtime imports as minified bundles. The primary size is gzip; raw minified size is included for diagnosis. An existing import growing by more than 10% is marked red. This report is informational.

Red	Yellow	Green	Unchanged	New	Removed
0	2	0	263	2	0

Compared 94485788 with 160c1473. Open workflow run.

Changed imports (4)

Status	Import	Base gzip	Head gzip	Delta
🟡	`agents/sessions#Sessions`	9.0 KiB	9.4 KiB	+418 B (+4.56%)
🟡	`agents/sessions#Session`	2.0 KiB	2.0 KiB	+22 B (+1.07%)
⚪	`agents/sessions#COMPACTION_PREFIX`	—	147 B	—
⚪	`agents/sessions#isCompactionMessage`	—	177 B	—

All 267 current runtime imports

Status	Import	Gzip	Raw minified
⚪	`agents#__DO_NOT_USE_WILL_BREAK__agentContext`	258.6 KiB	1130.0 KiB
⚪	`agents#__DO_NOT_USE_WILL_BREAK__withInvocationScope`	258.6 KiB	1130.0 KiB
⚪	`agents#Agent`	258.6 KiB	1130.0 KiB
⚪	`agents#AGENT_TOOL_MILESTONE_PART`	258.6 KiB	1130.0 KiB
⚪	`agents#AGENT_TOOL_PROGRESS_PART`	258.6 KiB	1130.0 KiB
⚪	`agents#buildAgentPath`	259.1 KiB	1132.3 KiB
⚪	`agents#buildAgentUrl`	259.3 KiB	1132.7 KiB
⚪	`agents#callable`	258.6 KiB	1130.1 KiB
⚪	`agents#camelCaseToKebabCase`	258.6 KiB	1130.0 KiB
⚪	`agents#createHeaderBasedEmailResolver`	258.8 KiB	1130.4 KiB
⚪	`agents#DEFAULT_AGENT_STATIC_OPTIONS`	258.6 KiB	1130.0 KiB
⚪	`agents#DurableObjectOAuthClientProvider`	258.6 KiB	1130.0 KiB
⚪	`agents#getAgentByName`	258.6 KiB	1130.0 KiB
⚪	`agents#getCurrentAgent`	258.6 KiB	1130.0 KiB
⚪	`agents#getSubAgentByName`	258.9 KiB	1130.7 KiB
⚪	`agents#isDurableObjectCodeUpdateReset`	258.6 KiB	1130.0 KiB
⚪	`agents#isDurableObjectMemoryLimitReset`	258.6 KiB	1130.0 KiB
⚪	`agents#isDurableObjectStorageReset`	258.6 KiB	1130.1 KiB
⚪	`agents#isPlatformTransientError`	258.6 KiB	1130.0 KiB
⚪	`agents#MCP_SERVER_ID_MAX_LENGTH`	258.6 KiB	1130.0 KiB
⚪	`agents#MessageType`	258.8 KiB	1130.3 KiB
⚪	`agents#normalizeServerId`	258.6 KiB	1130.0 KiB
⚪	`agents#parseSubAgentPath`	258.6 KiB	1130.0 KiB
⚪	`agents#routeAgentEmail`	258.9 KiB	1130.7 KiB
⚪	`agents#routeAgentRequest`	259.2 KiB	1131.9 KiB
⚪	`agents#routeSubAgentRequest`	258.8 KiB	1130.6 KiB
⚪	`agents#SqlError`	258.6 KiB	1130.0 KiB
⚪	`agents#StreamingResponse`	258.6 KiB	1130.0 KiB
⚪	`agents#SUB_PREFIX`	258.6 KiB	1130.0 KiB
⚪	`agents#unstable_callable`	258.7 KiB	1130.2 KiB
⚪	`agents/agent-tools#agentTool`	112.5 KiB	538.2 KiB
⚪	`agents/browser#BrowserConnector`	50.5 KiB	176.6 KiB
⚪	`agents/browser#browserContent`	36.3 KiB	127.4 KiB
⚪	`agents/browser#browserExtract`	36.3 KiB	127.4 KiB
⚪	`agents/browser#browserLinks`	36.3 KiB	127.4 KiB
⚪	`agents/browser#browserMarkdown`	36.3 KiB	127.4 KiB
⚪	`agents/browser#browserPdf`	36.3 KiB	127.3 KiB
⚪	`agents/browser#BrowserRenderingError`	36.0 KiB	126.7 KiB
⚪	`agents/browser#browserScrape`	36.3 KiB	127.4 KiB
⚪	`agents/browser#browserScreenshot`	36.3 KiB	127.3 KiB
⚪	`agents/browser#browserSnapshot`	36.3 KiB	127.4 KiB
⚪	`agents/browser#CdpSession`	37.2 KiB	129.8 KiB
⚪	`agents/browser#CodemodeRuntime`	39.6 KiB	139.0 KiB
⚪	`agents/browser#connectBrowser`	37.8 KiB	131.4 KiB
⚪	`agents/browser#connectBrowserSession`	37.5 KiB	130.4 KiB
⚪	`agents/browser#connectUrl`	37.6 KiB	130.5 KiB
⚪	`agents/browser#createBrowserSession`	36.3 KiB	127.5 KiB
⚪	`agents/browser#DEFAULT_EXEC_SWEEP_IDLE_MS`	36.0 KiB	126.6 KiB
⚪	`agents/browser#DEFAULT_SWEEP_IDLE_MS`	36.0 KiB	126.6 KiB
⚪	`agents/browser#deleteBrowserSession`	36.1 KiB	126.9 KiB
⚪	`agents/browser#DurableBrowserSessionStore`	36.4 KiB	127.6 KiB
⚪	`agents/browser#getBrowserRecording`	36.2 KiB	127.1 KiB
⚪	`agents/browser#listBrowserTargets`	36.1 KiB	126.9 KiB
⚪	`agents/browser#loadCdpSpec`	36.6 KiB	128.3 KiB
⚪	`agents/browser#runQuickAction`	36.0 KiB	126.6 KiB
⚪	`agents/browser/ai#createBrowserRuntime`	146.0 KiB	630.3 KiB
⚪	`agents/browser/ai#createBrowserTools`	146.0 KiB	630.3 KiB
⚪	`agents/browser/ai#createQuickActionTools`	122.5 KiB	554.3 KiB
⚪	`agents/browser/tanstack-ai#createBrowserTools`	161.7 KiB	699.2 KiB
⚪	`agents/chat#AbortRegistry`	2.5 KiB	8.9 KiB
⚪	`agents/chat#AGENT_TOOL_STREAM_PROGRESS_BUMP_THROTTLE_MS`	2.3 KiB	8.2 KiB
⚪	`agents/chat#AgentToolProgressEmitter`	2.7 KiB	9.5 KiB
⚪	`agents/chat#AgentToolStreamProgressThrottle`	2.4 KiB	8.3 KiB
⚪	`agents/chat#aiSdkRecoveryCodec`	2.3 KiB	8.2 KiB
⚪	`agents/chat#applyAgentToolEvent`	3.2 KiB	11.0 KiB
⚪	`agents/chat#applyChunkToParts`	2.3 KiB	8.2 KiB
⚪	`agents/chat#applyToolUpdate`	2.4 KiB	8.4 KiB
⚪	`agents/chat#AutoContinuationController`	2.3 KiB	8.2 KiB
⚪	`agents/chat#awaitWithDeadline`	2.4 KiB	8.4 KiB
⚪	`agents/chat#broadcastTransition`	3.2 KiB	11.4 KiB
⚪	`agents/chat#buildChatRecoveringFrame`	2.4 KiB	8.4 KiB
⚪	`agents/chat#buildInClauseStrings`	2.4 KiB	8.4 KiB
⚪	`agents/chat#bumpChatRecoveryProgress`	2.4 KiB	8.3 KiB
⚪	`agents/chat#byteLength`	2.5 KiB	8.4 KiB
⚪	`agents/chat#CHAT_LAST_TERMINAL_KEY`	2.3 KiB	8.2 KiB
⚪	`agents/chat#CHAT_MESSAGE_TYPES`	2.3 KiB	8.2 KiB
⚪	`agents/chat#CHAT_RECOVERING_FLAG_TTL_MS`	2.3 KiB	8.2 KiB
⚪	`agents/chat#CHAT_RECOVERING_KEY`	2.3 KiB	8.2 KiB
⚪	`agents/chat#CHAT_RECOVERY_ALARM_DEBOUNCE_MS`	2.3 KiB	8.2 KiB
⚪	`agents/chat#CHAT_RECOVERY_INCIDENT_KEY_PREFIX`	2.3 KiB	8.2 KiB
⚪	`agents/chat#CHAT_RECOVERY_INCIDENT_TTL_MS`	2.3 KiB	8.2 KiB
⚪	`agents/chat#CHAT_RECOVERY_PROGRESS_KEY`	2.3 KiB	8.2 KiB
⚪	`agents/chat#CHAT_RECOVERY_STABLE_RETRY_DELAY_SECONDS`	2.3 KiB	8.2 KiB
⚪	`agents/chat#CHAT_RECOVERY_TASK_NAME`	2.3 KiB	8.2 KiB
⚪	`agents/chat#CHAT_STREAM_PROGRESS_CREDIT_THROTTLE_MS`	2.3 KiB	8.2 KiB
⚪	`agents/chat#ChatRecoveryEngine`	4.4 KiB	15.3 KiB
⚪	`agents/chat#chatRecoveryTaskRunOptions`	2.4 KiB	8.6 KiB
⚪	`agents/chat#ChatStreamStalledError`	2.4 KiB	8.3 KiB
⚪	`agents/chat#classifyAgentToolChildRecovery`	2.4 KiB	8.5 KiB
⚪	`agents/chat#clearChatTerminal`	2.3 KiB	8.3 KiB
⚪	`agents/chat#clientResolvableToolNames`	2.3 KiB	8.3 KiB
⚪	`agents/chat#ContinuationState`	2.7 KiB	9.8 KiB
⚪	`agents/chat#createAgentToolEventState`	2.3 KiB	8.3 KiB
⚪	`agents/chat#createChatFiberSnapshot`	2.5 KiB	8.6 KiB
⚪	`agents/chat#createChatRecoveryTaskDefinition`	2.6 KiB	9.0 KiB
⚪	`agents/chat#createChatStreams`	6.3 KiB	21.7 KiB
⚪	`agents/chat#createChatTurnTaskDefinition`	2.7 KiB	8.9 KiB
⚪	`agents/chat#createToolsFromClientSchemas`	114.3 KiB	545.4 KiB
⚪	`agents/chat#crossMessageToolResultUpdate`	2.4 KiB	8.6 KiB
⚪	`agents/chat#DEFAULT_CHAT_RECOVERY_MAX_ATTEMPTS`	2.3 KiB	8.2 KiB
⚪	`agents/chat#DEFAULT_CHAT_RECOVERY_MAX_OOM_RETRIES`	2.3 KiB	8.2 KiB
⚪	`agents/chat#DEFAULT_CHAT_RECOVERY_MAX_WORK`	2.3 KiB	8.2 KiB
⚪	`agents/chat#DEFAULT_CHAT_RECOVERY_NO_PROGRESS_TIMEOUT_MS`	2.3 KiB	8.2 KiB
⚪	`agents/chat#DEFAULT_CHAT_RECOVERY_STABLE_TIMEOUT_MS`	2.3 KiB	8.2 KiB
⚪	`agents/chat#DEFAULT_CHAT_RECOVERY_TERMINAL_MESSAGE`	2.4 KiB	8.3 KiB
⚪	`agents/chat#dispatchChatRecoveryToHandoff`	3.0 KiB	9.9 KiB
⚪	`agents/chat#drainInteractionApplies`	2.3 KiB	8.3 KiB
⚪	`agents/chat#enforceRowSizeLimit`	3.5 KiB	11.2 KiB
⚪	`agents/chat#hasIncompleteToolBatch`	2.4 KiB	8.6 KiB
⚪	`agents/chat#interceptAgentToolBroadcast`	2.5 KiB	8.7 KiB
⚪	`agents/chat#isPlatformFailure`	2.6 KiB	8.9 KiB
⚪	`agents/chat#isReplayChunk`	2.4 KiB	8.7 KiB
⚪	`agents/chat#iterateWithStallWatchdog`	2.6 KiB	8.8 KiB
⚪	`agents/chat#KV_DELETE_MAX_KEYS`	2.3 KiB	8.2 KiB
⚪	`agents/chat#listActiveChatRecoveryIncidents`	2.4 KiB	8.4 KiB
⚪	`agents/chat#MAX_BOUND_PARAMS`	2.3 KiB	8.2 KiB
⚪	`agents/chat#MessageType`	2.4 KiB	9.0 KiB
⚪	`agents/chat#normalizeToolInput`	2.3 KiB	8.2 KiB
⚪	`agents/chat#parseProtocolMessage`	2.5 KiB	9.0 KiB
⚪	`agents/chat#partAwaitsClientInteraction`	2.4 KiB	8.6 KiB
⚪	`agents/chat#pausedExecutionUpdate`	2.4 KiB	8.4 KiB
⚪	`agents/chat#pendingChatTerminal`	2.3 KiB	8.3 KiB
⚪	`agents/chat#persistReconstructedOrphan`	3.1 KiB	11.0 KiB
⚪	`agents/chat#PreStreamTurns`	2.6 KiB	9.3 KiB
⚪	`agents/chat#readChatRecoveryProgress`	2.3 KiB	8.3 KiB
⚪	`agents/chat#reconcileMessages`	2.8 KiB	9.6 KiB
⚪	`agents/chat#reconcileOrphanPartial`	2.4 KiB	8.5 KiB
⚪	`agents/chat#recordChatTerminal`	2.4 KiB	8.3 KiB
⚪	`agents/chat#repairInterruptedToolParts`	2.6 KiB	9.1 KiB
⚪	`agents/chat#resolveChatRecoveryConfig`	2.6 KiB	8.9 KiB
⚪	`agents/chat#resolveToolMergeId`	2.4 KiB	8.5 KiB
⚪	`agents/chat#ResumableStream`	4.6 KiB	15.6 KiB
⚪	`agents/chat#ResumeHandshake`	3.0 KiB	10.4 KiB
⚪	`agents/chat#ROW_MAX_BYTES`	2.3 KiB	8.2 KiB
⚪	`agents/chat#runChatRecoveryExhaustion`	2.6 KiB	8.9 KiB
⚪	`agents/chat#sanitizeMessage`	2.5 KiB	8.8 KiB
⚪	`agents/chat#sendIfOpen`	2.4 KiB	8.4 KiB
⚪	`agents/chat#setChatRecovering`	2.5 KiB	8.5 KiB
⚪	`agents/chat#shouldCreditStreamProgress`	2.4 KiB	8.3 KiB
⚪	`agents/chat#STREAM_RESUME_NONE_REASONS`	2.3 KiB	8.2 KiB
⚪	`agents/chat#StreamAccumulator`	2.9 KiB	10.7 KiB
⚪	`agents/chat#StreamProgressCreditThrottle`	2.4 KiB	8.3 KiB
⚪	`agents/chat#SubmitConcurrencyController`	2.9 KiB	10.3 KiB
⚪	`agents/chat#sweepStaleChatRecoveryIncidents`	2.4 KiB	8.5 KiB
⚪	`agents/chat#TextSegmentJoiner`	2.7 KiB	9.2 KiB
⚪	`agents/chat#TIMED_OUT`	2.3 KiB	8.2 KiB
⚪	`agents/chat#toolApprovalUpdate`	2.4 KiB	8.5 KiB
⚪	`agents/chat#toolPartHasSettledResult`	2.3 KiB	8.4 KiB
⚪	`agents/chat#toolResultUpdate`	2.4 KiB	8.5 KiB
⚪	`agents/chat#truncateOlderMessages`	3.2 KiB	10.4 KiB
⚪	`agents/chat#TurnQueue`	2.6 KiB	9.2 KiB
⚪	`agents/chat#unwrapChatFiberSnapshot`	2.4 KiB	8.5 KiB
⚪	`agents/chat#wrapChatFiberSnapshot`	2.3 KiB	8.2 KiB
⚪	`agents/chat-sdk#ChatSdkStateAdapter`	261.0 KiB	1141.6 KiB
⚪	`agents/chat-sdk#ChatSdkStateAgent`	260.4 KiB	1139.1 KiB
⚪	`agents/chat-sdk#createChatSdkState`	261.0 KiB	1141.6 KiB
⚪	`agents/chat-sdk#defaultKeyShard`	258.8 KiB	1130.2 KiB
⚪	`agents/chat-sdk#defaultThreadShard`	258.7 KiB	1130.1 KiB
⚪	`agents/chat/react#detectToolsRequiringConfirmation`	3.3 KiB	8.3 KiB
⚪	`agents/chat/react#extractClientToolSchemas`	3.2 KiB	8.3 KiB
⚪	`agents/chat/react#getAgentMessages`	3.4 KiB	8.6 KiB
⚪	`agents/chat/react#getToolApproval`	3.1 KiB	8.0 KiB
⚪	`agents/chat/react#getToolCallId`	3.1 KiB	8.0 KiB
⚪	`agents/chat/react#getToolInput`	3.1 KiB	8.0 KiB
⚪	`agents/chat/react#getToolOutput`	3.1 KiB	8.0 KiB
⚪	`agents/chat/react#getToolPartState`	3.2 KiB	8.2 KiB
⚪	`agents/chat/react#useAgentChat`	132.9 KiB	609.7 KiB
⚪	`agents/chat/react#WebSocketChatTransport`	5.7 KiB	17.1 KiB
⚪	`agents/chat/transport#WebSocketChatTransport`	2.8 KiB	9.2 KiB
⚪	`agents/client#AgentClient`	5.7 KiB	16.6 KiB
⚪	`agents/client#AgentConnectionError`	582 B	993 B
⚪	`agents/client#agentFetch`	4.2 KiB	12.3 KiB
⚪	`agents/client#createStubProxy`	638 B	1.0 KiB
⚪	`agents/client#DEFAULT_CALL_TIMEOUT_MS`	473 B	770 B
⚪	`agents/client#isTerminalCloseEvent`	509 B	822 B
⚪	`agents/context#AgentContextProvider`	412 B	792 B
⚪	`agents/context#AgentSearchProvider`	640 B	1.3 KiB
⚪	`agents/context#ContextBlocks`	87.4 KiB	429.7 KiB
⚪	`agents/email#createAddressBasedEmailResolver`	193 B	227 B
⚪	`agents/email#createCatchAllEmailResolver`	110 B	97 B
⚪	`agents/email#createHeaderBasedEmailResolver`	334 B	492 B
⚪	`agents/email#createSecureReplyEmailResolver`	718 B	1.3 KiB
⚪	`agents/email#DEFAULT_MAX_AGE_SECONDS`	56 B	39 B
⚪	`agents/email#isAutoReplyEmail`	201 B	249 B
⚪	`agents/email#signAgentHeaders`	424 B	812 B
⚪	`agents/experimental/webmcp#registerWebMcp`	85.2 KiB	295.8 KiB
⚪	`agents/lifecycle#getCurrentAgent`	376 B	798 B
⚪	`agents/lifecycle#Lifecycle`	8.3 KiB	25.8 KiB
⚪	`agents/lifecycle#LifecycleCapability`	484 B	975 B
⚪	`agents/mcp#createLegacyMcpHandler`	375.9 KiB	1572.2 KiB
⚪	`agents/mcp#createMcpHandler`	388.2 KiB	1617.4 KiB
⚪	`agents/mcp#DurableObjectEventStore`	342.5 KiB	1430.7 KiB
⚪	`agents/mcp#ElicitRequestSchema`	342.5 KiB	1430.7 KiB
⚪	`agents/mcp#experimental_createMcpHandler`	376.1 KiB	1572.5 KiB
⚪	`agents/mcp#getMcpAuthContext`	342.5 KiB	1430.8 KiB
⚪	`agents/mcp#MCP_SERVER_ID_MAX_LENGTH`	342.5 KiB	1430.7 KiB
⚪	`agents/mcp#McpAgent`	342.5 KiB	1430.7 KiB
⚪	`agents/mcp#normalizeServerId`	342.5 KiB	1430.7 KiB
⚪	`agents/mcp#RPC_DO_PREFIX`	342.5 KiB	1430.7 KiB
⚪	`agents/mcp#RPCClientTransport`	342.5 KiB	1430.7 KiB
⚪	`agents/mcp#RPCServerTransport`	342.5 KiB	1430.7 KiB
⚪	`agents/mcp#SSEEdgeClientTransport`	342.6 KiB	1431.0 KiB
⚪	`agents/mcp#StreamableHTTPEdgeClientTransport`	342.6 KiB	1431.0 KiB
⚪	`agents/mcp#WorkerTransport`	345.8 KiB	1447.6 KiB
⚪	`agents/mcp/client#getNamespacedData`	62.9 KiB	240.0 KiB
⚪	`agents/mcp/client#MCP_SERVER_ID_MAX_LENGTH`	62.9 KiB	239.9 KiB
⚪	`agents/mcp/client#MCPClientManager`	158.6 KiB	702.6 KiB
⚪	`agents/mcp/client#normalizeServerId`	63.0 KiB	240.2 KiB
⚪	`agents/mcp/do-oauth-client-provider#DurableObjectOAuthClientProvider`	2.1 KiB	6.6 KiB
⚪	`agents/mcp/server#createMcpHandler`	80.5 KiB	307.2 KiB
⚪	`agents/mcp/server#getMcpAuthContext`	64.0 KiB	245.5 KiB
⚪	`agents/observability#channels`	259 B	549 B
⚪	`agents/observability#genericObservability`	470 B	1.2 KiB
⚪	`agents/observability#subscribe`	324 B	668 B
⚪	`agents/observability/ai#wrapAISDK`	8.8 KiB	30.5 KiB
⚪	`agents/react#_testUtils`	3.8 KiB	9.5 KiB
⚪	`agents/react#useAgent`	10.8 KiB	31.1 KiB
⚪	`agents/react#useAgentToolEvents`	5.6 KiB	16.8 KiB
⚪	`agents/routing#getAgentByName`	795 B	1.7 KiB
⚪	`agents/routing#routeAgentRequest`	1.6 KiB	3.6 KiB
⚪	`agents/routing#RoutedAgents`	2.4 KiB	6.2 KiB
⚪	`agents/schedule#getSchedulePrompt`	85.8 KiB	424.7 KiB
⚪	`agents/schedule#scheduleSchema`	85.3 KiB	423.6 KiB
⚪	`agents/schedule#unstable_getSchedulePrompt`	85.9 KiB	424.9 KiB
⚪	`agents/schedule#unstable_scheduleSchema`	85.3 KiB	423.6 KiB
⚪	`agents/schedules#Scheduler`	6.8 KiB	22.0 KiB
⚪	`agents/schedules/parser#getSchedulePrompt`	85.8 KiB	424.7 KiB
⚪	`agents/schedules/parser#scheduleSchema`	85.3 KiB	423.6 KiB
⚪	`agents/sessions#COMPACTION_PREFIX`	147 B	160 B
⚪	`agents/sessions#createCompactFunction`	1.7 KiB	4.0 KiB
⚪	`agents/sessions#isCompactionMessage`	177 B	200 B
🟡	`agents/sessions#Session`	2.0 KiB	6.2 KiB
🟡	`agents/sessions#Sessions`	9.4 KiB	33.6 KiB
⚪	`agents/skills#fromManifest`	309.8 KiB	1084.0 KiB
⚪	`agents/skills#parseSkillFrontmatter`	328.4 KiB	1146.2 KiB
⚪	`agents/skills#parseSkillMarkdown`	328.6 KiB	1146.5 KiB
⚪	`agents/skills#r2`	330.2 KiB	1150.4 KiB
⚪	`agents/skills#runner`	369.0 KiB	1297.8 KiB
⚪	`agents/skills#SkillRegistry`	416.4 KiB	1581.7 KiB
⚪	`agents/skills/compile#compileSkillScript`	15.4 KiB	43.4 KiB
⚪	`agents/skills/compile#isCompilableSkillScript`	15.4 KiB	43.3 KiB
⚪	`agents/streams#DEFAULT_MAX_CHUNK_BYTES`	94 B	96 B
⚪	`agents/streams#sseResponse`	855 B	1.6 KiB
⚪	`agents/streams#StreamClosedError`	175 B	212 B
⚪	`agents/streams#StreamNotFoundError`	215 B	276 B
⚪	`agents/streams#Streams`	4.1 KiB	13.5 KiB
⚪	`agents/streams#StreamSerializationError`	171 B	201 B
⚪	`agents/tasks#DuplicateTaskStepError`	328 B	463 B
⚪	`agents/tasks#MAX_SERIALIZED_BYTES`	190 B	232 B
⚪	`agents/tasks#MissingTaskDefinitionError`	358 B	536 B
⚪	`agents/tasks#NonRetryableError`	238 B	308 B
⚪	`agents/tasks#TaskReplayDivergedError`	341 B	483 B
⚪	`agents/tasks#Tasks`	8.9 KiB	31.8 KiB
⚪	`agents/tasks#TaskSerializationError`	258 B	339 B
⚪	`agents/types#MessageType`	211 B	365 B
⚪	`agents/vite#default`	353.8 KiB	1356.1 KiB
⚪	`agents/websockets#CALLABLES_RPC_QUERY`	12.4 KiB	43.3 KiB
⚪	`agents/websockets#CALLABLES_RPC_VALUE`	12.4 KiB	43.3 KiB
⚪	`agents/websockets#callablesFromDecorated`	12.7 KiB	44.3 KiB
⚪	`agents/websockets#callablesRpcUrl`	12.5 KiB	43.5 KiB
⚪	`agents/websockets#isCallablesRpcUpgrade`	12.4 KiB	43.4 KiB
⚪	`agents/websockets#WebSockets`	17.7 KiB	62.2 KiB
⚪	`agents/workflows#AgentWorkflow`	260.0 KiB	1134.8 KiB
⚪	`agents/workflows#WorkflowRejectedError`	258.7 KiB	1130.2 KiB
⚪	`agents/x402#normalizeNetwork`	14.7 KiB	61.1 KiB
⚪	`agents/x402#withX402`	23.0 KiB	89.2 KiB
⚪	`agents/x402#withX402Client`	104.1 KiB	346.5 KiB

Reported by agent-think[bot].

devin-ai-integration

Devin Review found 2 potential issues.

devin-ai-integration · 2026-09-08T17:51:17Z

+    this._cachedBytesSinceSync += JSON.stringify(message).length;
+    if (
+      this._lastHydration.totalContentBytes + this._cachedBytesSinceSync >
+      budget
+    ) {
+      this._cacheCoversActivePath = false;


🔴 Hydration budget misses cache growth

Large updates never reach _cachedBytesSinceSync, while non-ASCII appends count UTF-16 units instead of stored bytes. _cacheCoversActivePath can stay true beyond hydrationByteBudget, so later turns skip re-windowing and retain excess history.

Prompt for agents

Fix Think's post-hydration cache growth accounting in packages/think/src/think.ts. The current _noteCachedGrowth path only runs for linear appends and adds JSON.stringify(message).length, which is UTF-16 code units rather than the UTF-8/storage-sized units used by hydrationByteBudget. Updates can also enlarge cached messages without affecting the accounting. Track byte deltas for every cache mutation that can grow the cached transcript, using the same byte semantics as Sessions where possible, and mark _cacheCoversActivePath false once the refreshed total plus subsequent growth exceeds the budget. Preserve the optimization that avoids unnecessary full transcript reads.

Was this helpful? React with 👍 or 👎 to provide feedback.

devin-ai-integration · 2026-09-08T17:51:18Z

+    if (this._agedRowsHiddenFromCache()) {
+      if (
+        this._lastHydration !== null &&
+        this._mediaEvictionFruitlessAtBytes ===
+          this._lastHydration.totalContentBytes
+      ) {
+        return;


🟡 Media eviction stays suppressed after appends

After a fruitless compacted or windowed pass, appends leave _lastHydration.totalContentBytes unchanged. The stale equality suppresses later passes, so newly aged media remains inline until another refresh occurs.

Prompt for agents

Make the fruitless media-eviction gate reflect storage changes after the last hydration. In packages/think/src/think.ts, _mediaEvictionFruitlessAtBytes is compared against _lastHydration.totalContentBytes, but linear append events do not update that hydration snapshot. On compacted or windowed caches, a fruitless pass can therefore suppress all later append-triggered passes. Invalidate or advance the fruitless marker whenever a durable mutation changes candidate rows or total stored bytes, while retaining the no-repeat-scan optimization for genuinely unchanged storage.

Was this helpful? React with 👍 or 👎 to provide feedback.

pkg-pr-new · 2026-09-08T17:59:08Z

Open in StackBlitz

agents

npm i https://pkg.pr.new/agents@2219

@cloudflare/ai-chat

npm i https://pkg.pr.new/@cloudflare/ai-chat@2219

@cloudflare/codemode

npm i https://pkg.pr.new/@cloudflare/codemode@2219

hono-agents

npm i https://pkg.pr.new/hono-agents@2219

@cloudflare/shell

npm i https://pkg.pr.new/@cloudflare/shell@2219

@cloudflare/think

npm i https://pkg.pr.new/@cloudflare/think@2219

@cloudflare/voice

npm i https://pkg.pr.new/@cloudflare/voice@2219

@cloudflare/worker-bundler

npm i https://pkg.pr.new/@cloudflare/worker-bundler@2219

commit: 160c147

devin-ai-integration Bot reviewed Sep 8, 2026

View reviewed changes

Navigation Menu

Uh oh!

perf(think,sessions): stop re-reading the transcript during a turn - #2219

perf(think,sessions): stop re-reading the transcript during a turn#2219
mattzcarey wants to merge 1 commit into
mainfrom
perf/think-transcript-reads

mattzcarey commented Sep 8, 2026 •

edited by devin-ai-integration Bot

Loading

Uh oh!

changeset-bot Bot commented Sep 8, 2026

Uh oh!

agent-think Bot commented Sep 8, 2026

Uh oh!

devin-ai-integration Bot left a comment

Uh oh!

devin-ai-integration Bot Sep 8, 2026

Uh oh!

devin-ai-integration Bot Sep 8, 2026

Uh oh!

pkg-pr-new Bot commented Sep 8, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

Uh oh!

Conversation

mattzcarey commented Sep 8, 2026 • edited by devin-ai-integration Bot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

What

Measured

Not in this PR

Tests

Uh oh!

changeset-bot Bot commented Sep 8, 2026

🦋 Changeset detected

Uh oh!

agent-think Bot commented Sep 8, 2026

🟡 agents import sizes

Uh oh!

devin-ai-integration Bot left a comment

Choose a reason for hiding this comment

Uh oh!

devin-ai-integration Bot Sep 8, 2026

Choose a reason for hiding this comment

Uh oh!

devin-ai-integration Bot Sep 8, 2026

Choose a reason for hiding this comment

Uh oh!

pkg-pr-new Bot commented Sep 8, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

mattzcarey commented Sep 8, 2026 •

edited by devin-ai-integration Bot

Loading