fix(context-engine): honor assembled prompt authority in precheck (#74255)

Merged via squash.

Prepared head SHA: 650b02380b
Co-authored-by: 100yenadmin <239388517+100yenadmin@users.noreply.github.com>
Co-authored-by: jalehman <550978+jalehman@users.noreply.github.com>
Reviewed-by: @jalehman
This commit is contained in:
Andrew
2026-05-01 21:43:41 +07:00
committed by GitHub
parent 0ce0509856
commit 42584964ac
8 changed files with 177 additions and 5 deletions

View File

@@ -70,6 +70,7 @@ Docs: https://docs.openclaw.ai
- MCP/stdio: settle MCP stdio transport send() from the write callback instead of resolving immediately on buffer acceptance, so async write errors reject the promise instead of being lost. Refs #75438.
- Process/exec: add stdin error listener in runCommandWithTimeout so EPIPE from a prematurely-exited child is swallowed instead of escaping to uncaughtException. Refs #75438.
- Voice Call/realtime: add default-off fast memory/session context for `openclaw_agent_consult`, giving live calls a bounded answer-or-miss path before the full agent consult. Fixes #71849. Thanks @amzzzzzzz.
- Google Meet: interrupt Realtime provider output when local barge-in clears playback, so command-pair audio stops model speech instead of only restarting Chrome playback. Fixes #73850. (#73834) Thanks @shhtheonlyperson.
- Gateway/config: cap oversized plugin-owned schemas in the full `config.schema` response so large installed plugin sets cannot balloon Gateway RSS or crash schema clients. Thanks @vincentkoc.
- Plugins/update: skip ClawHub and marketplace plugin updates when the bundled version is newer than the recorded installed version, so `openclaw update` no longer overwrites working bundled plugins with older external packages. Fixes #75447. Thanks @amknight.
@@ -145,6 +146,7 @@ Docs: https://docs.openclaw.ai
- Pairing: surface unexpected allowlist filesystem stat errors instead of treating the allowlist as missing, so permission and I/O failures are visible during pairing authorization checks. (#63324) Thanks @franciscomaestre.
- macOS app: reserve layout space for exec approval command details so the allow dialog no longer overlaps the command, context, and action buttons. (#75470) Thanks @ngutman.
- Agents/failover: carry `sessionId`, `lane`, `provider`, `model`, and `profileId` attribution through `FailoverError` and `describeFailoverError`/`coerceToFailoverError` so structured error logs (e.g. `gateway.err.log` ingestion) can attribute exhausted-fallback wrapper errors to the originating session and last-attempted provider instead of dropping the metadata after the per-profile errors. Fixes #42713. (#73506) Thanks @wenxu007.
- Context Engine: treat assembled prompt as the default authority for preemptive overflow prechecks so engines that return a windowed, self-contained context no longer trigger false hard-fail compactions on huge raw history. Engines whose assembled view can hide overflow risk can opt back into the legacy behavior with `AssembleResult.promptAuthority: "preassembly_may_overflow"`. (#74255) Thanks @100yenadmin.
## 2026.4.29

View File

@@ -1,2 +1,2 @@
c1446005a26262d6b817d72493471d11c618b98441fad2014f1cf422bfe64bc9 plugin-sdk-api-baseline.json
1b7d71eaabcae7d957396e7ff242598ef22b51851bc3fe1f4b58f2c2e5bf1459 plugin-sdk-api-baseline.jsonl
37787172adf7a55a32097599b4bf5729fc7138c8743c6f4c9d58fc8d01df72a1 plugin-sdk-api-baseline.json
0ec4957528477832085c638a5f7f691c878ba199f3e81f330f162c27cfd9ebf4 plugin-sdk-api-baseline.jsonl

View File

@@ -197,6 +197,17 @@ Required members:
<ParamField path="systemPromptAddition" type="string">
Prepended to the system prompt.
</ParamField>
<ParamField path="promptAuthority" type='"assembled" | "preassembly_may_overflow"'>
Controls which token estimate the runner uses for preemptive overflow
prechecks. Defaults to `"assembled"`, which means only the assembled
prompt's estimate is checked — appropriate for engines that return a
windowed, self-contained context. Set to `"preassembly_may_overflow"` only
when your assembled view can hide overflow risk in the underlying
transcript; the runner then takes the maximum of the assembled estimate
and the pre-assembly (unwindowed) session-history estimate when deciding
whether to preemptively compact. Either way, the messages you return are
still what the model sees — `promptAuthority` only affects the precheck.
</ParamField>
`compact` returns a `CompactResult`. When compaction rotates the active
transcript, `result.sessionId` and `result.sessionFile` identify the successor

View File

@@ -322,6 +322,118 @@ describe("runEmbeddedAttempt context engine sessionKey forwarding", () => {
);
});
it("uses assembled context as the default precheck authority", async () => {
let sawPrompt = false;
const hugeHistory = "large raw history ".repeat(25_000);
const result = await createContextEngineAttemptRunner({
contextEngine: createTestContextEngine({
assemble: async () => ({
messages: [
{ role: "user", content: "small assembled context", timestamp: 1 },
] as AgentMessage[],
estimatedTokens: 8,
}),
}),
sessionKey,
tempPaths,
sessionMessages: [{ role: "user", content: hugeHistory, timestamp: 1 }] as AgentMessage[],
attemptOverrides: {
contextTokenBudget: 500,
},
sessionPrompt: async (session) => {
sawPrompt = true;
session.messages = [
...session.messages,
{ role: "assistant", content: "done", timestamp: 2 },
];
},
});
expect(sawPrompt).toBe(true);
expect(result.promptError).toBeNull();
expect(result.promptErrorSource).toBeNull();
expect(hoisted.preemptiveCompactionCalls.at(-1)).not.toHaveProperty("unwindowedMessages");
});
it("honors context engines that opt into preassembly overflow authority", async () => {
let sawPrompt = false;
const hugeHistory = "large raw history ".repeat(25_000);
const result = await createContextEngineAttemptRunner({
contextEngine: createTestContextEngine({
assemble: async () => ({
messages: [
{ role: "user", content: "small assembled context", timestamp: 1 },
] as AgentMessage[],
estimatedTokens: 8,
promptAuthority: "preassembly_may_overflow",
}),
}),
sessionKey,
tempPaths,
sessionMessages: [{ role: "user", content: hugeHistory, timestamp: 1 }] as AgentMessage[],
attemptOverrides: {
contextTokenBudget: 500,
},
sessionPrompt: async (session) => {
sawPrompt = true;
session.messages = [
...session.messages,
{ role: "assistant", content: "done", timestamp: 2 },
];
},
});
expect(sawPrompt).toBe(false);
expect(result.promptErrorSource).toBe("precheck");
expect(result.preflightRecovery?.route).toBe("compact_only");
expect(hoisted.preemptiveCompactionCalls.at(-1)).toHaveProperty("unwindowedMessages");
});
it("snapshots pre-assembly messages before assemble even when the engine windows in place", async () => {
const hugeHistory = "large raw history ".repeat(25_000);
const preassemblyMarker = { role: "user", content: hugeHistory, timestamp: 1 } as AgentMessage;
await createContextEngineAttemptRunner({
contextEngine: createTestContextEngine({
assemble: async ({ messages }: { messages: AgentMessage[] }) => {
// Simulate an engine that windows the input array IN PLACE.
// The assemble contract does not require immutability, so the
// runner must have already snapshotted before calling us.
messages.length = 0;
messages.push({ role: "user", content: "windowed", timestamp: 2 } as AgentMessage);
return {
messages: [
{ role: "user", content: "small assembled context", timestamp: 1 },
] as AgentMessage[],
estimatedTokens: 8,
promptAuthority: "preassembly_may_overflow",
};
},
}),
sessionKey,
tempPaths,
sessionMessages: [preassemblyMarker],
attemptOverrides: {
contextTokenBudget: 500,
},
sessionPrompt: async (session) => {
session.messages = [
...session.messages,
{ role: "assistant", content: "done", timestamp: 3 },
];
},
});
const lastCall = hoisted.preemptiveCompactionCalls.at(-1);
expect(lastCall).toHaveProperty("unwindowedMessages");
const unwindowed = (lastCall as { unwindowedMessages?: AgentMessage[] }).unwindowedMessages;
// The snapshot must reflect the true pre-assembly state, not the in-place
// windowed array that assemble mutated.
expect(unwindowed).toEqual([preassemblyMarker]);
});
it("keeps gateway model runs independent from agent context and session history", async () => {
const bootstrap = vi.fn(async () => ({ bootstrapped: true }));
const assemble = vi.fn(async ({ messages }: { messages: AgentMessage[] }) => ({

View File

@@ -26,6 +26,8 @@ type SubscribeEmbeddedPiSessionFn =
typeof import("../../pi-embedded-subscribe.js").subscribeEmbeddedPiSession;
type AcquireSessionWriteLockFn =
typeof import("../../session-write-lock.js").acquireSessionWriteLock;
type ShouldPreemptivelyCompactBeforePromptFn =
typeof import("./preemptive-compaction.js").shouldPreemptivelyCompactBeforePrompt;
type SubscriptionMock = ReturnType<SubscribeEmbeddedPiSessionFn>;
type UnknownMock = Mock<(...args: unknown[]) => unknown>;
@@ -78,6 +80,7 @@ type AttemptSpawnWorkspaceHoisted = {
(sessionKey: string | undefined, config: unknown) => number | undefined
>;
limitHistoryTurnsMock: Mock<<T>(messages: T, limit: number | undefined) => T>;
preemptiveCompactionCalls: Parameters<ShouldPreemptivelyCompactBeforePromptFn>[0][];
sessionManager: SessionManagerMocks;
};
@@ -148,6 +151,7 @@ const hoisted = vi.hoisted((): AttemptSpawnWorkspaceHoisted => {
const limitHistoryTurnsMock = vi.fn<<T>(messages: T, limit: number | undefined) => T>(
(messages) => messages,
);
const preemptiveCompactionCalls: Parameters<ShouldPreemptivelyCompactBeforePromptFn>[0][] = [];
const sessionManager = {
getLeafEntry: vi.fn(() => null),
branch: vi.fn(),
@@ -181,6 +185,7 @@ const hoisted = vi.hoisted((): AttemptSpawnWorkspaceHoisted => {
runContextEngineMaintenanceMock,
getDmHistoryLimitFromSessionKeyMock,
limitHistoryTurnsMock,
preemptiveCompactionCalls,
sessionManager,
};
});
@@ -583,6 +588,19 @@ vi.mock("../compaction-runtime-context.js", () => ({
buildEmbeddedCompactionRuntimeContext: () => ({}),
}));
vi.mock("./preemptive-compaction.js", async (importOriginal) => {
const actual = await importOriginal<typeof import("./preemptive-compaction.js")>();
return {
...actual,
shouldPreemptivelyCompactBeforePrompt: (
params: Parameters<typeof actual.shouldPreemptivelyCompactBeforePrompt>[0],
) => {
hoisted.preemptiveCompactionCalls.push(params);
return actual.shouldPreemptivelyCompactBeforePrompt(params);
},
};
});
vi.mock("../compaction-safety-timeout.js", () => ({
resolveCompactionTimeoutMs: () => undefined,
}));
@@ -770,6 +788,7 @@ export function resetEmbeddedAttemptHarness(
hoisted.runContextEngineMaintenanceMock.mockReset().mockResolvedValue(undefined);
hoisted.getDmHistoryLimitFromSessionKeyMock.mockReset().mockReturnValue(undefined);
hoisted.limitHistoryTurnsMock.mockReset().mockImplementation((messages) => messages);
hoisted.preemptiveCompactionCalls.length = 0;
hoisted.sessionManager.getLeafEntry.mockReset().mockReturnValue(null);
hoisted.sessionManager.branch.mockReset();
hoisted.sessionManager.resetLeaf.mockReset();

View File

@@ -10,6 +10,7 @@ import {
import { isAcpRuntimeSpawnAvailable } from "../../../acp/runtime/availability.js";
import { filterHeartbeatPairs } from "../../../auto-reply/heartbeat-filter.js";
import { getRuntimeConfig } from "../../../config/config.js";
import type { AssembleResult } from "../../../context-engine/types.js";
import { emitTrustedDiagnosticEvent } from "../../../infra/diagnostic-events.js";
import {
createChildDiagnosticTraceContext,
@@ -1527,6 +1528,8 @@ export async function runEmbeddedAttempt(
}
let prePromptMessageCount = activeSession.messages.length;
let unwindowedContextEngineMessagesForPrecheck: AgentMessage[] | undefined;
let contextEnginePromptAuthority: NonNullable<AssembleResult["promptAuthority"]> =
"assembled";
abortSessionForYield = () => {
yieldAbortSettled = Promise.resolve(activeSession.abort());
};
@@ -2071,7 +2074,11 @@ export async function runEmbeddedAttempt(
if (activeContextEngine) {
try {
unwindowedContextEngineMessagesForPrecheck = activeSession.messages.slice();
// Snapshot before assemble: the assemble contract does not require
// the input array to be treated immutably, so an engine that windows
// history in place would otherwise leave the precheck reading
// already-windowed messages instead of the true pre-assembly state.
const preassemblyContextEngineMessagesForPrecheck = activeSession.messages.slice();
const assembled = await assembleAttemptContextEngine({
contextEngine: activeContextEngine,
sessionId: params.sessionId,
@@ -2089,6 +2096,11 @@ export async function runEmbeddedAttempt(
if (assembled.messages !== activeSession.messages) {
activeSession.agent.state.messages = assembled.messages;
}
contextEnginePromptAuthority = assembled.promptAuthority ?? "assembled";
if (contextEnginePromptAuthority === "preassembly_may_overflow") {
unwindowedContextEngineMessagesForPrecheck =
preassemblyContextEngineMessagesForPrecheck;
}
if (assembled.systemPromptAddition) {
systemPromptText = prependSystemPromptAddition({
systemPrompt: systemPromptText,
@@ -2760,7 +2772,9 @@ export async function runEmbeddedAttempt(
const preemptiveCompaction = shouldPreemptivelyCompactBeforePrompt({
messages: activeSession.messages,
unwindowedMessages: unwindowedContextEngineMessagesForPrecheck,
...(contextEnginePromptAuthority === "preassembly_may_overflow"
? { unwindowedMessages: unwindowedContextEngineMessagesForPrecheck }
: {}),
systemPrompt: systemPromptText,
prompt: effectivePrompt,
contextTokenBudget,

View File

@@ -93,7 +93,7 @@ describe("preemptive-compaction", () => {
expect(result.estimatedPromptTokens).toBeLessThan(result.promptBudgetBeforeReserve);
});
it("uses the larger unwindowed message estimate when context engine assembly windows history", () => {
it("uses the larger unwindowed message estimate when explicitly provided", () => {
const result = shouldPreemptivelyCompactBeforePrompt({
messages: [makeAssistantHistory("small assembled window")],
unwindowedMessages: [makeAssistantHistory(verboseHistory.repeat(4))],

View File

@@ -8,6 +8,20 @@ export type AssembleResult = {
messages: AgentMessage[];
/** Estimated total tokens in assembled context */
estimatedTokens: number;
/**
* Controls which token estimate the runner treats as authoritative for
* preemptive overflow prechecks. The returned `messages` are always the
* prompt sent to the model; this only affects the precheck's token comparison.
*
* - "assembled": the precheck uses only the assembled prompt's estimate.
* - "preassembly_may_overflow": the precheck takes the maximum of the
* assembled estimate and the pre-assembly (unwindowed) session-history
* estimate. Engines opt into this when their assembled view can hide an
* overflow that would still affect the underlying transcript.
*
* Defaults to "assembled".
*/
promptAuthority?: "assembled" | "preassembly_may_overflow";
/** Optional context-engine-provided instructions prepended to the runtime system prompt */
systemPromptAddition?: string;
};