Skip to content

fix(tools): MemoryReviewer --dry-run builds the prompts and stops #2109

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
pai-scaffolde wants to merge 1 commit into
base: main
Choose a base branch
from
Open

fix(tools): MemoryReviewer --dry-run builds the prompts and stops #2109

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
29 changes: 28 additions & 1 deletion LifeOS/install/LIFEOS/TOOLS/MemoryReviewer.ts
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -634,6 +634,9 @@ export interface ReviewResult {
// A skip is a healthy no-op (nothing to curate — e.g. a just-/clear'ed
// transcript), NOT a failure. Health checks must not count it as one.
skipped?: boolean;
/** --dry-run: prompts were built and nothing else happened (no inference, no run dir, no ledger row). */
dry_run?: boolean;
prompt_chars?: { system: number; user: number };
dispatch_summary?: DispatchSummary;
error?: string;
}
Expand Down Expand Up @@ -666,6 +669,18 @@ export async function review(opts: ReviewOptions = {}): Promise<ReviewResult> {
// prompts reach the model. No-op in the live tree.
const systemPrompt = renderNames(REVIEWER_SYSTEM_PROMPT);
const userPrompt = renderNames(buildReviewerUserPrompt(exchanges, snapshot));

// --dry-run stops here: extract + prompt, no inference. It also writes no
// run directory and no ledger row — CortexHealth reads both as evidence of
// a real run (an orphaned run dir grades as "timed-out"), so a dry run must
// leave nothing behind for it to misread. (public issue #2073)
if (opts.dryRun) {
return {
ok: true, runId, transcript, exchanges: exchanges.length, inference_duration_ms: 0, parse_ok: true,
dry_run: true, prompt_chars: { system: systemPrompt.length, user: userPrompt.length },
};
}

writeRunDebug(runId, {
"prompt.system.md": systemPrompt,
"prompt.user.md": userPrompt,
Expand Down Expand Up @@ -731,7 +746,7 @@ export async function review(opts: ReviewOptions = {}): Promise<ReviewResult> {
writeRunDebug(runId, { "response.parsed.json": JSON.stringify(parsed.output, null, 2) });

// 6. Dispatch
const { summary, results } = dispatchItems(parsed.output.items, { dryRun: opts.dryRun });
const { summary, results } = dispatchItems(parsed.output.items);
writeRunDebug(runId, {
"dispatch.log": [
`Items: ${summary.total} (succeeded=${summary.succeeded} failed=${summary.failed} skipped_guard=${summary.skipped_guard})`,
Expand Down Expand Up @@ -845,6 +860,18 @@ async function smokeTest(): Promise<number> {
check("E2E: proposal enqueue succeeded", r.dispatch_summary?.by_type.proposal === 1);
check("E2E: zero dispatch failures", r.dispatch_summary?.failed === 0);

// 7b. --dry-run: builds the prompts and stops — no inference, no run dir, no
// ledger row (public issue #2073). Counted against the same synth transcript.
const runDirs = () => existsSync(RUNS_DEBUG_DIR) ? readdirSync(RUNS_DEBUG_DIR).length : 0;
const ledgerRows = () => existsSync(RUNS_LOG_PATH) ? readFileSync(RUNS_LOG_PATH, "utf8").split("\n").filter(Boolean).length : 0;
const dirsBefore = runDirs(), rowsBefore = ledgerRows();
const dry = await review({ input: synthPath, turns: 5, dryRun: true, mockInferenceResponse: mockResponse });
check("dry-run: returns ok with dry_run marker", dry.ok && dry.dry_run === true && dry.exchanges === 1);
check("dry-run: no inference", dry.inference_duration_ms === 0 && dry.dispatch_summary === undefined);
check("dry-run: no run directory written", runDirs() === dirsBefore, `${dirsBefore} → ${runDirs()}`);
check("dry-run: no ledger row written", ledgerRows() === rowsBefore, `${rowsBefore} → ${ledgerRows()}`);
check("dry-run: reports prompt sizes", (dry.prompt_chars?.system ?? 0) > 0 && (dry.prompt_chars?.user ?? 0) > 0);

// Cleanup synth transcript + reviewer-runs debug dir for this run
try {
const { rmSync } = await import("node:fs");
Expand Down