hippo-memory 1.59.0 → 1.61.0

package/dist/predictions.js

@@ -1,620 +0,0 @@
-/**
- * E2 prediction first-class object (v0.31 / docs/plans/2026-05-26-e2-prediction-object.md).
- *
- * Canonical store for ex-ante claims that can be closed against ex-post
- * outcomes. The `predictions` table holds every field (including
- * `claim_text`); a memory row mirrors the claim for recall/inspect surfaces
- * but is NOT the source of truth — ON DELETE SET NULL on memory_id means
- * memory deletion gracefully orphans the prediction without losing data.
- *
- * Tenant scoping: every helper requires tenantId. The schema's BEFORE INSERT
- * + BEFORE UPDATE triggers (`trg_predictions_tenant_match_*`) enforce that
- * `predictions.tenant_id` matches the referenced memory's tenant_id when
- * `memory_id IS NOT NULL`. Cross-tenant references are unrepresentable at
- * the schema level.
- *
- * Dual-write atomicity: `savePrediction` writes the memory + predictions
- * row inside `writeEntry`'s SAVEPOINT 'write_entry' (store.ts:1196). The
- * afterWrite hook (store.ts:1199-1201) runs inside the same SAVEPOINT, so
- * a failure in either step rolls back both. Pattern matches supersede
- * (api.ts:1486) and the Slack/GitHub connectors.
- *
- * J3 (reference-class / planning-fallacy detector) reads from
- * `loadPredictionsByClass` to compute per-class base rates from
- * (estimate_value, actual_value) at query time. J3 is a follow-up episode;
- * this module ships the data layer.
- */
-import { BadRequestError, NotFoundError } from './api-errors.js';
-import { openHippoDb, closeHippoDb } from './db.js';
-import { writeEntry } from './store.js';
-import { assertTenantId } from './tenant.js';
-import { createMemory, Layer } from './memory.js';
-import { appendAuditEvent } from './audit.js';
-import { loadConfig } from './config.js';
-import { detectForwardClaim } from './forward-claim-detector.js';
-export const VALID_CLOSURE_STATES = new Set([
-    'open',
-    'closed',
-    'closed-unknown',
-]);
-function rowToPrediction(row) {
-    return {
-        id: row.id,
-        memoryId: row.memory_id,
-        tenantId: row.tenant_id,
-        classTag: row.class_tag,
-        claimText: row.claim_text,
-        estimateValue: row.estimate_value,
-        estimateUnit: row.estimate_unit,
-        targetDate: row.target_date,
-        actualValue: row.actual_value,
-        // SAFETY: closure_state is DB-constrained to VALID_CLOSURE_STATES; this
-        // module is the only writer and always inserts one of those literals.
-        closureState: row.closure_state,
-        closedAt: row.closed_at,
-        closureNote: row.closure_note,
-        createdAt: row.created_at,
-    };
-}
-// ---------------------------------------------------------------------------
-// Public API
-// ---------------------------------------------------------------------------
-/**
- * Create a new prediction. Writes a memory mirror + a predictions table
- * row atomically inside `writeEntry`'s SAVEPOINT 'write_entry'. On any
- * failure (audit write, predictions INSERT, trigger ABORT), the SAVEPOINT
- * rolls back — neither the memory row nor the predictions row lands.
- *
- * The memory is tagged `['prediction', classTag]` with `source='prediction'`
- * and `kind='distilled'`. It surfaces in `hippo recall` so the agent can
- * see open predictions naturally; the predictions table is the canonical
- * structured store used by J3.
- */
-export function savePrediction(hippoRoot, tenantId, opts, actor = 'cli') {
-    assertTenantId('savePrediction', tenantId);
-    if (!opts.classTag)
-        throw new BadRequestError('savePrediction: classTag is required');
-    if (!opts.claimText)
-        throw new BadRequestError('savePrediction: claimText is required');
-    const now = new Date().toISOString();
-    const mem = createMemory(opts.claimText, {
-        tags: ['prediction', opts.classTag],
-        layer: Layer.Semantic,
-        confidence: 'observed',
-        source: 'prediction',
-        // SAFETY: 'distilled' is a valid MemoryKind literal (see memory.ts).
-        kind: 'distilled',
-        baseHalfLifeDays: loadConfig(hippoRoot).defaultHalfLifeDays,
-        tenantId,
-    });
-    // Captured for the return value; populated inside afterWrite hook so the
-    // INSERT and the memory write share a SAVEPOINT.
-    let savedRow;
-    writeEntry(hippoRoot, mem, {
-        actor,
-        afterWrite: (db, memoryId) => {
-            const result = db.prepare(`
-        INSERT INTO predictions(
-          memory_id, tenant_id, class_tag, claim_text,
-          estimate_value, estimate_unit, target_date,
-          actual_value, closure_state, closed_at, closure_note, created_at
-        ) VALUES (?, ?, ?, ?, ?, ?, ?, ?, 'open', NULL, NULL, ?)
-      `).run(memoryId, tenantId, opts.classTag, opts.claimText, opts.estimateValue ?? null, opts.estimateUnit ?? null, opts.targetDate ?? null, null, // actual_value — null until close
-            now);
-            const predictionId = Number(result.lastInsertRowid ?? 0);
-            // SAFETY: row's shape matches the columns named in the SELECT above.
-            const row = db.prepare(`
-        SELECT id, memory_id, tenant_id, class_tag, claim_text,
-               estimate_value, estimate_unit, target_date,
-               actual_value, closure_state, closed_at, closure_note, created_at
-        FROM predictions WHERE id = ?
-      `).get(predictionId);
-            if (!row) {
-                throw new Error('Failed to reload saved prediction row');
-            }
-            savedRow = row;
-            // GDPR-light audit metadata: prediction_id + class_tag + flags only.
-            // No claim_text in metadata; the predictions table holds it canonically.
-            appendAuditEvent(db, {
-                tenantId,
-                actor,
-                op: 'predict_create',
-                targetId: String(predictionId),
-                metadata: {
-                    prediction_id: predictionId,
-                    class_tag: opts.classTag,
-                    has_estimate: opts.estimateValue !== undefined && opts.estimateValue !== null,
-                    target_date: opts.targetDate ?? null,
-                },
-            });
-        },
-    });
-    if (!savedRow) {
-        // Cannot reach here without the afterWrite throwing first; defensive.
-        throw new Error('savePrediction: afterWrite did not populate the row');
-    }
-    return rowToPrediction(savedRow);
-}
-/**
- * Close an existing open prediction. Updates the predictions row only;
- * the memory mirror is NOT mutated in v1 (predictions table is canonical).
- * J3 computes accuracy (clean vs regressed) from (estimateValue,
- * actualValue) at query time.
- */
-export function closePrediction(hippoRoot, tenantId, id, opts, actor = 'cli') {
-    assertTenantId('closePrediction', tenantId);
-    if (!VALID_CLOSURE_STATES.has(opts.closureState)) {
-        throw new BadRequestError(`closePrediction: closureState must be one of ${Array.from(VALID_CLOSURE_STATES).join('|')}; got ${opts.closureState}`);
-    }
-    const now = new Date().toISOString();
-    const db = openHippoDb(hippoRoot);
-    try {
-        db.exec('BEGIN IMMEDIATE');
-        try {
-            // Codex review finding 2026-05-26: WHERE clause requires
-            // closure_state='open' so duplicate close requests / retries against
-            // an already-closed prediction return a clear error instead of
-            // silently overwriting actual_value + emitting a duplicate
-            // predict_close audit row. Zero changed rows → caller decides
-            // whether it's a "not found" or "already closed" case based on the
-            // load-then-close pattern.
-            const updateResult = db.prepare(`
-        UPDATE predictions
-        SET actual_value = ?, closure_state = ?, closed_at = ?, closure_note = ?
-        WHERE id = ? AND tenant_id = ? AND closure_state = 'open'
-      `).run(opts.actualValue ?? null, opts.closureState, now, opts.closureNote ?? null, id, tenantId);
-            if (updateResult.changes === 0) {
-                // Distinguish "not found" from "already closed" so callers (CLI, HTTP)
-                // can surface the right error to the user.
-                // SAFETY: row shape matches the single `closure_state` column named
-                // in the SELECT above.
-                const existing = db.prepare(`
-          SELECT closure_state FROM predictions WHERE id = ? AND tenant_id = ?
-        `).get(id, tenantId);
-                if (!existing) {
-                    throw new NotFoundError(`closePrediction: prediction ${id} not found for tenant ${tenantId}`);
-                }
-                throw new BadRequestError(`closePrediction: prediction ${id} is already closed (state='${existing.closure_state}'); ` +
-                    `cannot re-close. Open predictions only.`);
-            }
-            // SAFETY: row's shape matches the columns named in the SELECT above.
-            const row = db.prepare(`
-        SELECT id, memory_id, tenant_id, class_tag, claim_text,
-               estimate_value, estimate_unit, target_date,
-               actual_value, closure_state, closed_at, closure_note, created_at
-        FROM predictions WHERE id = ? AND tenant_id = ?
-      `).get(id, tenantId);
-            if (!row) {
-                throw new NotFoundError(`closePrediction: prediction ${id} not found after UPDATE`);
-            }
-            appendAuditEvent(db, {
-                tenantId,
-                actor,
-                op: 'predict_close',
-                targetId: String(id),
-                metadata: {
-                    prediction_id: id,
-                    closure_state: opts.closureState,
-                    has_actual: opts.actualValue !== undefined && opts.actualValue !== null,
-                },
-            });
-            db.exec('COMMIT');
-            return rowToPrediction(row);
-        }
-        catch (e) {
-            try {
-                db.exec('ROLLBACK');
-            }
-            catch {
-                // Ignore rollback failures — the throw below is what matters.
-            }
-            throw e;
-        }
-    }
-    finally {
-        closeHippoDb(db);
-    }
-}
-export function loadPredictionById(hippoRoot, tenantId, id) {
-    assertTenantId('loadPredictionById', tenantId);
-    const db = openHippoDb(hippoRoot);
-    try {
-        // SAFETY: row's shape matches the columns named in the SELECT above.
-        const row = db.prepare(`
-      SELECT id, memory_id, tenant_id, class_tag, claim_text,
-             estimate_value, estimate_unit, target_date,
-             actual_value, closure_state, closed_at, closure_note, created_at
-      FROM predictions WHERE id = ? AND tenant_id = ?
-    `).get(id, tenantId);
-        return row ? rowToPrediction(row) : null;
-    }
-    finally {
-        closeHippoDb(db);
-    }
-}
-export function loadPredictionsByClass(hippoRoot, tenantId, classTag, opts = {}) {
-    assertTenantId('loadPredictionsByClass', tenantId);
-    const limit = opts.limit ?? 100;
-    const db = openHippoDb(hippoRoot);
-    try {
-        let rows;
-        if (opts.closureState) {
-            if (!VALID_CLOSURE_STATES.has(opts.closureState)) {
-                throw new BadRequestError(`loadPredictionsByClass: closureState must be one of ${Array.from(VALID_CLOSURE_STATES).join('|')}; got ${opts.closureState}`);
-            }
-            // SAFETY: rows' shape matches the columns named in the SELECT above.
-            rows = db.prepare(`
-        SELECT id, memory_id, tenant_id, class_tag, claim_text,
-               estimate_value, estimate_unit, target_date,
-               actual_value, closure_state, closed_at, closure_note, created_at
-        FROM predictions
-        WHERE tenant_id = ? AND class_tag = ? AND closure_state = ?
-        ORDER BY created_at DESC, id DESC
-        LIMIT ?
-      `).all(tenantId, classTag, opts.closureState, limit);
-        }
-        else {
-            // SAFETY: rows' shape matches the columns named in the SELECT above.
-            rows = db.prepare(`
-        SELECT id, memory_id, tenant_id, class_tag, claim_text,
-               estimate_value, estimate_unit, target_date,
-               actual_value, closure_state, closed_at, closure_note, created_at
-        FROM predictions
-        WHERE tenant_id = ? AND class_tag = ?
-        ORDER BY created_at DESC, id DESC
-        LIMIT ?
-      `).all(tenantId, classTag, limit);
-        }
-        return rows.map(rowToPrediction);
-    }
-    finally {
-        closeHippoDb(db);
-    }
-}
-/**
- * Compute base-rate stats for closed predictions in a class. Used by J3
- * reference-class / planning-fallacy detector. Direct application of
- * Lovallo-Kahneman (2003) inside-vs-outside view.
- *
- * Filter: closure_state='closed' AND estimate_value IS NOT NULL AND
- * actual_value IS NOT NULL. Excludes closed-unknown (no actual to
- * compare against) and open (not yet resolved).
- *
- * Audit-emit is BUILT IN here (single source of truth, no caller-site
- * drift risk). Plan-eng-critic round 1 HIGH recommendation: emit inside
- * helper, not at 3 call sites.
- */
-export function computePredictionBaserate(hippoRoot, tenantId, classTag, actor = 'cli', 
-/** v0.32 / J3.2 — when false, skip the predict_baserate audit emit. The
- *  J3.2 orchestrator (computePlanningFallacyOutput, below) calls this with
- *  emitAudit=false and emits its own `recall_autodebias_hint` audit row
- *  instead, so the predict_baserate channel stays scoped to deliberate
- *  CLI / HTTP / MCP predict-baserate calls and does NOT pollute on every
- *  recall containing a forward-claim phrase. Default true preserves the
- *  v1.13.0 J3 audit semantics for the 3 direct callers (cmdPredict
- *  baserate, /v1/predictions/stats route, hippo_predict_baserate MCP
- *  handler) — none of them pass this argument. */
-emitAudit = true) {
-    assertTenantId('computePredictionBaserate', tenantId);
-    if (!classTag)
-        throw new BadRequestError('computePredictionBaserate: classTag is required');
-    const db = openHippoDb(hippoRoot);
-    try {
-        // SAFETY: rows' shape matches the two columns named in the SELECT above.
-        const rows = db.prepare(`
-      SELECT estimate_value, actual_value
-      FROM predictions
-      WHERE tenant_id = ?
-        AND class_tag = ?
-        AND closure_state = 'closed'
-        AND estimate_value IS NOT NULL
-        AND actual_value IS NOT NULL
-    `).all(tenantId, classTag);
-        const nClosed = rows.length;
-        if (nClosed === 0) {
-            // Audit zero-result reads too — agents probing empty classes is
-            // a signal worth recording. Skipped when emitAudit=false (J3.2
-            // orchestrator path; its own recall_autodebias_hint audit fires
-            // only when nClosed > 0 anyway, so no signal is lost).
-            if (emitAudit) {
-                appendAuditEvent(db, {
-                    tenantId,
-                    actor,
-                    op: 'predict_baserate',
-                    targetId: classTag,
-                    metadata: { class_tag: classTag, n_closed: 0 },
-                });
-            }
-            return {
-                classTag,
-                nClosed: 0,
-                nRatioEligible: 0,
-                meanEstimate: null,
-                meanActual: null,
-                meanRatio: null,
-                p50Ratio: null,
-                mae: null,
-                summary: '',
-            };
-        }
-        const ratioEligible = rows.filter((r) => r.estimate_value > 0);
-        const nRatioEligible = ratioEligible.length;
-        const meanEstimate = rows.reduce((s, r) => s + r.estimate_value, 0) / nClosed;
-        const meanActual = rows.reduce((s, r) => s + r.actual_value, 0) / nClosed;
-        const mae = rows.reduce((s, r) => s + Math.abs(r.actual_value - r.estimate_value), 0) / nClosed;
-        let meanRatio = null;
-        let p50Ratio = null;
-        if (nRatioEligible > 0) {
-            const ratios = ratioEligible.map((r) => r.actual_value / r.estimate_value);
-            meanRatio = ratios.reduce((s, x) => s + x, 0) / nRatioEligible;
-            const sorted = ratios.slice().sort((a, b) => a - b);
-            p50Ratio = nRatioEligible % 2 === 1
-                ? sorted[(nRatioEligible - 1) / 2]
-                : (sorted[nRatioEligible / 2 - 1] + sorted[nRatioEligible / 2]) / 2;
-        }
-        const ratioPart = meanRatio !== null
-            ? `averaged ${meanRatio.toFixed(2)}x actual`
-            : 'no ratio-eligible rows (all estimates were 0)';
-        const summary = `Last ${nClosed} estimate${nClosed === 1 ? '' : 's'} in class ${classTag} ${ratioPart} (MAE ${mae.toFixed(2)}).`;
-        if (emitAudit) {
-            appendAuditEvent(db, {
-                tenantId,
-                actor,
-                op: 'predict_baserate',
-                targetId: classTag,
-                metadata: { class_tag: classTag, n_closed: nClosed },
-            });
-        }
-        return {
-            classTag,
-            nClosed,
-            nRatioEligible,
-            meanEstimate,
-            meanActual,
-            meanRatio,
-            p50Ratio,
-            mae,
-            summary,
-        };
-    }
-    finally {
-        closeHippoDb(db);
-    }
-}
-export function loadOpenPredictions(hippoRoot, tenantId, opts = {}) {
-    assertTenantId('loadOpenPredictions', tenantId);
-    const limit = opts.limit ?? 100;
-    const db = openHippoDb(hippoRoot);
-    try {
-        let rows;
-        if (opts.classTag) {
-            // SAFETY: rows' shape matches the columns named in the SELECT above.
-            rows = db.prepare(`
-        SELECT id, memory_id, tenant_id, class_tag, claim_text,
-               estimate_value, estimate_unit, target_date,
-               actual_value, closure_state, closed_at, closure_note, created_at
-        FROM predictions
-        WHERE tenant_id = ? AND class_tag = ? AND closure_state = 'open'
-        ORDER BY created_at DESC, id DESC
-        LIMIT ?
-      `).all(tenantId, opts.classTag, limit);
-        }
-        else {
-            // SAFETY: rows' shape matches the columns named in the SELECT above.
-            rows = db.prepare(`
-        SELECT id, memory_id, tenant_id, class_tag, claim_text,
-               estimate_value, estimate_unit, target_date,
-               actual_value, closure_state, closed_at, closure_note, created_at
-        FROM predictions
-        WHERE tenant_id = ? AND closure_state = 'open'
-        ORDER BY created_at DESC, id DESC
-        LIMIT ?
-      `).all(tenantId, limit);
-        }
-        return rows.map(rowToPrediction);
-    }
-    finally {
-        closeHippoDb(db);
-    }
-}
-/**
- * Resolve a query-token set to a unique best-matching class_tag for the
- * tenant. Scores by lower-cased token overlap; requires best score ≥ 1
- * AND strictly greater than the 2nd-best score.
- *
- * Indexed via idx_predictions_tenant_class (db.ts:1015) → O(log n) seek
- * plus a small DISTINCT scan over the per-tenant class-tag set.
- *
- * Scope behaviour (v1 design choice, independent-review-critic round 1
- * MED): class_tag selection is TENANT-GLOBAL, NOT scope-filtered against
- * the recall's opts.scope. The class_tag is an aggregator label across
- * historical predictions in the class, not a per-memory scope-bound
- * property. A no-scope recall CAN surface a class_tag from a privately-
- * scoped prediction's class in PlanningFallacyHint.classTag — by design,
- * because base-rate reasoning needs the full historical sample.
- * Implications:
- *   - The hint payload itself carries no memory content (only the aggregate
- *     summary string + numeric stats), so memory bodies do not leak.
- *   - The class_tag NAME is the side-channel. If sensitive labels are a
- *     concern, callers should either use opaque class names (e.g. hashes
- *     or numeric tokens) or set HIPPO_AUTODEBIAS=off.
- *   - tests/api-recall-autodebias.test.ts locks this with an explicit
- *     test asserting that scope-set predictions surface via no-scope
- *     recalls (so future "fix" attempts that scope-filter trip CI).
- */
-function resolveClassFromTokens(hippoRoot, tenantId, queryTokens) {
-    if (queryTokens.length === 0)
-        return { classTag: null, tiebreak: false };
-    const db = openHippoDb(hippoRoot);
-    try {
-        // SAFETY: rows' shape matches the single `class_tag` column named in the SELECT above.
-        const rows = db.prepare(`SELECT DISTINCT class_tag FROM predictions WHERE tenant_id = ?`).all(tenantId);
-        const querySet = new Set(queryTokens);
-        let bestScore = 0;
-        let bestClass = null;
-        let secondBest = 0;
-        for (const { class_tag } of rows) {
-            const classTokens = class_tag
-                .toLowerCase()
-                .split(/[-_\s]+/)
-                .filter((t) => t.length >= 3);
-            let score = 0;
-            for (const t of classTokens)
-                if (querySet.has(t))
-                    score++;
-            if (score > bestScore) {
-                secondBest = bestScore;
-                bestScore = score;
-                bestClass = class_tag;
-            }
-            else if (score === bestScore && score > 0) {
-                // Tie at current best — bump secondBest. Do NOT update bestClass
-                // (alphabetical tiebreak would pick wrong class on ambiguous query
-                // like "migration will take 3 days" between migration-effort vs
-                // migration-risk; silent on tie instead).
-                secondBest = score;
-            }
-            else if (score > secondBest) {
-                secondBest = score;
-            }
-        }
-        if (bestScore < 1)
-            return { classTag: null, tiebreak: false };
-        if (bestScore === secondBest)
-            return { classTag: null, tiebreak: true };
-        return { classTag: bestClass, tiebreak: false };
-    }
-    finally {
-        closeHippoDb(db);
-    }
-}
-/**
- * J3.2 orchestrator.
- *
- * Composes the forward-claim detector + class resolver + baserate compute,
- * with telemetry-grade audit emission at every decision point (success,
- * no-class-match, tiebreak).
- *
- * Returns `{}` (neither hint nor watching) on:
- *   - mode === 'off' (env-disabled; pays only the env read, skips regex)
- *   - empty queryText
- *   - no forward-claim regex match
- *   - resolved class has nClosed=0 (no historical data yet; silent)
- *
- * Returns `{ watching: ... }` on (v1.13.4 NEW — was silent null pre-1.13.4):
- *   - resolver returns no class (no overlap ≥ 1; emits no_class_match audit)
- *   - resolver returns tiebreak (≥2 classes tied at best; emits tiebreak audit)
- *
- * Returns `{ hint: ... }` on success: calls computePredictionBaserate(...,
- * emitAudit=false) so the predict_baserate audit channel stays scoped to
- * deliberate predict-baserate calls (the orchestrator's own recall_autodebias_hint
- * audit carries n_closed + mean_ratio in metadata so no telemetry is lost),
- * then emits recall_autodebias_hint audit + returns the hint.
- *
- * Latency budget (plan §Latency): ~50us regex-only on miss; ~750-850us
- * on full match+resolve+baserate path. Well under 50ms target.
- */
-export function computePlanningFallacyOutput(hippoRoot, tenantId, queryText, opts = {}) {
-    // Env read FIRST so AUTODEBIAS=off pays zero regex cost. Per-call read
-    // (rather than module-load cache) is deliberate: tests env-toggle this
-    // via process.env mutation without module reload.
-    const mode = opts.mode ?? (process.env.HIPPO_AUTODEBIAS === 'off' ? 'off' : 'regex');
-    if (mode === 'off')
-        return {};
-    if (!queryText)
-        return {};
-    const match = detectForwardClaim(queryText);
-    if (!match)
-        return {};
-    const actor = opts.actor ?? 'recall';
-    const resolution = resolveClassFromTokens(hippoRoot, tenantId, match.classQueryTokens);
-    if (resolution.tiebreak) {
-        // Telemetry: forward-claim detected, ≥2 classes tied at best overlap.
-        // v1.13.4: now ALSO returns a watching variant so the caller surface
-        // can render a "watching but no baserate (tiebreak)" line. Audit emission
-        // unchanged (the audit channel is the telemetry-grade source of truth).
-        const db = openHippoDb(hippoRoot);
-        try {
-            appendAuditEvent(db, {
-                tenantId,
-                actor,
-                op: 'recall_autodebias_hint_tiebreak',
-                targetId: match.phrase.slice(0, 100),
-                metadata: { detected_phrase: match.phrase, token_count: match.classQueryTokens.length },
-            });
-        }
-        finally {
-            closeHippoDb(db);
-        }
-        return {
-            watching: {
-                detectedPhrase: match.phrase,
-                reason: 'tiebreak',
-                suggestion: 'Multiple prediction classes tied on this query. Refine the query or rename overlapping classes to break the tie.',
-            },
-        };
-    }
-    if (!resolution.classTag) {
-        // Telemetry: forward-claim detected, no class scored ≥ 1.
-        // This is the channel that drives the embedding-fallback decision
-        // for J3.3 — high volume here = regex+token-overlap is missing
-        // legitimate forward-claims that have NO obvious class signal.
-        // v1.13.4: now ALSO returns a watching variant so the caller surface
-        // can render a "watching but no baserate (no class match)" line.
-        const db = openHippoDb(hippoRoot);
-        try {
-            appendAuditEvent(db, {
-                tenantId,
-                actor,
-                op: 'recall_autodebias_hint_no_class_match',
-                targetId: match.phrase.slice(0, 100),
-                metadata: { detected_phrase: match.phrase, token_count: match.classQueryTokens.length },
-            });
-        }
-        finally {
-            closeHippoDb(db);
-        }
-        return {
-            watching: {
-                detectedPhrase: match.phrase,
-                reason: 'no_class_match',
-                suggestion: 'No matching prediction class for this forward-claim. Tag your prediction with `hippo predict --class <name>` to start tracking this class.',
-            },
-        };
-    }
-    // emitAudit=false: avoid double-write to predict_baserate channel.
-    // The recall_autodebias_hint audit below carries n_closed + mean_ratio.
-    const baserate = computePredictionBaserate(hippoRoot, tenantId, resolution.classTag, actor, 
-    /*emitAudit=*/ false);
-    if (baserate.nClosed === 0)
-        return {}; // Silent — wait for closed data.
-    const db = openHippoDb(hippoRoot);
-    try {
-        appendAuditEvent(db, {
-            tenantId,
-            actor,
-            op: 'recall_autodebias_hint',
-            targetId: resolution.classTag,
-            metadata: {
-                class_tag: resolution.classTag,
-                detected_phrase: match.phrase,
-                n_closed: baserate.nClosed,
-                mean_ratio: baserate.meanRatio,
-            },
-        });
-    }
-    finally {
-        closeHippoDb(db);
-    }
-    return {
-        hint: {
-            classTag: resolution.classTag,
-            baserateSummary: baserate.summary,
-            source: 'j3.2-auto',
-            detectedPhrase: match.phrase,
-            nClosed: baserate.nClosed,
-            meanRatio: baserate.meanRatio,
-        },
-    };
-}
-//# sourceMappingURL=predictions.js.map