@mneme-ai/mcp 1.17.6 → 1.18.0

package/dist/tools/quant.js

@@ -1,26 +1,319 @@
 /**
- * Quant — Wall-Street-inspired analysis applied to engineering data.
- * Drawdown, alpha, Greeks, black-swan, moneyball, insider-trading detector.
+ * Quant — engineering analysis BORROWED from Wall Street, not Wall-Street
+ * jargon hidden behind cute names. Every quant tool ships a `jargon`
+ * dictionary that translates the finance term inline so a cold-start AI
+ * agent (or a human who never traded a stock) can understand the result
+ * without leaving the response.
+ *
+ * The 10 quant tools were the lowest-scoring on the v1.17 lint pass —
+ * v1.18 promotes them to FULL contract: WHEN, OUTPUT, EXAMPLES,
+ * PITFALLS, COMPOSE_WITH, JARGON. They now score ≥80/100 in
+ * `mneme.tool.lint`.
  */
 import { passthroughHandler } from "./_runtime.js";
-const quant = (name, description, triggers, cli, argMap = () => [], inputSchema = { type: "object", properties: {} }, followUp = [], wisdom = () => "Result returned — summarize key fields for the user.") => ({
-    name,
+const quant = (c) => ({
+    name: c.name,
     category: "quant",
-    description,
-    triggers,
-    inputSchema,
-    handler: passthroughHandler(cli, argMap, { wisdom, followUp, confidence: "medium" }),
+    description: c.description,
+    whenToUse: c.whenToUse,
+    triggers: c.triggers,
+    inputSchema: c.inputSchema ?? { type: "object", properties: {} },
+    outputSchema: c.outputSchema ?? {
+        type: "object",
+        properties: {
+            result: { type: "object", description: "CLI passthrough — see the named CLI command for the exact shape." },
+        },
+    },
+    examples: c.examples,
+    pitfalls: c.pitfalls,
+    composeWith: c.composeWith,
+    jargon: c.jargon,
+    handler: passthroughHandler(c.cli, c.argMap ?? (() => []), {
+        wisdom: c.wisdom ?? (() => "Result returned — summarize key fields for the user."),
+        followUp: c.followUp ?? [],
+        confidence: "medium",
+    }),
 });
 export const quantTools = [
-    quant("mneme.quant.drawdown", "Worst losing streaks — periods of pure firefighting. Use WHEN user asks: 'when were our worst weeks?', 'firefighting periods', 'drawdown'.", ["worst weeks of firefighting", "drawdown"], "drawdown"),
-    quant("mneme.quant.alpha", "Per-author 'alpha' — Kelly-criterion allocation across tech-debt items. Use WHEN user asks: 'who delivers alpha?', 'allocation of effort'.", ["who delivers alpha", "Kelly allocation"], "alpha", (a) => (a["items"] ? ["--items", String(a["items"])] : []), { type: "object", properties: { items: { type: "string", description: "JSON file path with items" } } }),
-    quant("mneme.quant.backtest", "Validate any binary predictor against historical outcomes. Use WHEN user asks: 'backtest this metric', 'validate predictor'.", ["backtest this metric", "validate predictor"], "backtest", (a) => (a["samples"] ? ["--samples", String(a["samples"])] : []), { type: "object", properties: { samples: { type: "string" } } }),
-    quant("mneme.quant.black_swan", "Tail-risk scan — rare but catastrophic file patterns. Use WHEN user asks: 'tail risks', 'black swan candidates'.", ["tail risks", "black swan candidates"], "black-swan"),
-    quant("mneme.quant.insider_trading", "Authors who repeatedly fix bugs they introduced themselves. Use WHEN user asks: 'who fixes their own bugs?', 'insider trading patterns'.", ["who fixes their own bugs", "insider patterns"], "insider-trading"),
-    quant("mneme.quant.moneyball", "Undervalued contributors — high impact, low LOC volume. Use WHEN user asks: 'who is undervalued?', 'moneyball contributors'.", ["undervalued contributors", "moneyball"], "moneyball", (a) => (a["topN"] ? ["--top", String(a["topN"])] : []), { type: "object", properties: { topN: { type: "number" } } }, ["mneme.people.influence"]),
-    quant("mneme.quant.greek", "Codebase Greeks (Δ Γ Θ) — sensitivity analysis across files. Use WHEN user asks: 'sensitivity analysis', 'codebase Greeks'.", ["sensitivity Greeks", "Δ Γ Θ"], "greek"),
-    quant("mneme.quant.correlation_matrix", "Hidden behavioral coupling between files (no static deps needed). Use WHEN user asks: 'which files move together?', 'behavioral coupling'.", ["files that change together", "behavioral coupling"], "correlation-matrix"),
-    quant("mneme.quant.implied_volatility", "Project chaos predicted from commit-message tone. Use WHEN user asks: 'how chaotic is this project?', 'implied volatility from tone'.", ["chaos from commit tone", "implied volatility"], "implied-volatility"),
-    quant("mneme.quant.tax_loss_harvest", "Dead-code candidates — delete to offset technical debt. Use WHEN user asks: 'what can we delete?', 'dead code', 'tax loss harvest'.", ["what can we delete", "dead code candidates"], "tax-loss-harvest", () => [], { type: "object", properties: {} }, ["mneme.insights.fossil", "mneme.insights.runaway"]),
+    quant({
+        name: "mneme.quant.drawdown",
+        description: "Find the worst losing streaks in the repo's history — periods when the team " +
+            "spent more time fixing regressions than shipping features. Returns the " +
+            "deepest valleys (most consecutive 'putting-out-fires' commits) with start/end " +
+            "dates and what the team was firefighting. Use WHEN the user asks 'when were " +
+            "our worst weeks?' or wants to understand historical incident clusters.",
+        whenToUse: "User wants to identify historical periods of pure firefighting / regression-fixing — useful for postmortems and capacity planning.",
+        triggers: ["worst weeks of firefighting", "drawdown", "when were we firefighting"],
+        cli: "drawdown",
+        jargon: {
+            drawdown: "Borrowed from finance — the size of the largest peak-to-valley drop in a portfolio. Here: the longest stretch of pure remediation work between two productive peaks.",
+        },
+        examples: [
+            {
+                userQuery: "When were our worst firefighting weeks?",
+                expectedOutput: "Returns the top 3-5 drawdown periods with: startDate, endDate, durationDays, dominantTopic (e.g., 'auth bugs'), and recoveredAt commit hash.",
+            },
+        ],
+        pitfalls: [
+            "Heuristic: depends on commit-message tone (revert/hotfix/fix words). A team that always says 'patch' or 'tweak' will be under-counted.",
+            "Doesn't account for incident severity — a 1-day P0 firedrill won't show up if commit count is small.",
+        ],
+        composeWith: ["mneme.insights.regret", "mneme.insights.premortem"],
+    }),
+    quant({
+        name: "mneme.quant.alpha",
+        description: "Per-author 'alpha' score — risk-adjusted impact-per-unit-effort. Measures " +
+            "which contributors deliver disproportionate value relative to commit count " +
+            "(e.g., a small fix that prevents a class of regressions counts MORE than 1000 " +
+            "lines of routine refactor). Use WHEN the user asks 'who delivers the most " +
+            "leverage?' or wants to challenge a 'most LOC = best engineer' assumption.",
+        whenToUse: "User wants to identify high-leverage contributors who deliver outsized impact relative to commit volume.",
+        triggers: ["who delivers alpha", "Kelly allocation", "high-leverage contributors"],
+        cli: "alpha",
+        jargon: {
+            alpha: "Wall Street: returns earned ABOVE the market baseline — i.e., skill, not just exposure. Here: an author's impact relative to what their commit volume alone would predict.",
+            "Kelly criterion": "A betting formula that sizes positions optimally given win probability + payoff. We use it to allocate 'effort budget' across tech-debt items based on each item's regret-history-derived expected payoff.",
+        },
+        examples: [
+            {
+                userQuery: "Who on the team delivers the most leverage per commit?",
+                args: { items: ".mneme/tech-debt.json" },
+                expectedOutput: "Returns each author with alpha (real impact - commit-count baseline), risk-adjusted alpha (alpha / volatility), and rank.",
+            },
+        ],
+        pitfalls: [
+            "Requires `items` (a JSON of tech-debt scope items) for the Kelly allocation portion — alpha-per-author works without it.",
+            "Doesn't capture invisible work (mentoring, code review, design docs) — those don't show up in commits.",
+        ],
+        composeWith: ["mneme.people.influence", "mneme.quant.moneyball"],
+        argMap: (a) => (a["items"] ? ["--items", String(a["items"])] : []),
+        inputSchema: {
+            type: "object",
+            properties: {
+                items: {
+                    type: "string",
+                    description: "Optional JSON file path with tech-debt scope items for Kelly-criterion allocation.",
+                },
+            },
+        },
+    }),
+    quant({
+        name: "mneme.quant.backtest",
+        description: "Validate any binary predictor (e.g., 'commits with > 5 reverts ⇒ regression') " +
+            "against historical outcomes. Returns precision / recall / F1 / Brier score / " +
+            "calibration curve so you can tell whether a heuristic actually predicts what " +
+            "it claims, or just pattern-matches noise. Use WHEN the user wants to validate " +
+            "a metric or rule before depending on it in CI.",
+        whenToUse: "User has a 'commits with property X tend to be problematic' hypothesis and wants to test it against the repo's actual outcomes.",
+        triggers: ["backtest this metric", "validate predictor", "does this rule actually work"],
+        cli: "backtest",
+        jargon: {
+            backtest: "Trading-system validation: replay a strategy through historical data to see how it would have performed. Here: replay a heuristic through past commits to measure its accuracy.",
+            "Brier score": "A calibration metric for probabilistic predictions — measures how close 0.7 confidence is to actually being right 70% of the time. Lower is better.",
+        },
+        examples: [
+            {
+                userQuery: "Does the 'files touching auth + payments in same commit ⇒ risky' rule actually work?",
+                args: { samples: ".mneme/predictor-samples.json" },
+                expectedOutput: "Returns precision, recall, F1, Brier score, and a confusion matrix on historical labeled commits.",
+            },
+        ],
+        pitfalls: [
+            "Garbage-in, garbage-out — your samples file must label historical outcomes correctly, otherwise the score reflects label noise.",
+            "Doesn't extrapolate — a predictor that worked on the last 6 months may not work on the next 6 months.",
+        ],
+        composeWith: ["mneme.insights.premortem", "mneme.audit.conscience"],
+        argMap: (a) => (a["samples"] ? ["--samples", String(a["samples"])] : []),
+        inputSchema: {
+            type: "object",
+            properties: {
+                samples: {
+                    type: "string",
+                    description: "JSON file path with labeled historical samples (predicted vs actual).",
+                },
+            },
+        },
+    }),
+    quant({
+        name: "mneme.quant.black_swan",
+        description: "Tail-risk scan — identify rare-but-catastrophic patterns in the repo " +
+            "(e.g., a single file that, when broken, has historically caused 10× the " +
+            "incident response of any other). Returns files ranked by tail-risk weight " +
+            "(probability × impact). Use WHEN the user asks 'what could blow up?' or " +
+            "wants the inverse-of-routine risk view.",
+        whenToUse: "User wants to find rare but catastrophic risk hotspots — files / areas where past failures had outsized blast radius.",
+        triggers: ["tail risks", "black swan candidates", "what could blow up"],
+        cli: "black-swan",
+        jargon: {
+            "black swan": "Nassim Taleb's term for a rare event that's nearly impossible to predict but has massive impact. Here: a file/area whose past failures had blast radius far above the median.",
+            "tail risk": "Risk concentrated in the unlikely-but-extreme outcomes (the 'tails' of a distribution), as opposed to the common-case middle.",
+        },
+        examples: [
+            {
+                userQuery: "What files in our repo are most likely to cause a major incident if they break?",
+                expectedOutput: "Top 5-10 files ranked by tail-risk score: each with past-incident count, average blast radius (files affected per incident), and recovery time.",
+            },
+        ],
+        pitfalls: [
+            "Past behavior ≠ future risk. A new feature module won't appear in tail-risk results no matter how risky it actually is.",
+            "Bias toward old, frequently-touched files — newly-rewritten modules look 'safe' until they aren't.",
+        ],
+        composeWith: ["mneme.audit.conscience", "mneme.insights.premortem"],
+    }),
+    quant({
+        name: "mneme.quant.insider_trading",
+        description: "Detect authors who repeatedly fix bugs they introduced themselves — the " +
+            "'creating their own work' anti-pattern. Returns each author with: bugs " +
+            "introduced, bugs they personally fixed, and the ratio (high = self-correcting, " +
+            "very high = possibly intentional churn). Use WHEN the user asks about " +
+            "engineering quality or wants to flag busy-but-not-productive contributors.",
+        whenToUse: "User wants to identify authors with a high self-introduced-bug-fix ratio (often a sign of rushed shipping or unclear specs).",
+        triggers: ["who fixes their own bugs", "insider patterns", "self-correcting commits"],
+        cli: "insider-trading",
+        jargon: {
+            "insider trading": "Wall Street: profiting from non-public knowledge of one's own actions. Here (re-purposed metaphor): an author 'profits' from their own future bugs by fixing what they introduced — busy work that LOOKS productive.",
+        },
+        examples: [
+            {
+                userQuery: "Is anyone on the team mostly fixing bugs they introduced?",
+                expectedOutput: "Returns each author with: bugsIntroduced, bugsFixed, selfFixRatio (selfBugsFixed / bugsIntroduced), and recentExamples (commit pairs).",
+            },
+        ],
+        pitfalls: [
+            "A high ratio isn't always bad — senior engineers fix their own subtle bugs faster than junior ones could. Read in context.",
+            "Depends on linking 'introduce' commit ↔ 'fix' commit, which uses heuristic overlap (file + nearby lines + temporal proximity). False positives possible.",
+        ],
+        composeWith: ["mneme.people.atrophy", "mneme.insights.regret"],
+    }),
+    quant({
+        name: "mneme.quant.moneyball",
+        description: "Surface undervalued contributors — high impact, low LOC volume. The opposite " +
+            "of LOC-counting culture: returns engineers whose small surgical changes prevent " +
+            "regressions, fix root causes, or improve architecture per-commit far more than " +
+            "the team median. Use WHEN the user asks 'who is undervalued?' or wants " +
+            "promotion / retention candidates not visible in commit volume.",
+        whenToUse: "User wants to identify low-LOC, high-impact contributors hidden by volume-based metrics — promotion / retention signal.",
+        triggers: ["undervalued contributors", "moneyball", "high impact low LOC"],
+        cli: "moneyball",
+        jargon: {
+            moneyball: "From Michael Lewis's book about the Oakland A's baseball team finding undervalued players via stats nobody else looked at. Here: finding engineers whose value isn't captured by LOC or commit count.",
+        },
+        examples: [
+            {
+                userQuery: "Who on the team delivers the most value but writes the fewest lines?",
+                args: { topN: 5 },
+                expectedOutput: "Returns top-5 'moneyball' authors with: impactScore, locTotal, impactPerLoc ratio, signature contributions (commit hashes).",
+            },
+        ],
+        pitfalls: [
+            "Impact is heuristic — derived from regret avoidance, file criticality, and commit reach. Not the ground truth on engineering value.",
+            "Underweights work that's high-volume by necessity (e.g., test infrastructure, refactors).",
+        ],
+        composeWith: ["mneme.people.influence", "mneme.quant.alpha"],
+        argMap: (a) => (a["topN"] ? ["--top", String(a["topN"])] : []),
+        inputSchema: { type: "object", properties: { topN: { type: "number", description: "How many candidates to return. Default 10." } } },
+    }),
+    quant({
+        name: "mneme.quant.greek",
+        description: "Codebase 'Greeks' (Δ delta, Γ gamma, Θ theta) — sensitivity analysis across " +
+            "files. Δ = how much each file 'moves' (changes) per unit time. Γ = how the " +
+            "rate of change is changing (acceleration). Θ = decay — how fast knowledge " +
+            "ages out. Use WHEN the user wants to understand WHICH parts of the codebase " +
+            "are stable, accelerating, or atrophying.",
+        whenToUse: "User wants per-file sensitivity metrics: which files are changing fast, accelerating, or atrophying.",
+        triggers: ["sensitivity Greeks", "Δ Γ Θ", "which files are changing fast"],
+        cli: "greek",
+        jargon: {
+            delta: "Options trading: Δ = how much an option's price moves per $1 move in the underlying. Here: how much a file changes per unit time (commits/week or LOC/week).",
+            gamma: "Options trading: Γ = the RATE at which Δ changes — second-derivative sensitivity. Here: file-change acceleration. Positive Γ = a file getting noisier; negative Γ = settling down.",
+            theta: "Options trading: Θ = time-decay — how much value an option loses per day, all else equal. Here: knowledge atrophy — how fast a file's expertise ages out (Ebbinghaus curve).",
+        },
+        examples: [
+            {
+                userQuery: "Which files are changing fastest right now and which are atrophying?",
+                expectedOutput: "Returns per-file: delta (commits/week), gamma (acceleration), theta (knowledge half-life in weeks). Sortable by any.",
+            },
+        ],
+        pitfalls: [
+            "Greeks are MOMENTUM indicators — they tell you direction, not whether the change is good or bad.",
+            "Theta uses a default 90-day half-life; configurable per repo via .mneme/config.",
+        ],
+        composeWith: ["mneme.people.atrophy", "mneme.quant.implied_volatility"],
+    }),
+    quant({
+        name: "mneme.quant.correlation_matrix",
+        description: "Find hidden BEHAVIORAL coupling between files — pairs that tend to change " +
+            "together even though they have no static dependency (no import, no shared " +
+            "type). Returns a ranked list of file pairs with co-change frequency + " +
+            "correlation strength. Use WHEN the user wants to find architectural debt " +
+            "static analysis can't see.",
+        whenToUse: "User wants to find file pairs that move together in commits despite having no compile-time dependency — hidden coupling.",
+        triggers: ["files that change together", "behavioral coupling", "hidden dependencies"],
+        cli: "correlation-matrix",
+        jargon: {
+            "correlation matrix": "Statistics: a table of pairwise correlations between variables. Here: between files, where 'observation' = 'commit'. High correlation between files A and B = they tend to change in the same commit.",
+        },
+        examples: [
+            {
+                userQuery: "Which files in our repo always seem to change together?",
+                expectedOutput: "Top file pairs by correlation: { fileA, fileB, coChangeCount, correlation (0-1), staticallyDependent: boolean }.",
+            },
+        ],
+        pitfalls: [
+            "Correlation isn't causation — two files may co-change because of a 3rd file (e.g., they both depend on a shared schema).",
+            "Sensitive to commit granularity — repos that batch many concerns per commit will have inflated correlations.",
+        ],
+        composeWith: ["mneme.insights.cluster", "mneme.insights.network"],
+    }),
+    quant({
+        name: "mneme.quant.implied_volatility",
+        description: "Estimate project chaos from commit-message TONE (urgency words, frustration " +
+            "words, 'temp / hack / fixme' frequency). Returns a daily volatility series — " +
+            "high = stressful periods (firefighting / deadline crunch), low = calm flow. " +
+            "Use WHEN the user wants a leading indicator of team stress that doesn't " +
+            "depend on incident reports.",
+        whenToUse: "User wants a tone-derived stress signal independent of incident tickets — daily volatility from commit messages.",
+        triggers: ["chaos from commit tone", "implied volatility", "team stress signal"],
+        cli: "implied-volatility",
+        jargon: {
+            "implied volatility": "Options trading: market's forward-looking estimate of how much a price will move, derived from option prices. Here: forward-looking chaos estimate derived from how the team is TALKING in commits, not what they're shipping.",
+        },
+        examples: [
+            {
+                userQuery: "When was our team most stressed based on how we wrote commits?",
+                expectedOutput: "Returns a daily series: { date, volatility (0-1), topUrgencyWords, sampleCommitSubject } with peaks and rolling 7-day average.",
+            },
+        ],
+        pitfalls: [
+            "Tone-based — sarcasm, dry humor, and culture-specific phrasing can fool it.",
+            "Lags the actual stressful event by 1-3 days (people commit AFTER firefighting, not during).",
+        ],
+        composeWith: ["mneme.quant.drawdown", "mneme.quality.heartbeat"],
+    }),
+    quant({
+        name: "mneme.quant.tax_loss_harvest",
+        description: "Surface dead-code candidates — code that hasn't been touched in N months, " +
+            "is reachable but never imported, or whose tests don't actually exercise it. " +
+            "Returns deletion candidates ranked by 'safe-to-delete' confidence + " +
+            "estimated LOC savings. Use WHEN the user asks 'what can we delete?' or " +
+            "wants to offset technical debt by removing surface area.",
+        whenToUse: "User wants concrete dead-code deletion candidates with safety ratings — to reduce surface area / pay down debt.",
+        triggers: ["what can we delete", "dead code candidates", "tax loss harvest"],
+        cli: "tax-loss-harvest",
+        jargon: {
+            "tax loss harvesting": "Investing: deliberately selling losing positions to offset capital gains tax. Here (re-purposed): deliberately deleting dead/legacy code to offset 'cognitive tax' of carrying it.",
+        },
+        examples: [
+            {
+                userQuery: "What chunks of our codebase can we safely delete?",
+                expectedOutput: "Top deletion candidates: { path, lastTouched, importedBy, testedBy, safeDeleteScore (0-1), locSaved }.",
+            },
+        ],
+        pitfalls: [
+            "'Reachable but never imported' is heuristic — dynamic imports, plugin systems, and runtime reflection can fool it.",
+            "Always run the test suite + grep for path strings before deleting; the safeDeleteScore is a recommendation, not a guarantee.",
+        ],
+        composeWith: ["mneme.insights.fossil", "mneme.insights.runaway"],
+    }),
 ];
 //# sourceMappingURL=quant.js.map

package/dist/tools/quant.js.map

@@ -1 +1 @@
-{"version":3,"file":"quant.js","sourceRoot":"","sources":["../../src/tools/quant.ts"],"names":[],"mappings":"AAAA;;;GAGG;AAEH,OAAO,EAAE,kBAAkB,EAAE,MAAM,eAAe,CAAC;AAGnD,MAAM,KAAK,GAAG,CACZ,IAAY,EACZ,WAAmB,EACnB,QAAkB,EAClB,GAAW,EACX,SAAmD,GAAG,EAAE,CAAC,EAAE,EAC3D,cAAwC,EAAE,IAAI,EAAE,QAAQ,EAAE,UAAU,EAAE,EAAE,EAAE,EAC1E,WAAqB,EAAE,EACvB,SAAiC,GAAG,EAAE,CAAC,sDAAsD,EAClF,EAAE,CAAC,CAAC;IACf,IAAI;IACJ,QAAQ,EAAE,OAAO;IACjB,WAAW;IACX,QAAQ;IACR,WAAW;IACX,OAAO,EAAE,kBAAkB,CAAC,GAAG,EAAE,MAAM,EAAE,EAAE,MAAM,EAAE,QAAQ,EAAE,UAAU,EAAE,QAAQ,EAAE,CAAC;CACrF,CAAC,CAAC;AAEH,MAAM,CAAC,MAAM,UAAU,GAAgB;IACrC,KAAK,CACH,sBAAsB,EACtB,4IAA4I,EAC5I,CAAC,6BAA6B,EAAE,UAAU,CAAC,EAC3C,UAAU,CACX;IACD,KAAK,CACH,mBAAmB,EACnB,4IAA4I,EAC5I,CAAC,oBAAoB,EAAE,kBAAkB,CAAC,EAC1C,OAAO,EACP,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,CAAC,OAAO,CAAC,CAAC,CAAC,CAAC,CAAC,SAAS,EAAE,MAAM,CAAC,CAAC,CAAC,OAAO,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,EAAE,CAAC,EAC1D,EAAE,IAAI,EAAE,QAAQ,EAAE,UAAU,EAAE,EAAE,KAAK,EAAE,EAAE,IAAI,EAAE,QAAQ,EAAE,WAAW,EAAE,2BAA2B,EAAE,EAAE,EAAE,CACxG;IACD,KAAK,CACH,sBAAsB,EACtB,8HAA8H,EAC9H,CAAC,sBAAsB,EAAE,oBAAoB,CAAC,EAC9C,UAAU,EACV,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,CAAC,SAAS,CAAC,CAAC,CAAC,CAAC,CAAC,WAAW,EAAE,MAAM,CAAC,CAAC,CAAC,SAAS,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,EAAE,CAAC,EAChE,EAAE,IAAI,EAAE,QAAQ,EAAE,UAAU,EAAE,EAAE,OAAO,EAAE,EAAE,IAAI,EAAE,QAAQ,EAAE,EAAE,EAAE,CAChE;IACD,KAAK,CACH,wBAAwB,EACxB,kHAAkH,EAClH,CAAC,YAAY,EAAE,uBAAuB,CAAC,EACvC,YAAY,CACb;IACD,KAAK,CACH,6BAA6B,EAC7B,0IAA0I,EAC1I,CAAC,0BAA0B,EAAE,kBAAkB,CAAC,EAChD,iBAAiB,CAClB;IACD,KAAK,CACH,uBAAuB,EACvB,8HAA8H,EAC9H,CAAC,0BAA0B,EAAE,WAAW,CAAC,EACzC,WAAW,EACX,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,CAAC,MAAM,CAAC,CAAC,CAAC,CAAC,CAAC,OAAO,EAAE,MAAM,CAAC,CAAC,CAAC,MAAM,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,EAAE,CAAC,EACtD,EAAE,IAAI,EAAE,QAAQ,EAAE,UAAU,EAAE,EAAE,IAAI,EAAE,EAAE,IAAI,EAAE,QAAQ,EAAE,EAAE,EAAE,EAC5D,CAAC,wBAAwB,CAAC,CAC3B;IACD,KAAK,CACH,mBAAmB,EACnB,6HAA6H,EAC7H,CAAC,oBAAoB,EAAE,OAAO,CAAC,EAC/B,OAAO,CACR;IACD,KAAK,CACH,gCAAgC,EAChC,4IAA4I,EAC5I,CAAC,4BAA4B,EAAE,qBAAqB,CAAC,EACrD,oBAAoB,CACrB;IACD,KAAK,CACH,gCAAgC,EAChC,uIAAuI,EACvI,CAAC,wBAAwB,EAAE,oBAAoB,CAAC,EAChD,oBAAoB,CACrB;IACD,KAAK,CACH,8BAA8B,EAC9B,qIAAqI,EACrI,CAAC,oBAAoB,EAAE,sBAAsB,CAAC,EAC9C,kBAAkB,EAClB,GAAG,EAAE,CAAC,EAAE,EACR,EAAE,IAAI,EAAE,QAAQ,EAAE,UAAU,EAAE,EAAE,EAAE,EAClC,CAAC,uBAAuB,EAAE,wBAAwB,CAAC,CACpD;CACF,CAAC"}
+{"version":3,"file":"quant.js","sourceRoot":"","sources":["../../src/tools/quant.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;GAWG;AAEH,OAAO,EAAE,kBAAkB,EAAE,MAAM,eAAe,CAAC;AAoBnD,MAAM,KAAK,GAAG,CAAC,CAAgB,EAAa,EAAE,CAAC,CAAC;IAC9C,IAAI,EAAE,CAAC,CAAC,IAAI;IACZ,QAAQ,EAAE,OAAO;IACjB,WAAW,EAAE,CAAC,CAAC,WAAW;IAC1B,SAAS,EAAE,CAAC,CAAC,SAAS;IACtB,QAAQ,EAAE,CAAC,CAAC,QAAQ;IACpB,WAAW,EAAE,CAAC,CAAC,WAAW,IAAI,EAAE,IAAI,EAAE,QAAQ,EAAE,UAAU,EAAE,EAAE,EAAE;IAChE,YAAY,EAAE,CAAC,CAAC,YAAY,IAAI;QAC9B,IAAI,EAAE,QAAQ;QACd,UAAU,EAAE;YACV,MAAM,EAAE,EAAE,IAAI,EAAE,QAAQ,EAAE,WAAW,EAAE,kEAAkE,EAAE;SAC5G;KACF;IACD,QAAQ,EAAE,CAAC,CAAC,QAAQ;IACpB,QAAQ,EAAE,CAAC,CAAC,QAAQ;IACpB,WAAW,EAAE,CAAC,CAAC,WAAW;IAC1B,MAAM,EAAE,CAAC,CAAC,MAAM;IAChB,OAAO,EAAE,kBAAkB,CAAC,CAAC,CAAC,GAAG,EAAE,CAAC,CAAC,MAAM,IAAI,CAAC,GAAG,EAAE,CAAC,EAAE,CAAC,EAAE;QACzD,MAAM,EAAE,CAAC,CAAC,MAAM,IAAI,CAAC,GAAG,EAAE,CAAC,sDAAsD,CAAC;QAClF,QAAQ,EAAE,CAAC,CAAC,QAAQ,IAAI,EAAE;QAC1B,UAAU,EAAE,QAAQ;KACrB,CAAC;CACH,CAAC,CAAC;AAEH,MAAM,CAAC,MAAM,UAAU,GAAgB;IACrC,KAAK,CAAC;QACJ,IAAI,EAAE,sBAAsB;QAC5B,WAAW,EACT,8EAA8E;YAC9E,yEAAyE;YACzE,gFAAgF;YAChF,8EAA8E;YAC9E,wEAAwE;QAC1E,SAAS,EACP,oIAAoI;QACtI,QAAQ,EAAE,CAAC,6BAA6B,EAAE,UAAU,EAAE,2BAA2B,CAAC;QAClF,GAAG,EAAE,UAAU;QACf,MAAM,EAAE;YACN,QAAQ,EACN,sKAAsK;SACzK;QACD,QAAQ,EAAE;YACR;gBACE,SAAS,EAAE,yCAAyC;gBACpD,cAAc,EACZ,8IAA8I;aACjJ;SACF;QACD,QAAQ,EAAE;YACR,wIAAwI;YACxI,sGAAsG;SACvG;QACD,WAAW,EAAE,CAAC,uBAAuB,EAAE,0BAA0B,CAAC;KACnE,CAAC;IACF,KAAK,CAAC;QACJ,IAAI,EAAE,mBAAmB;QACzB,WAAW,EACT,4EAA4E;YAC5E,6EAA6E;YAC7E,gFAAgF;YAChF,4EAA4E;YAC5E,2EAA2E;QAC7E,SAAS,EACP,0GAA0G;QAC5G,QAAQ,EAAE,CAAC,oBAAoB,EAAE,kBAAkB,EAAE,4BAA4B,CAAC;QAClF,GAAG,EAAE,OAAO;QACZ,MAAM,EAAE;YACN,KAAK,EACH,4KAA4K;YAC9K,iBAAiB,EACf,4MAA4M;SAC/M;QACD,QAAQ,EAAE;YACR;gBACE,SAAS,EAAE,wDAAwD;gBACnE,IAAI,EAAE,EAAE,KAAK,EAAE,uBAAuB,EAAE;gBACxC,cAAc,EACZ,2HAA2H;aAC9H;SACF;QACD,QAAQ,EAAE;YACR,0HAA0H;YAC1H,wGAAwG;SACzG;QACD,WAAW,EAAE,CAAC,wBAAwB,EAAE,uBAAuB,CAAC;QAChE,MAAM,EAAE,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,CAAC,OAAO,CAAC,CAAC,CAAC,CAAC,CAAC,SAAS,EAAE,MAAM,CAAC,CAAC,CAAC,OAAO,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,EAAE,CAAC;QAClE,WAAW,EAAE;YACX,IAAI,EAAE,QAAQ;YACd,UAAU,EAAE;gBACV,KAAK,EAAE;oBACL,IAAI,EAAE,QAAQ;oBACd,WAAW,EAAE,oFAAoF;iBAClG;aACF;SACF;KACF,CAAC;IACF,KAAK,CAAC;QACJ,IAAI,EAAE,sBAAsB;QAC5B,WAAW,EACT,gFAAgF;YAChF,+EAA+E;YAC/E,+EAA+E;YAC/E,gFAAgF;YAChF,gDAAgD;QAClD,SAAS,EACP,iIAAiI;QACnI,QAAQ,EAAE,CAAC,sBAAsB,EAAE,oBAAoB,EAAE,8BAA8B,CAAC;QACxF,GAAG,EAAE,UAAU;QACf,MAAM,EAAE;YACN,QAAQ,EACN,iLAAiL;YACnL,aAAa,EACX,qJAAqJ;SACxJ;QACD,QAAQ,EAAE;YACR;gBACE,SAAS,EAAE,sFAAsF;gBACjG,IAAI,EAAE,EAAE,OAAO,EAAE,+BAA+B,EAAE;gBAClD,cAAc,EACZ,mGAAmG;aACtG;SACF;QACD,QAAQ,EAAE;YACR,iIAAiI;YACjI,uGAAuG;SACxG;QACD,WAAW,EAAE,CAAC,0BAA0B,EAAE,wBAAwB,CAAC;QACnE,MAAM,EAAE,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,CAAC,SAAS,CAAC,CAAC,CAAC,CAAC,CAAC,WAAW,EAAE,MAAM,CAAC,CAAC,CAAC,SAAS,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,EAAE,CAAC;QACxE,WAAW,EAAE;YACX,IAAI,EAAE,QAAQ;YACd,UAAU,EAAE;gBACV,OAAO,EAAE;oBACP,IAAI,EAAE,QAAQ;oBACd,WAAW,EAAE,uEAAuE;iBACrF;aACF;SACF;KACF,CAAC;IACF,KAAK,CAAC;QACJ,IAAI,EAAE,wBAAwB;QAC9B,WAAW,EACT,uEAAuE;YACvE,0EAA0E;YAC1E,4EAA4E;YAC5E,0EAA0E;YAC1E,yCAAyC;QAC3C,SAAS,EACP,uHAAuH;QACzH,QAAQ,EAAE,CAAC,YAAY,EAAE,uBAAuB,EAAE,oBAAoB,CAAC;QACvE,GAAG,EAAE,YAAY;QACjB,MAAM,EAAE;YACN,YAAY,EACV,+KAA+K;YACjL,WAAW,EACT,+HAA+H;SAClI;QACD,QAAQ,EAAE;YACR;gBACE,SAAS,EAAE,iFAAiF;gBAC5F,cAAc,EACZ,iJAAiJ;aACpJ;SACF;QACD,QAAQ,EAAE;YACR,yHAAyH;YACzH,oGAAoG;SACrG;QACD,WAAW,EAAE,CAAC,wBAAwB,EAAE,0BAA0B,CAAC;KACpE,CAAC;IACF,KAAK,CAAC;QACJ,IAAI,EAAE,6BAA6B;QACnC,WAAW,EACT,0EAA0E;YAC1E,yEAAyE;YACzE,iFAAiF;YACjF,wEAAwE;YACxE,4EAA4E;QAC9E,SAAS,EACP,8HAA8H;QAChI,QAAQ,EAAE,CAAC,0BAA0B,EAAE,kBAAkB,EAAE,yBAAyB,CAAC;QACrF,GAAG,EAAE,iBAAiB;QACtB,MAAM,EAAE;YACN,iBAAiB,EACf,sNAAsN;SACzN;QACD,QAAQ,EAAE;YACR;gBACE,SAAS,EAAE,2DAA2D;gBACtE,cAAc,EACZ,wIAAwI;aAC3I;SACF;QACD,QAAQ,EAAE;YACR,4HAA4H;YAC5H,0JAA0J;SAC3J;QACD,WAAW,EAAE,CAAC,sBAAsB,EAAE,uBAAuB,CAAC;KAC/D,CAAC;IACF,KAAK,CAAC;QACJ,IAAI,EAAE,uBAAuB;QAC7B,WAAW,EACT,+EAA+E;YAC/E,kFAAkF;YAClF,iFAAiF;YACjF,yEAAyE;YACzE,gEAAgE;QAClE,SAAS,EACP,yHAAyH;QAC3H,QAAQ,EAAE,CAAC,0BAA0B,EAAE,WAAW,EAAE,qBAAqB,CAAC;QAC1E,GAAG,EAAE,WAAW;QAChB,MAAM,EAAE;YACN,SAAS,EACP,uMAAuM;SAC1M;QACD,QAAQ,EAAE;YACR;gBACE,SAAS,EAAE,sEAAsE;gBACjF,IAAI,EAAE,EAAE,IAAI,EAAE,CAAC,EAAE;gBACjB,cAAc,EACZ,6HAA6H;aAChI;SACF;QACD,QAAQ,EAAE;YACR,qIAAqI;YACrI,2FAA2F;SAC5F;QACD,WAAW,EAAE,CAAC,wBAAwB,EAAE,mBAAmB,CAAC;QAC5D,MAAM,EAAE,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,CAAC,MAAM,CAAC,CAAC,CAAC,CAAC,CAAC,OAAO,EAAE,MAAM,CAAC,CAAC,CAAC,MAAM,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,EAAE,CAAC;QAC9D,WAAW,EAAE,EAAE,IAAI,EAAE,QAAQ,EAAE,UAAU,EAAE,EAAE,IAAI,EAAE,EAAE,IAAI,EAAE,QAAQ,EAAE,WAAW,EAAE,4CAA4C,EAAE,EAAE,EAAE;KACrI,CAAC;IACF,KAAK,CAAC;QACJ,IAAI,EAAE,mBAAmB;QACzB,WAAW,EACT,8EAA8E;YAC9E,6EAA6E;YAC7E,4EAA4E;YAC5E,8EAA8E;YAC9E,0CAA0C;QAC5C,SAAS,EACP,sGAAsG;QACxG,QAAQ,EAAE,CAAC,oBAAoB,EAAE,OAAO,EAAE,+BAA+B,CAAC;QAC1E,GAAG,EAAE,OAAO;QACZ,MAAM,EAAE;YACN,KAAK,EACH,8JAA8J;YAChK,KAAK,EACH,oLAAoL;YACtL,KAAK,EACH,8KAA8K;SACjL;QACD,QAAQ,EAAE;YACR;gBACE,SAAS,EAAE,sEAAsE;gBACjF,cAAc,EACZ,sHAAsH;aACzH;SACF;QACD,QAAQ,EAAE;YACR,kGAAkG;YAClG,iFAAiF;SAClF;QACD,WAAW,EAAE,CAAC,sBAAsB,EAAE,gCAAgC,CAAC;KACxE,CAAC;IACF,KAAK,CAAC;QACJ,IAAI,EAAE,gCAAgC;QACtC,WAAW,EACT,4EAA4E;YAC5E,4EAA4E;YAC5E,wEAAwE;YACxE,2EAA2E;YAC3E,4BAA4B;QAC9B,SAAS,EACP,0HAA0H;QAC5H,QAAQ,EAAE,CAAC,4BAA4B,EAAE,qBAAqB,EAAE,qBAAqB,CAAC;QACtF,GAAG,EAAE,oBAAoB;QACzB,MAAM,EAAE;YACN,oBAAoB,EAClB,uMAAuM;SAC1M;QACD,QAAQ,EAAE;YACR;gBACE,SAAS,EAAE,yDAAyD;gBACpE,cAAc,EACZ,kHAAkH;aACrH;SACF;QACD,QAAQ,EAAE;YACR,0HAA0H;YAC1H,8GAA8G;SAC/G;QACD,WAAW,EAAE,CAAC,wBAAwB,EAAE,wBAAwB,CAAC;KAClE,CAAC;IACF,KAAK,CAAC;QACJ,IAAI,EAAE,gCAAgC;QACtC,WAAW,EACT,8EAA8E;YAC9E,+EAA+E;YAC/E,8EAA8E;YAC9E,0EAA0E;YAC1E,6BAA6B;QAC/B,SAAS,EACP,kHAAkH;QACpH,QAAQ,EAAE,CAAC,wBAAwB,EAAE,oBAAoB,EAAE,oBAAoB,CAAC;QAChF,GAAG,EAAE,oBAAoB;QACzB,MAAM,EAAE;YACN,oBAAoB,EAClB,gOAAgO;SACnO;QACD,QAAQ,EAAE;YACR;gBACE,SAAS,EAAE,gEAAgE;gBAC3E,cAAc,EACZ,gIAAgI;aACnI;SACF;QACD,QAAQ,EAAE;YACR,6EAA6E;YAC7E,6FAA6F;SAC9F;QACD,WAAW,EAAE,CAAC,sBAAsB,EAAE,yBAAyB,CAAC;KACjE,CAAC;IACF,KAAK,CAAC;QACJ,IAAI,EAAE,8BAA8B;QACpC,WAAW,EACT,4EAA4E;YAC5E,8EAA8E;YAC9E,sEAAsE;YACtE,yEAAyE;YACzE,0DAA0D;QAC5D,SAAS,EACP,iHAAiH;QACnH,QAAQ,EAAE,CAAC,oBAAoB,EAAE,sBAAsB,EAAE,kBAAkB,CAAC;QAC5E,GAAG,EAAE,kBAAkB;QACvB,MAAM,EAAE;YACN,qBAAqB,EACnB,oLAAoL;SACvL;QACD,QAAQ,EAAE;YACR;gBACE,SAAS,EAAE,mDAAmD;gBAC9D,cAAc,EACZ,wGAAwG;aAC3G;SACF;QACD,QAAQ,EAAE;YACR,oHAAoH;YACpH,8HAA8H;SAC/H;QACD,WAAW,EAAE,CAAC,uBAAuB,EAAE,wBAAwB,CAAC;KACjE,CAAC;CACH,CAAC"}

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@mneme-ai/mcp",
-  "version": "1.17.6",
+  "version": "1.18.0",
   "description": "MCP server that exposes Mneme to Claude Code, Cursor, Continue, and other AI clients",
   "type": "module",
   "main": "./dist/index.js",
@@ -37,8 +37,8 @@
     "node": ">=22.13.0 <25.0.0"
   },
   "dependencies": {
-    "@mneme-ai/core": "1.17.6",
-    "@mneme-ai/embeddings": "1.17.6",
+    "@mneme-ai/core": "1.18.0",
+    "@mneme-ai/embeddings": "1.18.0",
     "@modelcontextprotocol/sdk": "^1.0.4"
   }
 }