daftari 1.9.1 → 1.11.0

package/dist/tools/themes.js

@@ -0,0 +1,425 @@
+// vault_themes — thematic clustering over document-pooled embeddings.
+//
+// Cluster DOCUMENTS, not chunks: for each document we mean-pool its chunk
+// embeddings into one vector, L2-normalise it, and cluster the resulting
+// ~N-doc set. ~3.5k document vectors instead of ~44k chunk vectors makes
+// every algorithm cheap (silhouette becomes tractable on the full set; no
+// sampling needed at this scale).
+//
+// v1 constraints (locked):
+//   - One-doc-one-theme. Each document lives in exactly one cluster.
+//   - Heuristic labels from TF-IDF over titles + tags. No LLM call.
+//   - No new storage — the tool is read-only against the existing
+//     embeddings / chunks / documents tables.
+//   - Default k-sweep over {10, 15, 20, 25}; an explicit `k` argument
+//     skips the sweep.
+//   - Deterministic via a fixed seed for the k-means RNG.
+import { canRead } from "../access/rbac.js";
+import { err, ok } from "../frontmatter/types.js";
+import { getProvider } from "../search/vector.js";
+import { blobToEmbedding } from "../storage/index-db.js";
+import { clusterCoherence, kmeans, meanPoolL2, pickK, seededRng, selectSecondaryMemberships, } from "../themes/clustering.js";
+import { ensureIndexReady, openIndexForActiveProvider } from "./search.js";
+// Fixed seed for the k-means RNG. The whole point of vault_themes is that
+// the same vault produces the same themes — a random seed would give the
+// user different themes every run and erode trust in the output.
+const THEMES_RNG_SEED = 0x7da17f1; // arbitrary; constant is what matters
+// Default k-sweep range (from the issue dialogue). Each is clamped to the
+// number of clusterable documents inside pickK.
+const DEFAULT_K_CANDIDATES = [10, 15, 20, 25];
+// k-means iteration cap. Documents on the unit sphere converge fast in
+// practice; this is a safety bound for pathological inputs.
+const KMEANS_MAX_ITER = 50;
+// Representative-doc count returned per theme. 5 is enough to be useful in
+// the UI without dragging in marginal members.
+const REPRESENTATIVE_DOCS_PER_THEME = 5;
+// Tag-frequency cutoff per theme. Five most-common tags is enough signal
+// without burying the user in noise.
+const RELATED_TAGS_PER_THEME = 5;
+// Secondary-membership tuning. A doc qualifies as a secondary member of a
+// non-primary cluster when its centroid similarity is BOTH within DELTA of
+// the primary AND above MIN_ABS_SIMILARITY. The cap prevents a centroid-
+// of-mass doc from showing up in every cluster. Defaults are deliberately
+// conservative — secondaries should surface cross-cutting docs, not bury
+// the primary signal.
+const SECONDARY_DELTA = 0.1;
+const SECONDARY_MIN_SIMILARITY = 0.5;
+const SECONDARY_MAX_PER_DOC = 2;
+// Hard cap on secondaries reported per theme. Same UX rationale as
+// REPRESENTATIVE_DOCS_PER_THEME — enough to be useful, bounded enough to
+// not bury the user.
+const SECONDARY_DOCS_PER_THEME = 5;
+// Tokenisation for TF-IDF labels: lowercase, split on non-alphanumeric,
+// drop stopwords. This is deliberately a smaller list than bm25's — the
+// label step doesn't need FTS5 alignment, and we keep "tag-like" tokens.
+const LABEL_STOPWORDS = new Set([
+    "a",
+    "an",
+    "and",
+    "are",
+    "as",
+    "at",
+    "be",
+    "but",
+    "by",
+    "for",
+    "from",
+    "has",
+    "have",
+    "in",
+    "is",
+    "it",
+    "its",
+    "of",
+    "on",
+    "or",
+    "that",
+    "the",
+    "their",
+    "them",
+    "they",
+    "this",
+    "to",
+    "was",
+    "were",
+    "will",
+    "with",
+    "you",
+    "your",
+]);
+function labelTokens(text) {
+    return text
+        .toLowerCase()
+        .split(/[^a-z0-9]+/)
+        .filter((t) => t.length > 1 && !LABEL_STOPWORDS.has(t));
+}
+// Loads every (path, content_hash, embedding) row for the active model, in
+// path-then-chunk-index order, and groups embeddings by document path.
+function loadEmbeddingsByPath(db, model) {
+    const rows = db
+        .prepare(`SELECT c.path AS path, e.embedding AS embedding, e.dim AS dim
+         FROM chunks c
+         LEFT JOIN embeddings e
+           ON e.content_hash = c.content_hash AND e.model = ?
+        ORDER BY c.path, c.chunk_index`)
+        .all(model);
+    const provider = getProvider();
+    const expectedDim = provider.dim;
+    const out = new Map();
+    for (const row of rows) {
+        if (!row.embedding)
+            continue;
+        // Defense-in-depth: skip rows whose stored dim disagrees with the
+        // provider's expected dim — same guard `getAllChunks` applies.
+        const blobOk = row.embedding.length === expectedDim * 4;
+        const dimOk = row.dim === expectedDim;
+        if (!blobOk || !dimOk)
+            continue;
+        const vec = blobToEmbedding(row.embedding);
+        const list = out.get(row.path);
+        if (list)
+            list.push(vec);
+        else
+            out.set(row.path, [vec]);
+    }
+    return out;
+}
+// Applies the scope filter (collection, tags), RBAC, and the doc-has-an-
+// embedded-chunk requirement; returns the clusterable doc set and the count
+// of docs the embedding requirement skipped.
+function buildScopedDocs(documents, embeddingsByPath, filters, access) {
+    const scoped = [];
+    let skipped = 0;
+    for (const doc of documents) {
+        if (filters.collection && doc.collection !== filters.collection)
+            continue;
+        if (filters.tags && filters.tags.length > 0) {
+            const hasAll = filters.tags.every((t) => doc.tags.includes(t));
+            if (!hasAll)
+                continue;
+        }
+        if (access && !canRead(access.role, doc.collection))
+            continue;
+        const chunkVecs = embeddingsByPath.get(doc.path) ?? [];
+        const pooled = meanPoolL2(chunkVecs);
+        if (!pooled) {
+            // No embedded chunks (or an all-zero pool): count as skipped per
+            // the v1 contract.
+            skipped += 1;
+            continue;
+        }
+        scoped.push({
+            path: doc.path,
+            title: doc.title,
+            collection: doc.collection,
+            tags: doc.tags,
+            vector: pooled,
+        });
+    }
+    return { scoped, skipped };
+}
+// Cosine similarity for L2-normalised vectors == dot product. Used for
+// representative-doc ranking against a cluster centroid.
+function dotNormalized(a, b) {
+    let s = 0;
+    const n = Math.min(a.length, b.length);
+    for (let i = 0; i < n; i++)
+        s += a[i] * b[i];
+    return s;
+}
+// TF-IDF over the cluster: term frequency among the cluster's titles +
+// tags, divided by the document-frequency across ALL clusters (so a term
+// that names every theme — e.g. "doc" — doesn't dominate any one of them).
+// Falls back to the most common tags if TF-IDF yields nothing useful, and
+// finally to a generic "theme N" placeholder.
+function buildLabel(clusterDocs, globalDocCount, globalDf, themeIndex) {
+    const tf = new Map();
+    for (const doc of clusterDocs) {
+        const seen = new Set();
+        for (const tok of labelTokens(doc.title))
+            seen.add(tok);
+        for (const tag of doc.tags) {
+            for (const tok of labelTokens(tag))
+                seen.add(tok);
+        }
+        for (const tok of seen)
+            tf.set(tok, (tf.get(tok) ?? 0) + 1);
+    }
+    const scored = [];
+    for (const [term, freq] of tf) {
+        const df = globalDf.get(term) ?? 1;
+        // Standard smoothed IDF.
+        const idf = Math.log((globalDocCount + 1) / (df + 1)) + 1;
+        if (idf <= 0)
+            continue;
+        scored.push({ term, score: freq * idf });
+    }
+    scored.sort((a, b) => b.score - a.score);
+    const top = scored.slice(0, 3).map((s) => s.term);
+    if (top.length > 0)
+        return top.join(" / ");
+    // Fallback: most common tags.
+    const tagFreq = new Map();
+    for (const doc of clusterDocs) {
+        for (const tag of doc.tags)
+            tagFreq.set(tag, (tagFreq.get(tag) ?? 0) + 1);
+    }
+    const topTags = [...tagFreq.entries()]
+        .sort((a, b) => b[1] - a[1])
+        .slice(0, 3)
+        .map(([t]) => t);
+    if (topTags.length > 0)
+        return topTags.join(" / ");
+    return `theme ${themeIndex + 1}`;
+}
+// Builds the global document-frequency map across the clustered set. Used
+// by buildLabel; computed once and shared because every cluster's IDF must
+// agree on the document base.
+function buildGlobalDf(docs) {
+    const df = new Map();
+    for (const doc of docs) {
+        const seen = new Set();
+        for (const tok of labelTokens(doc.title))
+            seen.add(tok);
+        for (const tag of doc.tags) {
+            for (const tok of labelTokens(tag))
+                seen.add(tok);
+        }
+        for (const tok of seen)
+            df.set(tok, (df.get(tok) ?? 0) + 1);
+    }
+    return df;
+}
+function topTagsForCluster(clusterDocs) {
+    const freq = new Map();
+    for (const doc of clusterDocs) {
+        for (const tag of doc.tags)
+            freq.set(tag, (freq.get(tag) ?? 0) + 1);
+    }
+    return [...freq.entries()]
+        .sort((a, b) => b[1] - a[1] || a[0].localeCompare(b[0]))
+        .slice(0, RELATED_TAGS_PER_THEME)
+        .map(([t]) => t);
+}
+function representativesForCluster(clusterDocs, centroid) {
+    return [...clusterDocs]
+        .map((d) => ({ path: d.path, sim: dotNormalized(d.vector, centroid) }))
+        .sort((a, b) => b.sim - a.sim || a.path.localeCompare(b.path))
+        .slice(0, REPRESENTATIVE_DOCS_PER_THEME)
+        .map((d) => d.path);
+}
+function parseArgs(args) {
+    let k;
+    if (args.k !== undefined) {
+        if (typeof args.k !== "number" || !Number.isInteger(args.k) || args.k <= 0) {
+            return err(new Error("vault_themes: 'k' must be a positive integer when provided"));
+        }
+        k = args.k;
+    }
+    const collection = typeof args.collection === "string" && args.collection.length > 0 ? args.collection : undefined;
+    let tags;
+    if (Array.isArray(args.tags)) {
+        const filtered = args.tags.filter((t) => typeof t === "string" && t.length > 0);
+        if (filtered.length > 0)
+            tags = filtered;
+    }
+    return ok({ k, collection, tags });
+}
+export async function vaultThemes(vaultRoot, args, access) {
+    const parsed = parseArgs(args);
+    if (!parsed.ok)
+        return parsed;
+    const ready = await ensureIndexReady(vaultRoot);
+    if (!ready.ok)
+        return ready;
+    const dbResult = openIndexForActiveProvider(vaultRoot);
+    if (!dbResult.ok)
+        return dbResult;
+    const db = dbResult.value;
+    try {
+        const provider = getProvider();
+        const docRows = db
+            .prepare("SELECT path, title, collection, tags FROM documents ORDER BY path")
+            .all();
+        const documents = docRows.map((r) => ({
+            path: r.path,
+            title: r.title,
+            collection: r.collection,
+            domain: "",
+            status: "",
+            confidence: "",
+            updated: "",
+            tags: JSON.parse(r.tags),
+            content: "",
+            tokens: [],
+            ttlDays: null,
+            created: "",
+            supersededBy: null,
+        }));
+        const embeddingsByPath = loadEmbeddingsByPath(db, provider.id);
+        const { scoped, skipped } = buildScopedDocs(documents, embeddingsByPath, parsed.value, access);
+        if (scoped.length === 0) {
+            return ok({
+                themes: [],
+                totalDocuments: 0,
+                skippedDocuments: skipped,
+                selectedK: 0,
+                clusteredAt: new Date().toISOString(),
+            });
+        }
+        const vectors = scoped.map((s) => s.vector);
+        const rng = seededRng(THEMES_RNG_SEED);
+        let selectedK;
+        let assignments;
+        let centroids;
+        if (parsed.value.k !== undefined) {
+            const result = kmeans(vectors, parsed.value.k, rng, KMEANS_MAX_ITER);
+            selectedK = result.centroids.length;
+            assignments = result.assignments;
+            centroids = result.centroids;
+        }
+        else {
+            const result = pickK(vectors, DEFAULT_K_CANDIDATES, rng, KMEANS_MAX_ITER);
+            selectedK = result.centroids.length;
+            assignments = result.assignments;
+            centroids = result.centroids;
+        }
+        // Group docs by cluster.
+        const byCluster = new Map();
+        for (let i = 0; i < scoped.length; i++) {
+            const c = assignments[i];
+            const doc = scoped[i];
+            const list = byCluster.get(c);
+            if (list)
+                list.push(doc);
+            else
+                byCluster.set(c, [doc]);
+        }
+        const globalDf = buildGlobalDf(scoped);
+        // Secondary memberships: for each doc, find non-primary clusters its
+        // pooled vector also aligns with. This is the soft-reporting layer on
+        // top of the hard partition — it does NOT change documentCount, which
+        // still reflects primary membership only.
+        const secondaryByCluster = selectSecondaryMemberships(vectors, assignments, centroids, {
+            delta: SECONDARY_DELTA,
+            minSimilarity: SECONDARY_MIN_SIMILARITY,
+            maxPerDoc: SECONDARY_MAX_PER_DOC,
+        });
+        const themes = [];
+        let themeIndex = 0;
+        for (const [clusterIdx, clusterDocs] of byCluster) {
+            if (clusterDocs.length === 0)
+                continue;
+            const centroid = centroids[clusterIdx];
+            if (!centroid)
+                continue;
+            // Singleton clusters have no pairs to average — return null rather
+            // than the mathematically-trivial 1.0, which would falsely imply
+            // tightness.
+            const coherence = clusterDocs.length < 2 ? null : clusterCoherence(clusterDocs.map((d) => d.vector));
+            const secondaries = (secondaryByCluster.get(clusterIdx) ?? [])
+                .slice(0, SECONDARY_DOCS_PER_THEME)
+                .map((s) => scoped[s.docIndex].path);
+            themes.push({
+                label: buildLabel(clusterDocs, scoped.length, globalDf, themeIndex),
+                documentCount: clusterDocs.length,
+                coherence,
+                representativeDocs: representativesForCluster(clusterDocs, centroid),
+                secondaryDocs: secondaries,
+                relatedTags: topTagsForCluster(clusterDocs),
+            });
+            themeIndex += 1;
+        }
+        themes.sort((a, b) => b.documentCount - a.documentCount || a.label.localeCompare(b.label));
+        return ok({
+            themes,
+            totalDocuments: scoped.length,
+            skippedDocuments: skipped,
+            selectedK,
+            clusteredAt: new Date().toISOString(),
+        });
+    }
+    finally {
+        db.close();
+    }
+}
+// ---------------------------------------------------------------------------
+// MCP tool definition
+// ---------------------------------------------------------------------------
+export const themesTools = [
+    {
+        name: "vault_themes",
+        description: "Surface thematic clusters across the vault using k-means over " +
+            "document-pooled embeddings. Each document's chunk vectors are " +
+            "mean-pooled into one vector, L2-normalised, and clustered. By " +
+            "default the tool sweeps k ∈ {10, 15, 20, 25} and picks the k with " +
+            "the best mean silhouette; pass `k` to skip the sweep. Each theme " +
+            "reports a heuristic label (TF-IDF over titles + tags), a coherence " +
+            "score (mean pairwise cosine inside the cluster), representative " +
+            "documents nearest the centroid, and the most frequent tags. " +
+            "Output is deterministic for the same vault.",
+        inputSchema: {
+            type: "object",
+            properties: {
+                k: {
+                    type: "integer",
+                    description: "Optional explicit cluster count. When omitted, the tool sweeps " +
+                        "k ∈ {10, 15, 20, 25} and picks the k with the best silhouette.",
+                    minimum: 1,
+                },
+                collection: {
+                    type: "string",
+                    description: "Restrict clustering to documents in this collection.",
+                },
+                tags: {
+                    type: "array",
+                    items: { type: "string" },
+                    description: "Restrict to documents that have all of these tags.",
+                },
+            },
+            additionalProperties: false,
+        },
+        handler: (vaultRoot, args, access) => vaultThemes(vaultRoot, args, access),
+    },
+];
+//# sourceMappingURL=themes.js.map

package/dist/tools/themes.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"themes.js","sourceRoot":"","sources":["../../src/tools/themes.ts"],"names":[],"mappings":"AAAA,sEAAsE;AACtE,EAAE;AACF,0EAA0E;AAC1E,yEAAyE;AACzE,yEAAyE;AACzE,0EAA0E;AAC1E,kCAAkC;AAClC,EAAE;AACF,2BAA2B;AAC3B,qEAAqE;AACrE,oEAAoE;AACpE,kEAAkE;AAClE,8CAA8C;AAC9C,sEAAsE;AACtE,uBAAuB;AACvB,0DAA0D;AAE1D,OAAO,EAAsB,OAAO,EAAE,MAAM,mBAAmB,CAAC;AAChE,OAAO,EAAE,GAAG,EAAE,EAAE,EAAe,MAAM,yBAAyB,CAAC;AAC/D,OAAO,EAAE,WAAW,EAAE,MAAM,qBAAqB,CAAC;AAClD,OAAO,EAAE,eAAe,EAAsC,MAAM,wBAAwB,CAAC;AAC7F,OAAO,EACL,gBAAgB,EAChB,MAAM,EACN,UAAU,EACV,KAAK,EACL,SAAS,EACT,0BAA0B,GAC3B,MAAM,yBAAyB,CAAC;AAEjC,OAAO,EAAE,gBAAgB,EAAE,0BAA0B,EAAE,MAAM,aAAa,CAAC;AAE3E,0EAA0E;AAC1E,yEAAyE;AACzE,iEAAiE;AACjE,MAAM,eAAe,GAAG,SAAS,CAAC,CAAC,sCAAsC;AAEzE,0EAA0E;AAC1E,gDAAgD;AAChD,MAAM,oBAAoB,GAAG,CAAC,EAAE,EAAE,EAAE,EAAE,EAAE,EAAE,EAAE,CAAC,CAAC;AAE9C,uEAAuE;AACvE,4DAA4D;AAC5D,MAAM,eAAe,GAAG,EAAE,CAAC;AAE3B,2EAA2E;AAC3E,+CAA+C;AAC/C,MAAM,6BAA6B,GAAG,CAAC,CAAC;AAExC,yEAAyE;AACzE,qCAAqC;AACrC,MAAM,sBAAsB,GAAG,CAAC,CAAC;AAEjC,0EAA0E;AAC1E,2EAA2E;AAC3E,yEAAyE;AACzE,0EAA0E;AAC1E,yEAAyE;AACzE,sBAAsB;AACtB,MAAM,eAAe,GAAG,GAAG,CAAC;AAC5B,MAAM,wBAAwB,GAAG,GAAG,CAAC;AACrC,MAAM,qBAAqB,GAAG,CAAC,CAAC;AAEhC,mEAAmE;AACnE,yEAAyE;AACzE,qBAAqB;AACrB,MAAM,wBAAwB,GAAG,CAAC,CAAC;AAoCnC,wEAAwE;AACxE,wEAAwE;AACxE,yEAAyE;AACzE,MAAM,eAAe,GAAG,IAAI,GAAG,CAAC;IAC9B,GAAG;IACH,IAAI;IACJ,KAAK;IACL,KAAK;IACL,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,KAAK;IACL,IAAI;IACJ,KAAK;IACL,MAAM;IACN,KAAK;IACL,MAAM;IACN,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,KAAK;IACL,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,MAAM;IACN,KAAK;IACL,OAAO;IACP,MAAM;IACN,MAAM;IACN,MAAM;IACN,IAAI;IACJ,KAAK;IACL,MAAM;IACN,MAAM;IACN,MAAM;IACN,KAAK;IACL,MAAM;CACP,CAAC,CAAC;AAEH,SAAS,WAAW,CAAC,IAAY;IAC/B,OAAO,IAAI;SACR,WAAW,EAAE;SACb,KAAK,CAAC,YAAY,CAAC;SACnB,MAAM,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,MAAM,GAAG,CAAC,IAAI,CAAC,eAAe,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC,CAAC;AAC5D,CAAC;AAED,2EAA2E;AAC3E,uEAAuE;AACvE,SAAS,oBAAoB,CAAC,EAAW,EAAE,KAAa;IAMtD,MAAM,IAAI,GAAG,EAAE;SACZ,OAAO,CACN;;;;uCAIiC,CAClC;SACA,GAAG,CAAC,KAAK,CAAU,CAAC;IACvB,MAAM,QAAQ,GAAG,WAAW,EAAE,CAAC;IAC/B,MAAM,WAAW,GAAG,QAAQ,CAAC,GAAG,CAAC;IACjC,MAAM,GAAG,GAAG,IAAI,GAAG,EAA0B,CAAC;IAC9C,KAAK,MAAM,GAAG,IAAI,IAAI,EAAE,CAAC;QACvB,IAAI,CAAC,GAAG,CAAC,SAAS;YAAE,SAAS;QAC7B,kEAAkE;QAClE,+DAA+D;QAC/D,MAAM,MAAM,GAAG,GAAG,CAAC,SAAS,CAAC,MAAM,KAAK,WAAW,GAAG,CAAC,CAAC;QACxD,MAAM,KAAK,GAAG,GAAG,CAAC,GAAG,KAAK,WAAW,CAAC;QACtC,IAAI,CAAC,MAAM,IAAI,CAAC,KAAK;YAAE,SAAS;QAChC,MAAM,GAAG,GAAG,eAAe,CAAC,GAAG,CAAC,SAAS,CAAC,CAAC;QAC3C,MAAM,IAAI,GAAG,GAAG,CAAC,GAAG,CAAC,GAAG,CAAC,IAAI,CAAC,CAAC;QAC/B,IAAI,IAAI;YAAE,IAAI,CAAC,IAAI,CAAC,GAAG,CAAC,CAAC;;YACpB,GAAG,CAAC,GAAG,CAAC,GAAG,CAAC,IAAI,EAAE,CAAC,GAAG,CAAC,CAAC,CAAC;IAChC,CAAC;IACD,OAAO,GAAG,CAAC;AACb,CAAC;AAOD,yEAAyE;AACzE,4EAA4E;AAC5E,6CAA6C;AAC7C,SAAS,eAAe,CACtB,SAA4B,EAC5B,gBAA6C,EAC7C,OAAsB,EACtB,MAAiC;IAEjC,MAAM,MAAM,GAAgB,EAAE,CAAC;IAC/B,IAAI,OAAO,GAAG,CAAC,CAAC;IAChB,KAAK,MAAM,GAAG,IAAI,SAAS,EAAE,CAAC;QAC5B,IAAI,OAAO,CAAC,UAAU,IAAI,GAAG,CAAC,UAAU,KAAK,OAAO,CAAC,UAAU;YAAE,SAAS;QAC1E,IAAI,OAAO,CAAC,IAAI,IAAI,OAAO,CAAC,IAAI,CAAC,MAAM,GAAG,CAAC,EAAE,CAAC;YAC5C,MAAM,MAAM,GAAG,OAAO,CAAC,IAAI,CAAC,KAAK,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,GAAG,CAAC,IAAI,CAAC,QAAQ,CAAC,CAAC,CAAC,CAAC,CAAC;YAC/D,IAAI,CAAC,MAAM;gBAAE,SAAS;QACxB,CAAC;QACD,IAAI,MAAM,IAAI,CAAC,OAAO,CAAC,MAAM,CAAC,IAAI,EAAE,GAAG,CAAC,UAAU,CAAC;YAAE,SAAS;QAE9D,MAAM,SAAS,GAAG,gBAAgB,CAAC,GAAG,CAAC,GAAG,CAAC,IAAI,CAAC,IAAI,EAAE,CAAC;QACvD,MAAM,MAAM,GAAG,UAAU,CAAC,SAAS,CAAC,CAAC;QACrC,IAAI,CAAC,MAAM,EAAE,CAAC;YACZ,iEAAiE;YACjE,mBAAmB;YACnB,OAAO,IAAI,CAAC,CAAC;YACb,SAAS;QACX,CAAC;QACD,MAAM,CAAC,IAAI,CAAC;YACV,IAAI,EAAE,GAAG,CAAC,IAAI;YACd,KAAK,EAAE,GAAG,CAAC,KAAK;YAChB,UAAU,EAAE,GAAG,CAAC,UAAU;YAC1B,IAAI,EAAE,GAAG,CAAC,IAAI;YACd,MAAM,EAAE,MAAM;SACf,CAAC,CAAC;IACL,CAAC;IACD,OAAO,EAAE,MAAM,EAAE,OAAO,EAAE,CAAC;AAC7B,CAAC;AAED,uEAAuE;AACvE,yDAAyD;AACzD,SAAS,aAAa,CAAC,CAAe,EAAE,CAAe;IACrD,IAAI,CAAC,GAAG,CAAC,CAAC;IACV,MAAM,CAAC,GAAG,IAAI,CAAC,GAAG,CAAC,CAAC,CAAC,MAAM,EAAE,CAAC,CAAC,MAAM,CAAC,CAAC;IACvC,KAAK,IAAI,CAAC,GAAG,CAAC,EAAE,CAAC,GAAG,CAAC,EAAE,CAAC,EAAE;QAAE,CAAC,IAAK,CAAC,CAAC,CAAC,CAAY,GAAI,CAAC,CAAC,CAAC,CAAY,CAAC;IACrE,OAAO,CAAC,CAAC;AACX,CAAC;AAED,uEAAuE;AACvE,yEAAyE;AACzE,2EAA2E;AAC3E,0EAA0E;AAC1E,8CAA8C;AAC9C,SAAS,UAAU,CACjB,WAAwB,EACxB,cAAsB,EACtB,QAA6B,EAC7B,UAAkB;IAElB,MAAM,EAAE,GAAG,IAAI,GAAG,EAAkB,CAAC;IACrC,KAAK,MAAM,GAAG,IAAI,WAAW,EAAE,CAAC;QAC9B,MAAM,IAAI,GAAG,IAAI,GAAG,EAAU,CAAC;QAC/B,KAAK,MAAM,GAAG,IAAI,WAAW,CAAC,GAAG,CAAC,KAAK,CAAC;YAAE,IAAI,CAAC,GAAG,CAAC,GAAG,CAAC,CAAC;QACxD,KAAK,MAAM,GAAG,IAAI,GAAG,CAAC,IAAI,EAAE,CAAC;YAC3B,KAAK,MAAM,GAAG,IAAI,WAAW,CAAC,GAAG,CAAC;gBAAE,IAAI,CAAC,GAAG,CAAC,GAAG,CAAC,CAAC;QACpD,CAAC;QACD,KAAK,MAAM,GAAG,IAAI,IAAI;YAAE,EAAE,CAAC,GAAG,CAAC,GAAG,EAAE,CAAC,EAAE,CAAC,GAAG,CAAC,GAAG,CAAC,IAAI,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC;IAC9D,CAAC;IACD,MAAM,MAAM,GAAsC,EAAE,CAAC;IACrD,KAAK,MAAM,CAAC,IAAI,EAAE,IAAI,CAAC,IAAI,EAAE,EAAE,CAAC;QAC9B,MAAM,EAAE,GAAG,QAAQ,CAAC,GAAG,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC;QACnC,yBAAyB;QACzB,MAAM,GAAG,GAAG,IAAI,CAAC,GAAG,CAAC,CAAC,cAAc,GAAG,CAAC,CAAC,GAAG,CAAC,EAAE,GAAG,CAAC,CAAC,CAAC,GAAG,CAAC,CAAC;QAC1D,IAAI,GAAG,IAAI,CAAC;YAAE,SAAS;QACvB,MAAM,CAAC,IAAI,CAAC,EAAE,IAAI,EAAE,KAAK,EAAE,IAAI,GAAG,GAAG,EAAE,CAAC,CAAC;IAC3C,CAAC;IACD,MAAM,CAAC,IAAI,CAAC,CAAC,CAAC,EAAE,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,KAAK,GAAG,CAAC,CAAC,KAAK,CAAC,CAAC;IACzC,MAAM,GAAG,GAAG,MAAM,CAAC,KAAK,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,IAAI,CAAC,CAAC;IAClD,IAAI,GAAG,CAAC,MAAM,GAAG,CAAC;QAAE,OAAO,GAAG,CAAC,IAAI,CAAC,KAAK,CAAC,CAAC;IAE3C,8BAA8B;IAC9B,MAAM,OAAO,GAAG,IAAI,GAAG,EAAkB,CAAC;IAC1C,KAAK,MAAM,GAAG,IAAI,WAAW,EAAE,CAAC;QAC9B,KAAK,MAAM,GAAG,IAAI,GAAG,CAAC,IAAI;YAAE,OAAO,CAAC,GAAG,CAAC,GAAG,EAAE,CAAC,OAAO,CAAC,GAAG,CAAC,GAAG,CAAC,IAAI,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC;IAC5E,CAAC;IACD,MAAM,OAAO,GAAG,CAAC,GAAG,OAAO,CAAC,OAAO,EAAE,CAAC;SACnC,IAAI,CAAC,CAAC,CAAC,EAAE,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC,CAAC;SAC3B,KAAK,CAAC,CAAC,EAAE,CAAC,CAAC;SACX,GAAG,CAAC,CAAC,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,CAAC;IACnB,IAAI,OAAO,CAAC,MAAM,GAAG,CAAC;QAAE,OAAO,OAAO,CAAC,IAAI,CAAC,KAAK,CAAC,CAAC;IAEnD,OAAO,SAAS,UAAU,GAAG,CAAC,EAAE,CAAC;AACnC,CAAC;AAED,0EAA0E;AAC1E,2EAA2E;AAC3E,8BAA8B;AAC9B,SAAS,aAAa,CAAC,IAAiB;IACtC,MAAM,EAAE,GAAG,IAAI,GAAG,EAAkB,CAAC;IACrC,KAAK,MAAM,GAAG,IAAI,IAAI,EAAE,CAAC;QACvB,MAAM,IAAI,GAAG,IAAI,GAAG,EAAU,CAAC;QAC/B,KAAK,MAAM,GAAG,IAAI,WAAW,CAAC,GAAG,CAAC,KAAK,CAAC;YAAE,IAAI,CAAC,GAAG,CAAC,GAAG,CAAC,CAAC;QACxD,KAAK,MAAM,GAAG,IAAI,GAAG,CAAC,IAAI,EAAE,CAAC;YAC3B,KAAK,MAAM,GAAG,IAAI,WAAW,CAAC,GAAG,CAAC;gBAAE,IAAI,CAAC,GAAG,CAAC,GAAG,CAAC,CAAC;QACpD,CAAC;QACD,KAAK,MAAM,GAAG,IAAI,IAAI;YAAE,EAAE,CAAC,GAAG,CAAC,GAAG,EAAE,CAAC,EAAE,CAAC,GAAG,CAAC,GAAG,CAAC,IAAI,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC;IAC9D,CAAC;IACD,OAAO,EAAE,CAAC;AACZ,CAAC;AAED,SAAS,iBAAiB,CAAC,WAAwB;IACjD,MAAM,IAAI,GAAG,IAAI,GAAG,EAAkB,CAAC;IACvC,KAAK,MAAM,GAAG,IAAI,WAAW,EAAE,CAAC;QAC9B,KAAK,MAAM,GAAG,IAAI,GAAG,CAAC,IAAI;YAAE,IAAI,CAAC,GAAG,CAAC,GAAG,EAAE,CAAC,IAAI,CAAC,GAAG,CAAC,GAAG,CAAC,IAAI,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC;IACtE,CAAC;IACD,OAAO,CAAC,GAAG,IAAI,CAAC,OAAO,EAAE,CAAC;SACvB,IAAI,CAAC,CAAC,CAAC,EAAE,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC,IAAI,CAAC,CAAC,CAAC,CAAC,CAAC,aAAa,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC;SACvD,KAAK,CAAC,CAAC,EAAE,sBAAsB,CAAC;SAChC,GAAG,CAAC,CAAC,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,CAAC;AACrB,CAAC;AAED,SAAS,yBAAyB,CAAC,WAAwB,EAAE,QAAsB;IACjF,OAAO,CAAC,GAAG,WAAW,CAAC;SACpB,GAAG,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,EAAE,IAAI,EAAE,CAAC,CAAC,IAAI,EAAE,GAAG,EAAE,aAAa,CAAC,CAAC,CAAC,MAAM,EAAE,QAAQ,CAAC,EAAE,CAAC,CAAC;SACtE,IAAI,CAAC,CAAC,CAAC,EAAE,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,GAAG,GAAG,CAAC,CAAC,GAAG,IAAI,CAAC,CAAC,IAAI,CAAC,aAAa,CAAC,CAAC,CAAC,IAAI,CAAC,CAAC;SAC7D,KAAK,CAAC,CAAC,EAAE,6BAA6B,CAAC;SACvC,GAAG,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,IAAI,CAAC,CAAC;AACxB,CAAC;AAQD,SAAS,SAAS,CAAC,IAA6B;IAC9C,IAAI,CAAqB,CAAC;IAC1B,IAAI,IAAI,CAAC,CAAC,KAAK,SAAS,EAAE,CAAC;QACzB,IAAI,OAAO,IAAI,CAAC,CAAC,KAAK,QAAQ,IAAI,CAAC,MAAM,CAAC,SAAS,CAAC,IAAI,CAAC,CAAC,CAAC,IAAI,IAAI,CAAC,CAAC,IAAI,CAAC,EAAE,CAAC;YAC3E,OAAO,GAAG,CAAC,IAAI,KAAK,CAAC,4DAA4D,CAAC,CAAC,CAAC;QACtF,CAAC;QACD,CAAC,GAAG,IAAI,CAAC,CAAC,CAAC;IACb,CAAC;IACD,MAAM,UAAU,GACd,OAAO,IAAI,CAAC,UAAU,KAAK,QAAQ,IAAI,IAAI,CAAC,UAAU,CAAC,MAAM,GAAG,CAAC,CAAC,CAAC,CAAC,IAAI,CAAC,UAAU,CAAC,CAAC,CAAC,SAAS,CAAC;IAClG,IAAI,IAA0B,CAAC;IAC/B,IAAI,KAAK,CAAC,OAAO,CAAC,IAAI,CAAC,IAAI,CAAC,EAAE,CAAC;QAC7B,MAAM,QAAQ,GAAG,IAAI,CAAC,IAAI,CAAC,MAAM,CAAC,CAAC,CAAC,EAAe,EAAE,CAAC,OAAO,CAAC,KAAK,QAAQ,IAAI,CAAC,CAAC,MAAM,GAAG,CAAC,CAAC,CAAC;QAC7F,IAAI,QAAQ,CAAC,MAAM,GAAG,CAAC;YAAE,IAAI,GAAG,QAAQ,CAAC;IAC3C,CAAC;IACD,OAAO,EAAE,CAAC,EAAE,CAAC,EAAE,UAAU,EAAE,IAAI,EAAE,CAAC,CAAC;AACrC,CAAC;AAED,MAAM,CAAC,KAAK,UAAU,WAAW,CAC/B,SAAiB,EACjB,IAA6B,EAC7B,MAAsB;IAEtB,MAAM,MAAM,GAAG,SAAS,CAAC,IAAI,CAAC,CAAC;IAC/B,IAAI,CAAC,MAAM,CAAC,EAAE;QAAE,OAAO,MAAM,CAAC;IAE9B,MAAM,KAAK,GAAG,MAAM,gBAAgB,CAAC,SAAS,CAAC,CAAC;IAChD,IAAI,CAAC,KAAK,CAAC,EAAE;QAAE,OAAO,KAAK,CAAC;IAE5B,MAAM,QAAQ,GAAG,0BAA0B,CAAC,SAAS,CAAC,CAAC;IACvD,IAAI,CAAC,QAAQ,CAAC,EAAE;QAAE,OAAO,QAAQ,CAAC;IAClC,MAAM,EAAE,GAAG,QAAQ,CAAC,KAAK,CAAC;IAC1B,IAAI,CAAC;QACH,MAAM,QAAQ,GAAG,WAAW,EAAE,CAAC;QAO/B,MAAM,OAAO,GAAG,EAAE;aACf,OAAO,CAAC,mEAAmE,CAAC;aAC5E,GAAG,EAAc,CAAC;QACrB,MAAM,SAAS,GAAsB,OAAO,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC;YACvD,IAAI,EAAE,CAAC,CAAC,IAAI;YACZ,KAAK,EAAE,CAAC,CAAC,KAAK;YACd,UAAU,EAAE,CAAC,CAAC,UAAU;YACxB,MAAM,EAAE,EAAE;YACV,MAAM,EAAE,EAAE;YACV,UAAU,EAAE,EAAE;YACd,OAAO,EAAE,EAAE;YACX,IAAI,EAAE,IAAI,CAAC,KAAK,CAAC,CAAC,CAAC,IAAI,CAAa;YACpC,OAAO,EAAE,EAAE;YACX,MAAM,EAAE,EAAE;YACV,OAAO,EAAE,IAAI;YACb,OAAO,EAAE,EAAE;YACX,YAAY,EAAE,IAAI;SACnB,CAAC,CAAC,CAAC;QAEJ,MAAM,gBAAgB,GAAG,oBAAoB,CAAC,EAAE,EAAE,QAAQ,CAAC,EAAE,CAAC,CAAC;QAC/D,MAAM,EAAE,MAAM,EAAE,OAAO,EAAE,GAAG,eAAe,CAAC,SAAS,EAAE,gBAAgB,EAAE,MAAM,CAAC,KAAK,EAAE,MAAM,CAAC,CAAC;QAE/F,IAAI,MAAM,CAAC,MAAM,KAAK,CAAC,EAAE,CAAC;YACxB,OAAO,EAAE,CAAC;gBACR,MAAM,EAAE,EAAE;gBACV,cAAc,EAAE,CAAC;gBACjB,gBAAgB,EAAE,OAAO;gBACzB,SAAS,EAAE,CAAC;gBACZ,WAAW,EAAE,IAAI,IAAI,EAAE,CAAC,WAAW,EAAE;aACtC,CAAC,CAAC;QACL,CAAC;QAED,MAAM,OAAO,GAAG,MAAM,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,MAAM,CAAC,CAAC;QAC5C,MAAM,GAAG,GAAG,SAAS,CAAC,eAAe,CAAC,CAAC;QACvC,IAAI,SAAiB,CAAC;QACtB,IAAI,WAAqB,CAAC;QAC1B,IAAI,SAAyB,CAAC;QAC9B,IAAI,MAAM,CAAC,KAAK,CAAC,CAAC,KAAK,SAAS,EAAE,CAAC;YACjC,MAAM,MAAM,GAAG,MAAM,CAAC,OAAO,EAAE,MAAM,CAAC,KAAK,CAAC,CAAC,EAAE,GAAG,EAAE,eAAe,CAAC,CAAC;YACrE,SAAS,GAAG,MAAM,CAAC,SAAS,CAAC,MAAM,CAAC;YACpC,WAAW,GAAG,MAAM,CAAC,WAAW,CAAC;YACjC,SAAS,GAAG,MAAM,CAAC,SAAS,CAAC;QAC/B,CAAC;aAAM,CAAC;YACN,MAAM,MAAM,GAAG,KAAK,CAAC,OAAO,EAAE,oBAAoB,EAAE,GAAG,EAAE,eAAe,CAAC,CAAC;YAC1E,SAAS,GAAG,MAAM,CAAC,SAAS,CAAC,MAAM,CAAC;YACpC,WAAW,GAAG,MAAM,CAAC,WAAW,CAAC;YACjC,SAAS,GAAG,MAAM,CAAC,SAAS,CAAC;QAC/B,CAAC;QAED,yBAAyB;QACzB,MAAM,SAAS,GAAG,IAAI,GAAG,EAAuB,CAAC;QACjD,KAAK,IAAI,CAAC,GAAG,CAAC,EAAE,CAAC,GAAG,MAAM,CAAC,MAAM,EAAE,CAAC,EAAE,EAAE,CAAC;YACvC,MAAM,CAAC,GAAG,WAAW,CAAC,CAAC,CAAW,CAAC;YACnC,MAAM,GAAG,GAAG,MAAM,CAAC,CAAC,CAAc,CAAC;YACnC,MAAM,IAAI,GAAG,SAAS,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC;YAC9B,IAAI,IAAI;gBAAE,IAAI,CAAC,IAAI,CAAC,GAAG,CAAC,CAAC;;gBACpB,SAAS,CAAC,GAAG,CAAC,CAAC,EAAE,CAAC,GAAG,CAAC,CAAC,CAAC;QAC/B,CAAC;QAED,MAAM,QAAQ,GAAG,aAAa,CAAC,MAAM,CAAC,CAAC;QAEvC,qEAAqE;QACrE,sEAAsE;QACtE,sEAAsE;QACtE,0CAA0C;QAC1C,MAAM,kBAAkB,GAAG,0BAA0B,CAAC,OAAO,EAAE,WAAW,EAAE,SAAS,EAAE;YACrF,KAAK,EAAE,eAAe;YACtB,aAAa,EAAE,wBAAwB;YACvC,SAAS,EAAE,qBAAqB;SACjC,CAAC,CAAC;QAEH,MAAM,MAAM,GAAiB,EAAE,CAAC;QAChC,IAAI,UAAU,GAAG,CAAC,CAAC;QACnB,KAAK,MAAM,CAAC,UAAU,EAAE,WAAW,CAAC,IAAI,SAAS,EAAE,CAAC;YAClD,IAAI,WAAW,CAAC,MAAM,KAAK,CAAC;gBAAE,SAAS;YACvC,MAAM,QAAQ,GAAG,SAAS,CAAC,UAAU,CAAC,CAAC;YACvC,IAAI,CAAC,QAAQ;gBAAE,SAAS;YACxB,mEAAmE;YACnE,iEAAiE;YACjE,aAAa;YACb,MAAM,SAAS,GACb,WAAW,CAAC,MAAM,GAAG,CAAC,CAAC,CAAC,CAAC,IAAI,CAAC,CAAC,CAAC,gBAAgB,CAAC,WAAW,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,MAAM,CAAC,CAAC,CAAC;YACrF,MAAM,WAAW,GAAG,CAAC,kBAAkB,CAAC,GAAG,CAAC,UAAU,CAAC,IAAI,EAAE,CAAC;iBAC3D,KAAK,CAAC,CAAC,EAAE,wBAAwB,CAAC;iBAClC,GAAG,CAAC,CAAC,CAAC,EAAE,EAAE,CAAE,MAAM,CAAC,CAAC,CAAC,QAAQ,CAAe,CAAC,IAAI,CAAC,CAAC;YACtD,MAAM,CAAC,IAAI,CAAC;gBACV,KAAK,EAAE,UAAU,CAAC,WAAW,EAAE,MAAM,CAAC,MAAM,EAAE,QAAQ,EAAE,UAAU,CAAC;gBACnE,aAAa,EAAE,WAAW,CAAC,MAAM;gBACjC,SAAS;gBACT,kBAAkB,EAAE,yBAAyB,CAAC,WAAW,EAAE,QAAQ,CAAC;gBACpE,aAAa,EAAE,WAAW;gBAC1B,WAAW,EAAE,iBAAiB,CAAC,WAAW,CAAC;aAC5C,CAAC,CAAC;YACH,UAAU,IAAI,CAAC,CAAC;QAClB,CAAC;QAED,MAAM,CAAC,IAAI,CAAC,CAAC,CAAC,EAAE,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,aAAa,GAAG,CAAC,CAAC,aAAa,IAAI,CAAC,CAAC,KAAK,CAAC,aAAa,CAAC,CAAC,CAAC,KAAK,CAAC,CAAC,CAAC;QAE3F,OAAO,EAAE,CAAC;YACR,MAAM;YACN,cAAc,EAAE,MAAM,CAAC,MAAM;YAC7B,gBAAgB,EAAE,OAAO;YACzB,SAAS;YACT,WAAW,EAAE,IAAI,IAAI,EAAE,CAAC,WAAW,EAAE;SACtC,CAAC,CAAC;IACL,CAAC;YAAS,CAAC;QACT,EAAE,CAAC,KAAK,EAAE,CAAC;IACb,CAAC;AACH,CAAC;AAED,8EAA8E;AAC9E,sBAAsB;AACtB,8EAA8E;AAE9E,MAAM,CAAC,MAAM,WAAW,GAAqB;IAC3C;QACE,IAAI,EAAE,cAAc;QACpB,WAAW,EACT,gEAAgE;YAChE,gEAAgE;YAChE,gEAAgE;YAChE,oEAAoE;YACpE,mEAAmE;YACnE,qEAAqE;YACrE,kEAAkE;YAClE,8DAA8D;YAC9D,6CAA6C;QAC/C,WAAW,EAAE;YACX,IAAI,EAAE,QAAQ;YACd,UAAU,EAAE;gBACV,CAAC,EAAE;oBACD,IAAI,EAAE,SAAS;oBACf,WAAW,EACT,iEAAiE;wBACjE,gEAAgE;oBAClE,OAAO,EAAE,CAAC;iBACX;gBACD,UAAU,EAAE;oBACV,IAAI,EAAE,QAAQ;oBACd,WAAW,EAAE,sDAAsD;iBACpE;gBACD,IAAI,EAAE;oBACJ,IAAI,EAAE,OAAO;oBACb,KAAK,EAAE,EAAE,IAAI,EAAE,QAAQ,EAAE;oBACzB,WAAW,EAAE,oDAAoD;iBAClE;aACF;YACD,oBAAoB,EAAE,KAAK;SAC5B;QACD,OAAO,EAAE,CAAC,SAAS,EAAE,IAAI,EAAE,MAAM,EAAE,EAAE,CAAC,WAAW,CAAC,SAAS,EAAE,IAAI,EAAE,MAAM,CAAC;KAC3E;CACF,CAAC"}

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "daftari",
-  "version": "1.9.1",
+  "version": "1.11.0",
   "description": "An open-source, multi-user knowledge vault exposed to AI agents via an MCP server.",
   "license": "MIT",
   "author": "Mihir Wagle / mavaali",