@nicia-ai/typegraph 0.28.1 → 0.29.0

package/dist/{chunk-WVQXAFP3.js → chunk-454I6GDF.js}

@@ -32,7 +32,7 @@ function embedding(dimensions, options = {}) {
   });
 }
 function resolveEmbeddingIndex(options) {
-  const indexType = options.indexType ?? "hnsw";
+  const indexType = options.indexType ?? DEFAULT_EMBEDDING_INDEX_TYPE;
   const m = options.m ?? DEFAULT_HNSW_M;
   const efConstruction = options.efConstruction ?? DEFAULT_HNSW_EF_CONSTRUCTION;
   if (m <= 0 || !Number.isInteger(m)) {
@@ -51,7 +51,7 @@ function resolveEmbeddingIndex(options) {
     );
   }
   return Object.freeze({
-    metric: options.metric ?? "cosine",
+    metric: options.metric ?? DEFAULT_EMBEDDING_METRIC,
     indexType,
     m,
     efConstruction,
@@ -73,7 +73,48 @@ function getEmbeddingIndex(schema) {
   if (candidate === void 0) return void 0;
   return candidate;
 }
+var DEFAULT_EMBEDDING_METRIC = "cosine";
+var DEFAULT_EMBEDDING_INDEX_TYPE = "hnsw";
+function resolveEmbeddingFields(schema) {
+  if (schema.type !== "object") return [];
+  const shape = schema.def.shape;
+  if (!shape) return [];
+  const fields = [];
+  for (const [fieldPath, fieldSchema] of Object.entries(shape)) {
+    const dimensions = resolveEmbeddingDimensions(fieldSchema);
+    if (dimensions === void 0) continue;
+    const index = resolveEmbeddingIndexConfig(fieldSchema);
+    fields.push({
+      fieldPath,
+      dimensions,
+      metric: index?.metric ?? DEFAULT_EMBEDDING_METRIC,
+      indexType: index?.indexType ?? DEFAULT_EMBEDDING_INDEX_TYPE
+    });
+  }
+  return fields;
+}
+function resolveEmbeddingDimensions(schema) {
+  const direct = getEmbeddingDimensions(schema);
+  if (direct !== void 0) return direct;
+  const unwrapped = unwrapToEmbedding(schema);
+  return unwrapped === void 0 ? void 0 : getEmbeddingDimensions(unwrapped);
+}
+function resolveEmbeddingIndexConfig(schema) {
+  const direct = getEmbeddingIndex(schema);
+  if (direct !== void 0) return direct;
+  const unwrapped = unwrapToEmbedding(schema);
+  return unwrapped === void 0 ? void 0 : getEmbeddingIndex(unwrapped);
+}
+function unwrapToEmbedding(schema) {
+  const type = schema.type;
+  const innerType = schema.def.innerType;
+  if ((type === "optional" || type === "nullable" || type === "default" || type === "readonly") && innerType) {
+    if (isEmbeddingSchema(innerType)) return innerType;
+    return unwrapToEmbedding(innerType);
+  }
+  return void 0;
+}
 
-export { embedding, getEmbeddingDimensions, getEmbeddingIndex, isEmbeddingSchema };
-//# sourceMappingURL=chunk-WVQXAFP3.js.map
-//# sourceMappingURL=chunk-WVQXAFP3.js.map
+export { DEFAULT_EMBEDDING_INDEX_TYPE, DEFAULT_EMBEDDING_METRIC, embedding, getEmbeddingDimensions, getEmbeddingIndex, isEmbeddingSchema, resolveEmbeddingFields };
+//# sourceMappingURL=chunk-454I6GDF.js.map
+//# sourceMappingURL=chunk-454I6GDF.js.map

package/dist/chunk-454I6GDF.js.map

@@ -0,0 +1 @@
+{"version":3,"sources":["../src/core/embedding.ts"],"names":[],"mappings":";;;AAqCO,IAAM,wBAAA,GAA2B,sBAAA;AAOjC,IAAM,mBAAA,GAAsB,iBAAA;AA4DnC,IAAM,cAAA,GAAiB,EAAA;AACvB,IAAM,4BAAA,GAA+B,EAAA;AAyD9B,SAAS,SAAA,CACd,UAAA,EACA,OAAA,GAAiC,EAAC,EACd;AACpB,EAAA,IAAI,CAAC,MAAA,CAAO,SAAA,CAAU,UAAU,CAAA,IAAK,cAAc,CAAA,EAAG;AACpD,IAAA,MAAM,IAAI,KAAA;AAAA,MACR,yDAAyD,UAAU,CAAA;AAAA,KACrE;AAAA,EACF;AAEA,EAAA,MAAM,WAAA,GAAc,sBAAsB,OAAO,CAAA;AAIjD,EAAA,MAAM,SAAS,CAAA,CAAE,MAAA;AAAA,IACf,CAAC,KAAA,KAAmC;AAClC,MAAA,IAAI,CAAC,KAAA,CAAM,OAAA,CAAQ,KAAK,CAAA,EAAG;AACzB,QAAA,OAAO,KAAA;AAAA,MACT;AACA,MAAA,IAAI,KAAA,CAAM,WAAW,UAAA,EAAY;AAC/B,QAAA,OAAO,KAAA;AAAA,MACT;AACA,MAAA,OAAO,KAAA,CAAM,KAAA,CAAM,CAAC,CAAA,KAAM,OAAO,MAAM,QAAA,IAAY,MAAA,CAAO,QAAA,CAAS,CAAC,CAAC,CAAA;AAAA,IACvE,CAAA;AAAA,IACA;AAAA,MACE,OAAA,EAAS,wBAAwB,UAAU,CAAA,eAAA;AAAA;AAC7C,GACF;AAEA,EAAA,OAAO,MAAA,CAAO,OAAO,MAAA,EAAQ;AAAA,IAC3B,CAAC,wBAAwB,GAAG,UAAA;AAAA,IAC5B,CAAC,mBAAmB,GAAG;AAAA,GACxB,CAAA;AACH;AAEA,SAAS,sBACP,OAAA,EACwB;AACxB,EAAA,MAAM,SAAA,GAAY,QAAQ,SAAA,IAAa,4BAAA;AACvC,EAAA,MAAM,CAAA,GAAI,QAAQ,CAAA,IAAK,cAAA;AACvB,EAAA,MAAM,cAAA,GAAiB,QAAQ,cAAA,IAAkB,4BAAA;AAEjD,EAAA,IAAI,KAAK,CAAA,IAAK,CAAC,MAAA,CAAO,SAAA,CAAU,CAAC,CAAA,EAAG;AAClC,IAAA,MAAM,IAAI,KAAA;AAAA,MACR,iEAAiE,CAAC,CAAA;AAAA,KACpE;AAAA,EACF;AACA,EAAA,IAAI,kBAAkB,CAAA,IAAK,CAAC,MAAA,CAAO,SAAA,CAAU,cAAc,CAAA,EAAG;AAC5D,IAAA,MAAM,IAAI,KAAA;AAAA,MACR,8EAA8E,cAAc,CAAA;AAAA,KAC9F;AAAA,EACF;AACA,EAAA,IACE,OAAA,CAAQ,KAAA,KAAU,MAAA,KACjB,OAAA,CAAQ,KAAA,IAAS,CAAA,IAAK,CAAC,MAAA,CAAO,SAAA,CAAU,OAAA,CAAQ,KAAK,CAAA,CAAA,EACtD;AACA,IAAA,MAAM,IAAI,KAAA;AAAA,MACR,CAAA,kEAAA,EAAqE,QAAQ,KAAK,CAAA;AAAA,KACpF;AAAA,EACF;AAEA,EAAA,OAAO,OAAO,MAAA,CAAO;AAAA,IACnB,MAAA,EAAQ,QAAQ,MAAA,IAAU,wBAAA;AAAA,IAC1B,SAAA;AAAA,IACA,CAAA;AAAA,IACA,cAAA;AAAA,IACA,OAAO,OAAA,CAAQ;AAAA,GAChB,CAAA;AACH;AASO,SAAS,kBAAkB,KAAA,EAA0C;AAC1E,EAAA,OACE,OAAO,KAAA,KAAU,QAAA,IACjB,KAAA,KAAU,IAAA,IACV,4BAA4B,KAAA,IAC5B,OAAQ,KAAA,CAAkC,wBAAwB,CAAA,KAChE,QAAA;AAEN;AAMO,SAAS,uBAAuB,MAAA,EAAuC;AAC5E,EAAA,IAAI,iBAAA,CAAkB,MAAM,CAAA,EAAG;AAC7B,IAAA,OAAO,OAAO,wBAAwB,CAAA;AAAA,EACxC;AACA,EAAA,OAAO,MAAA;AACT;AAQO,SAAS,kBACd,MAAA,EACoC;AACpC,EAAA,IAAI,CAAC,iBAAA,CAAkB,MAAM,CAAA,EAAG,OAAO,MAAA;AACvC,EAAA,MAAM,SAAA,GAAa,OACjB,mBACF,CAAA;AACA,EAAA,IAAI,SAAA,KAAc,QAAW,OAAO,MAAA;AACpC,EAAA,OAAO,SAAA;AACT;AA2BO,IAAM,wBAAA,GAA4C;AAClD,IAAM,4BAAA,GAAmD;AAQzD,SAAS,uBACd,MAAA,EACmC;AACnC,EAAA,IAAI,MAAA,CAAO,IAAA,KAAS,QAAA,EAAU,OAAO,EAAC;AACtC,EAAA,MAAM,KAAA,GAAS,OAAO,GAAA,CAA8C,KAAA;AACpE,EAAA,IAAI,CAAC,KAAA,EAAO,OAAO,EAAC;AAEpB,EAAA,MAAM,SAAmC,EAAC;AAC1C,EAAA,KAAA,MAAW,CAAC,SAAA,EAAW,WAAW,KAAK,MAAA,CAAO,OAAA,CAAQ,KAAK,CAAA,EAAG;AAC5D,IAAA,MAAM,UAAA,GAAa,2BAA2B,WAAW,CAAA;AACzD,IAAA,IAAI,eAAe,MAAA,EAAW;AAC9B,IAAA,MAAM,KAAA,GAAQ,4BAA4B,WAAW,CAAA;AACrD,IAAA,MAAA,CAAO,IAAA,CAAK;AAAA,MACV,SAAA;AAAA,MACA,UAAA;AAAA,MACA,MAAA,EAAQ,OAAO,MAAA,IAAU,wBAAA;AAAA,MACzB,SAAA,EAAW,OAAO,SAAA,IAAa;AAAA,KAChC,CAAA;AAAA,EACH;AACA,EAAA,OAAO,MAAA;AACT;AAEA,SAAS,2BAA2B,MAAA,EAAuC;AACzE,EAAA,MAAM,MAAA,GAAS,uBAAuB,MAAM,CAAA;AAC5C,EAAA,IAAI,MAAA,KAAW,QAAW,OAAO,MAAA;AACjC,EAAA,MAAM,SAAA,GAAY,kBAAkB,MAAM,CAAA;AAC1C,EAAA,OAAO,SAAA,KAAc,MAAA,GAAY,MAAA,GAC7B,sBAAA,CAAuB,SAAS,CAAA;AAEtC;AAEA,SAAS,4BACP,MAAA,EACoC;AACpC,EAAA,MAAM,MAAA,GAAS,kBAAkB,MAAM,CAAA;AACvC,EAAA,IAAI,MAAA,KAAW,QAAW,OAAO,MAAA;AACjC,EAAA,MAAM,SAAA,GAAY,kBAAkB,MAAM,CAAA;AAC1C,EAAA,OAAO,SAAA,KAAc,MAAA,GAAY,MAAA,GAAY,iBAAA,CAAkB,SAAS,CAAA;AAC1E;AAOA,SAAS,kBAAkB,MAAA,EAA0C;AACnE,EAAA,MAAM,OAAO,MAAA,CAAO,IAAA;AACpB,EAAA,MAAM,SAAA,GAAa,OAAO,GAAA,CAAkC,SAAA;AAC5D,EAAA,IAAA,CACG,IAAA,KAAS,cACR,IAAA,KAAS,UAAA,IACT,SAAS,SAAA,IACT,IAAA,KAAS,eACX,SAAA,EACA;AACA,IAAA,IAAI,iBAAA,CAAkB,SAAS,CAAA,EAAG,OAAO,SAAA;AACzC,IAAA,OAAO,kBAAkB,SAAS,CAAA;AAAA,EACpC;AACA,EAAA,OAAO,MAAA;AACT","file":"chunk-454I6GDF.js","sourcesContent":["/**\n * Embedding type for vector search.\n *\n * Creates a Zod-compatible schema for vector embeddings with\n * dimension validation and metadata for query compilation.\n */\nimport { z } from \"zod\";\n\n// ============================================================\n// Embedding Brand Symbol\n// ============================================================\n\n/**\n * Symbol used to brand embedding values at the type level.\n * This allows TypeScript to distinguish embedding arrays from regular\n * number arrays in the query builder type system.\n */\ndeclare const EMBEDDING_BRAND: unique symbol;\n\n/**\n * Branded embedding type for type-level distinction.\n * At runtime this is just `readonly number[]`, but TypeScript can\n * distinguish it from regular arrays for query builder typing.\n */\nexport type EmbeddingValue = readonly number[] & {\n  readonly [EMBEDDING_BRAND]: true;\n};\n\n// ============================================================\n// Embedding Metadata Symbol\n// ============================================================\n\n/**\n * Symbol key for storing embedding dimensions on the schema.\n * This allows the schema introspector to detect embedding types\n * and extract dimension information.\n */\nexport const EMBEDDING_DIMENSIONS_KEY = \"_embeddingDimensions\" as const;\n\n/**\n * Symbol key for storing the embedding's preferred index configuration\n * on the schema. The schema introspector reads this when auto-deriving\n * `VectorIndexDeclaration` entries at `defineGraph()` time.\n */\nexport const EMBEDDING_INDEX_KEY = \"_embeddingIndex\" as const;\n\n/**\n * Distance metric used for vector similarity. Pinned at the embedding\n * brand because changing the metric usually means generating different\n * model output (e.g. cosine-normalized vs. raw inner-product), and the\n * stored index needs the same metric to score correctly.\n *\n * - `cosine`: cosine similarity. The default — works for most\n *   sentence-transformer / OpenAI-style embeddings that are already\n *   length-normalized.\n * - `l2`: Euclidean distance. Use when your model outputs vectors\n *   trained for L2 (e.g. some image embedding models).\n * - `inner_product`: dot-product distance. Use when your model outputs\n *   pre-normalized vectors AND you want to skip the cosine\n *   normalization step at query time.\n */\nexport type EmbeddingMetric = \"cosine\" | \"l2\" | \"inner_product\";\n\n/**\n * Vector index implementation. `hnsw` is the default and what pgvector\n * recommends for most workloads. `ivfflat` is the alternative for\n * larger datasets where memory is the bottleneck. `none` disables\n * automatic index creation for this embedding (the operator can still\n * call `backend.createVectorIndex` manually).\n */\nexport type EmbeddingIndexType = \"hnsw\" | \"ivfflat\" | \"none\";\n\n/**\n * Per-embedding configuration for the auto-derived\n * `VectorIndexDeclaration`. All fields are optional with sensible\n * defaults (`cosine` / `hnsw` / pgvector defaults: `m=16`,\n * `ef_construction=64`). Override only when your model or dataset has\n * a known reason to.\n */\nexport type EmbeddingIndexOptions = Readonly<{\n  /** Distance metric. Default `\"cosine\"`. */\n  metric?: EmbeddingMetric;\n  /** Vector index implementation. Default `\"hnsw\"`. */\n  indexType?: EmbeddingIndexType;\n  /** HNSW `m` parameter — max connections per layer. Default `16`. */\n  m?: number;\n  /** HNSW `ef_construction` parameter — build-time search depth. Default `64`. */\n  efConstruction?: number;\n  /** IVFFlat `lists` parameter — number of inverted-list partitions. */\n  lists?: number;\n}>;\n\n/**\n * Resolved embedding index configuration with all defaults applied.\n * What gets attached to the brand and read by the auto-derivation pass.\n */\nexport type ResolvedEmbeddingIndex = Readonly<{\n  metric: EmbeddingMetric;\n  indexType: EmbeddingIndexType;\n  m: number;\n  efConstruction: number;\n  lists: number | undefined;\n}>;\n\nconst DEFAULT_HNSW_M = 16;\nconst DEFAULT_HNSW_EF_CONSTRUCTION = 64;\n\n// ============================================================\n// Embedding Schema Type\n// ============================================================\n\n/**\n * A Zod schema for vector embeddings with attached dimension metadata.\n * Uses the branded EmbeddingValue type for type-level distinction.\n */\nexport type EmbeddingSchema<D extends number = number> =\n  z.ZodType<EmbeddingValue> &\n    Readonly<{\n      [EMBEDDING_DIMENSIONS_KEY]: D;\n      [EMBEDDING_INDEX_KEY]: ResolvedEmbeddingIndex;\n    }>;\n\n// ============================================================\n// Embedding Factory\n// ============================================================\n\n/**\n * Creates a Zod schema for vector embeddings.\n *\n * The dimension is validated at runtime and attached as metadata\n * for the schema introspector and query compiler.\n *\n * @param dimensions - The number of dimensions (e.g., 384, 512, 768, 1536, 3072)\n * @returns A Zod schema that validates embedding arrays\n *\n * @example\n * ```typescript\n * // OpenAI ada-002 embeddings\n * const Document = defineNode(\"Document\", {\n *   schema: z.object({\n *     title: z.string(),\n *     embedding: embedding(1536),\n *   }),\n * });\n *\n * // Sentence transformers\n * const Sentence = defineNode(\"Sentence\", {\n *   schema: z.object({\n *     text: z.string(),\n *     embedding: embedding(384), // all-MiniLM-L6-v2\n *   }),\n * });\n *\n * // Optional embeddings are supported\n * const Article = defineNode(\"Article\", {\n *   schema: z.object({\n *     content: z.string(),\n *     embedding: embedding(1536).optional(),\n *   }),\n * });\n * ```\n */\nexport function embedding<D extends number>(\n  dimensions: D,\n  options: EmbeddingIndexOptions = {},\n): EmbeddingSchema<D> {\n  if (!Number.isInteger(dimensions) || dimensions <= 0) {\n    throw new Error(\n      `Embedding dimensions must be a positive integer, got: ${dimensions}`,\n    );\n  }\n\n  const indexConfig = resolveEmbeddingIndex(options);\n\n  // Use EmbeddingValue as the output type for proper type branding.\n  // At runtime, validation accepts any number array with correct dimensions.\n  const schema = z.custom<EmbeddingValue>(\n    (value): value is EmbeddingValue => {\n      if (!Array.isArray(value)) {\n        return false;\n      }\n      if (value.length !== dimensions) {\n        return false;\n      }\n      return value.every((n) => typeof n === \"number\" && Number.isFinite(n));\n    },\n    {\n      message: `Expected an array of ${dimensions} finite numbers`,\n    },\n  );\n\n  return Object.assign(schema, {\n    [EMBEDDING_DIMENSIONS_KEY]: dimensions,\n    [EMBEDDING_INDEX_KEY]: indexConfig,\n  });\n}\n\nfunction resolveEmbeddingIndex(\n  options: EmbeddingIndexOptions,\n): ResolvedEmbeddingIndex {\n  const indexType = options.indexType ?? DEFAULT_EMBEDDING_INDEX_TYPE;\n  const m = options.m ?? DEFAULT_HNSW_M;\n  const efConstruction = options.efConstruction ?? DEFAULT_HNSW_EF_CONSTRUCTION;\n\n  if (m <= 0 || !Number.isInteger(m)) {\n    throw new Error(\n      `embedding() index option 'm' must be a positive integer, got: ${m}`,\n    );\n  }\n  if (efConstruction <= 0 || !Number.isInteger(efConstruction)) {\n    throw new Error(\n      `embedding() index option 'efConstruction' must be a positive integer, got: ${efConstruction}`,\n    );\n  }\n  if (\n    options.lists !== undefined &&\n    (options.lists <= 0 || !Number.isInteger(options.lists))\n  ) {\n    throw new Error(\n      `embedding() index option 'lists' must be a positive integer, got: ${options.lists}`,\n    );\n  }\n\n  return Object.freeze({\n    metric: options.metric ?? DEFAULT_EMBEDDING_METRIC,\n    indexType,\n    m,\n    efConstruction,\n    lists: options.lists,\n  });\n}\n\n// ============================================================\n// Type Guards\n// ============================================================\n\n/**\n * Checks if a value is an embedding schema.\n */\nexport function isEmbeddingSchema(value: unknown): value is EmbeddingSchema {\n  return (\n    typeof value === \"object\" &&\n    value !== null &&\n    EMBEDDING_DIMENSIONS_KEY in value &&\n    typeof (value as Record<string, unknown>)[EMBEDDING_DIMENSIONS_KEY] ===\n      \"number\"\n  );\n}\n\n/**\n * Gets the dimensions from an embedding schema.\n * Returns undefined if the schema is not an embedding schema.\n */\nexport function getEmbeddingDimensions(schema: z.ZodType): number | undefined {\n  if (isEmbeddingSchema(schema)) {\n    return schema[EMBEDDING_DIMENSIONS_KEY];\n  }\n  return undefined;\n}\n\n/**\n * Gets the resolved index configuration from an embedding schema.\n * Returns undefined if the schema is not an embedding schema. Used by\n * the auto-derivation pass at `defineGraph()` time to build vector\n * `IndexDeclaration` entries.\n */\nexport function getEmbeddingIndex(\n  schema: z.ZodType,\n): ResolvedEmbeddingIndex | undefined {\n  if (!isEmbeddingSchema(schema)) return undefined;\n  const candidate = (schema as unknown as Record<string, unknown>)[\n    EMBEDDING_INDEX_KEY\n  ];\n  if (candidate === undefined) return undefined;\n  return candidate as ResolvedEmbeddingIndex;\n}\n\n// ============================================================\n// Embedding Field Resolution\n// ============================================================\n\n/**\n * One resolved embedding field on a node schema — its dot-path plus the\n * fixed `dimensions` and the resolved index `(metric, indexType)`.\n * Sourced from the field's `embedding()` brand, unwrapping `.optional()`\n * / `.nullable()` / `.default()` / `.readonly()` wrappers. The single\n * source of truth shared by embedding sync (write path), the query\n * compiler's `field.similarTo(...)` slot resolution, and search.\n */\nexport type ResolvedEmbeddingField = Readonly<{\n  fieldPath: string;\n  dimensions: number;\n  metric: EmbeddingMetric;\n  indexType: EmbeddingIndexType;\n}>;\n\n/**\n * The single source of truth for an embedding field's default metric / index\n * type when unspecified — used by `embedding()` declaration resolution, the\n * field-fallback below, and the legacy-migration slot defaults\n * (`migrate-vectors.ts`) so the three never drift.\n */\nexport const DEFAULT_EMBEDDING_METRIC: EmbeddingMetric = \"cosine\";\nexport const DEFAULT_EMBEDDING_INDEX_TYPE: EmbeddingIndexType = \"hnsw\";\n\n/**\n * Extracts every top-level embedding field from a node's object schema.\n * Returns `[]` for non-object schemas or schemas with no embedding\n * fields. Embeddings nested inside object properties are out of scope by\n * design (storage keys at the top-level `(kind, fieldPath)` grain).\n */\nexport function resolveEmbeddingFields(\n  schema: z.ZodType,\n): readonly ResolvedEmbeddingField[] {\n  if (schema.type !== \"object\") return [];\n  const shape = (schema.def as { shape?: Record<string, z.ZodType> }).shape;\n  if (!shape) return [];\n\n  const fields: ResolvedEmbeddingField[] = [];\n  for (const [fieldPath, fieldSchema] of Object.entries(shape)) {\n    const dimensions = resolveEmbeddingDimensions(fieldSchema);\n    if (dimensions === undefined) continue;\n    const index = resolveEmbeddingIndexConfig(fieldSchema);\n    fields.push({\n      fieldPath,\n      dimensions,\n      metric: index?.metric ?? DEFAULT_EMBEDDING_METRIC,\n      indexType: index?.indexType ?? DEFAULT_EMBEDDING_INDEX_TYPE,\n    });\n  }\n  return fields;\n}\n\nfunction resolveEmbeddingDimensions(schema: z.ZodType): number | undefined {\n  const direct = getEmbeddingDimensions(schema);\n  if (direct !== undefined) return direct;\n  const unwrapped = unwrapToEmbedding(schema);\n  return unwrapped === undefined ? undefined : (\n      getEmbeddingDimensions(unwrapped)\n    );\n}\n\nfunction resolveEmbeddingIndexConfig(\n  schema: z.ZodType,\n): ResolvedEmbeddingIndex | undefined {\n  const direct = getEmbeddingIndex(schema);\n  if (direct !== undefined) return direct;\n  const unwrapped = unwrapToEmbedding(schema);\n  return unwrapped === undefined ? undefined : getEmbeddingIndex(unwrapped);\n}\n\n/**\n * Unwraps optional / nullable / default / readonly wrappers to find the\n * inner embedding schema, recursing through chained wrappers. Returns\n * `undefined` when no embedding schema is reachable.\n */\nfunction unwrapToEmbedding(schema: z.ZodType): z.ZodType | undefined {\n  const type = schema.type;\n  const innerType = (schema.def as { innerType?: z.ZodType }).innerType;\n  if (\n    (type === \"optional\" ||\n      type === \"nullable\" ||\n      type === \"default\" ||\n      type === \"readonly\") &&\n    innerType\n  ) {\n    if (isEmbeddingSchema(innerType)) return innerType;\n    return unwrapToEmbedding(innerType);\n  }\n  return undefined;\n}\n"]}

package/dist/chunk-4K6IHUW6.cjs

@@ -0,0 +1,183 @@
+'use strict';
+
+var chunkC2RWFIYP_cjs = require('./chunk-C2RWFIYP.cjs');
+var drizzleOrm = require('drizzle-orm');
+
+var TABLE_PREFIX = "tg_vec";
+var SQLITE_VEC_MAX_DIMENSIONS = 8e3;
+var SQLITE_VEC_CAPABILITIES = {
+  supported: true,
+  metrics: ["cosine", "l2"],
+  // vec0 always indexes inline, so the portable ANN intent ("hnsw", the
+  // `embedding()` default) is honored by real KNN; "none" stays brute-force.
+  // Both map to the same vec0 table — only `buildSearch` differs.
+  indexTypes: ["hnsw", "none"],
+  maxDimensions: SQLITE_VEC_MAX_DIMENSIONS
+};
+function usesAnnIndex(slot) {
+  return slot.indexType === "hnsw";
+}
+function sqliteVecMetricToken(metric) {
+  switch (metric) {
+    case "cosine": {
+      return "cosine";
+    }
+    case "l2": {
+      return "l2";
+    }
+    case "inner_product": {
+      throw new Error(
+        "sqlite-vec does not support inner_product. Use 'cosine' or 'l2'."
+      );
+    }
+    default: {
+      const _exhaustive = metric;
+      throw new Error(`Unsupported vector metric: ${String(_exhaustive)}`);
+    }
+  }
+}
+function vecF32Literal(embedding, name) {
+  chunkC2RWFIYP_cjs.assertFiniteEmbedding(embedding, name);
+  return drizzleOrm.sql`vec_f32(${JSON.stringify(embedding)})`;
+}
+function distanceExpression(embeddingColumn, queryEmbedding, metric) {
+  const query = vecF32Literal(queryEmbedding, "queryEmbedding");
+  switch (metric) {
+    case "cosine": {
+      return drizzleOrm.sql`vec_distance_cosine(${embeddingColumn}, ${query})`;
+    }
+    case "l2": {
+      return drizzleOrm.sql`vec_distance_l2(${embeddingColumn}, ${query})`;
+    }
+    case "inner_product": {
+      throw new Error(
+        "sqlite-vec does not support inner_product. Use 'cosine' or 'l2'."
+      );
+    }
+    default: {
+      const _exhaustive = metric;
+      throw new Error(`Unsupported vector metric: ${String(_exhaustive)}`);
+    }
+  }
+}
+var sqliteVecStrategy = {
+  name: "sqlite-vec",
+  capabilities: SQLITE_VEC_CAPABILITIES,
+  tableName(graphId, nodeKind, fieldPath) {
+    return chunkC2RWFIYP_cjs.vectorPhysicalName(TABLE_PREFIX, graphId, nodeKind, fieldPath);
+  },
+  ownedTables(slot) {
+    const table = this.tableName(slot.graphId, slot.nodeKind, slot.fieldPath);
+    const name = chunkC2RWFIYP_cjs.quoteIdentifier(table);
+    const metric = sqliteVecMetricToken(slot.metric);
+    return [
+      {
+        logicalName: `${chunkC2RWFIYP_cjs.VECTOR_CONTRIBUTION_PREFIX}:${slot.nodeKind}.${slot.fieldPath}`,
+        owner: "sqlite-vec",
+        tableName: table,
+        createDdl: [
+          `CREATE VIRTUAL TABLE IF NOT EXISTS ${name} USING vec0(
+  node_id TEXT PRIMARY KEY,
+  graph_id TEXT PARTITION KEY,
+  +created_at TEXT,
+  +updated_at TEXT,
+  embedding float[${slot.dimensions}] distance_metric=${metric}
+);`
+        ],
+        runtimeEnsure: true
+      }
+    ];
+  },
+  buildUpsert(slot, params, timestamp) {
+    const table = chunkC2RWFIYP_cjs.quotedTableName(
+      this.tableName(slot.graphId, slot.nodeKind, slot.fieldPath)
+    );
+    const value = vecF32Literal(params.embedding, "embedding");
+    return [
+      drizzleOrm.sql`
+        DELETE FROM ${table}
+        WHERE "node_id" = ${params.nodeId} AND "graph_id" = ${params.graphId}
+      `,
+      drizzleOrm.sql`
+        INSERT INTO ${table} ("node_id", "graph_id", "created_at", "updated_at", "embedding")
+        VALUES (${params.nodeId}, ${params.graphId}, ${timestamp}, ${timestamp}, ${value})
+      `
+    ];
+  },
+  buildDelete(slot, params) {
+    const table = chunkC2RWFIYP_cjs.quotedTableName(
+      this.tableName(slot.graphId, slot.nodeKind, slot.fieldPath)
+    );
+    return [
+      drizzleOrm.sql`
+        DELETE FROM ${table}
+        WHERE "node_id" = ${params.nodeId} AND "graph_id" = ${params.graphId}
+      `
+    ];
+  },
+  buildSearch(slot, params) {
+    const table = chunkC2RWFIYP_cjs.quotedTableName(
+      this.tableName(slot.graphId, slot.nodeKind, slot.fieldPath)
+    );
+    const embeddingColumn = drizzleOrm.sql`${table}."embedding"`;
+    const distance = distanceExpression(
+      embeddingColumn,
+      params.queryEmbedding,
+      params.metric
+    );
+    const score = chunkC2RWFIYP_cjs.vectorScoreExpression(distance, params.metric);
+    if (usesAnnIndex(slot)) {
+      const query = vecF32Literal(params.queryEmbedding, "queryEmbedding");
+      const conditions2 = [
+        drizzleOrm.sql`${table}."embedding" MATCH ${query}`,
+        drizzleOrm.sql`k = ${params.limit}`,
+        drizzleOrm.sql`${table}."graph_id" = ${params.graphId}`
+      ];
+      if (params.minScore !== void 0) {
+        conditions2.push(
+          chunkC2RWFIYP_cjs.vectorMinScoreCondition(distance, params.metric, params.minScore)
+        );
+      }
+      return drizzleOrm.sql`
+        SELECT ${table}."node_id" AS node_id, ${score} AS score
+        FROM ${table}
+        WHERE ${drizzleOrm.sql.join(conditions2, drizzleOrm.sql` AND `)}
+        ORDER BY distance ASC
+      `;
+    }
+    const conditions = [drizzleOrm.sql`${table}."graph_id" = ${params.graphId}`];
+    if (params.minScore !== void 0) {
+      conditions.push(
+        chunkC2RWFIYP_cjs.vectorMinScoreCondition(distance, params.metric, params.minScore)
+      );
+    }
+    return drizzleOrm.sql`
+      SELECT ${table}."node_id" AS node_id, ${score} AS score
+      FROM ${table}
+      WHERE ${drizzleOrm.sql.join(conditions, drizzleOrm.sql` AND `)}
+      ORDER BY ${distance} ASC
+      LIMIT ${params.limit}
+    `;
+  },
+  distanceExpression(embeddingColumn, queryEmbedding, metric) {
+    return distanceExpression(embeddingColumn, queryEmbedding, metric);
+  },
+  // vec0 indexes inline (the virtual table is the index), so there is no
+  // standalone ANN index to create or drop — both return undefined.
+  buildCreateIndex() {
+    return void 0;
+  },
+  buildDropIndex() {
+    return void 0;
+  },
+  buildDropStorage(slot) {
+    const table = chunkC2RWFIYP_cjs.quoteIdentifier(
+      this.tableName(slot.graphId, slot.nodeKind, slot.fieldPath)
+    );
+    return [`DROP TABLE IF EXISTS ${table}`];
+  }
+};
+
+exports.sqliteVecStrategy = sqliteVecStrategy;
+//# sourceMappingURL=chunk-4K6IHUW6.cjs.map
+//# sourceMappingURL=chunk-4K6IHUW6.cjs.map

package/dist/chunk-4K6IHUW6.cjs.map

@@ -0,0 +1 @@
+{"version":3,"sources":["../src/query/dialect/vector/sqlite-vec-strategy.ts"],"names":["assertFiniteEmbedding","sql","vectorPhysicalName","quoteIdentifier","VECTOR_CONTRIBUTION_PREFIX","quotedTableName","vectorScoreExpression","conditions","vectorMinScoreCondition"],"mappings":";;;;;AA6EA,IAAM,YAAA,GAAe,QAAA;AAOrB,IAAM,yBAAA,GAA4B,GAAA;AAElC,IAAM,uBAAA,GAA8C;AAAA,EAClD,SAAA,EAAW,IAAA;AAAA,EACX,OAAA,EAAS,CAAC,QAAA,EAAU,IAAI,CAAA;AAAA;AAAA;AAAA;AAAA,EAIxB,UAAA,EAAY,CAAC,MAAA,EAAQ,MAAM,CAAA;AAAA,EAC3B,aAAA,EAAe;AACjB,CAAA;AAGA,SAAS,aAAa,IAAA,EAA2B;AAC/C,EAAA,OAAO,KAAK,SAAA,KAAc,MAAA;AAC5B;AAGA,SAAS,qBAAqB,MAAA,EAA8B;AAC1D,EAAA,QAAQ,MAAA;AAAQ,IACd,KAAK,QAAA,EAAU;AACb,MAAA,OAAO,QAAA;AAAA,IACT;AAAA,IACA,KAAK,IAAA,EAAM;AACT,MAAA,OAAO,IAAA;AAAA,IACT;AAAA,IACA,KAAK,eAAA,EAAiB;AACpB,MAAA,MAAM,IAAI,KAAA;AAAA,QACR;AAAA,OACF;AAAA,IACF;AAAA,IACA,SAAS;AACP,MAAA,MAAM,WAAA,GAAqB,MAAA;AAC3B,MAAA,MAAM,IAAI,KAAA,CAAM,CAAA,2BAAA,EAA8B,MAAA,CAAO,WAAW,CAAC,CAAA,CAAE,CAAA;AAAA,IACrE;AAAA;AAEJ;AAQA,SAAS,aAAA,CAAc,WAA8B,IAAA,EAAmB;AACtE,EAAAA,uCAAA,CAAsB,WAAW,IAAI,CAAA;AACrC,EAAA,OAAOC,cAAA,CAAA,QAAA,EAAc,IAAA,CAAK,SAAA,CAAU,SAAS,CAAC,CAAA,CAAA,CAAA;AAChD;AAEA,SAAS,kBAAA,CACP,eAAA,EACA,cAAA,EACA,MAAA,EACK;AACL,EAAA,MAAM,KAAA,GAAQ,aAAA,CAAc,cAAA,EAAgB,gBAAgB,CAAA;AAC5D,EAAA,QAAQ,MAAA;AAAQ,IACd,KAAK,QAAA,EAAU;AACb,MAAA,OAAOA,cAAA,CAAA,oBAAA,EAA0B,eAAe,CAAA,EAAA,EAAK,KAAK,CAAA,CAAA,CAAA;AAAA,IAC5D;AAAA,IACA,KAAK,IAAA,EAAM;AACT,MAAA,OAAOA,cAAA,CAAA,gBAAA,EAAsB,eAAe,CAAA,EAAA,EAAK,KAAK,CAAA,CAAA,CAAA;AAAA,IACxD;AAAA,IACA,KAAK,eAAA,EAAiB;AACpB,MAAA,MAAM,IAAI,KAAA;AAAA,QACR;AAAA,OACF;AAAA,IACF;AAAA,IACA,SAAS;AACP,MAAA,MAAM,WAAA,GAAqB,MAAA;AAC3B,MAAA,MAAM,IAAI,KAAA,CAAM,CAAA,2BAAA,EAA8B,MAAA,CAAO,WAAW,CAAC,CAAA,CAAE,CAAA;AAAA,IACrE;AAAA;AAEJ;AAEO,IAAM,iBAAA,GAAoC;AAAA,EAC/C,IAAA,EAAM,YAAA;AAAA,EACN,YAAA,EAAc,uBAAA;AAAA,EAEd,SAAA,CAAU,OAAA,EAAS,QAAA,EAAU,SAAA,EAAW;AACtC,IAAA,OAAOC,oCAAA,CAAmB,YAAA,EAAc,OAAA,EAAS,QAAA,EAAU,SAAS,CAAA;AAAA,EACtE,CAAA;AAAA,EAEA,YAAY,IAAA,EAA4C;AACtD,IAAA,MAAM,KAAA,GAAQ,KAAK,SAAA,CAAU,IAAA,CAAK,SAAS,IAAA,CAAK,QAAA,EAAU,KAAK,SAAS,CAAA;AACxE,IAAA,MAAM,IAAA,GAAOC,kCAAgB,KAAK,CAAA;AAClC,IAAA,MAAM,MAAA,GAAS,oBAAA,CAAqB,IAAA,CAAK,MAAM,CAAA;AAO/C,IAAA,OAAO;AAAA,MACL;AAAA,QACE,WAAA,EAAa,GAAGC,4CAA0B,CAAA,CAAA,EAAI,KAAK,QAAQ,CAAA,CAAA,EAAI,KAAK,SAAS,CAAA,CAAA;AAAA,QAC7E,KAAA,EAAO,YAAA;AAAA,QACP,SAAA,EAAW,KAAA;AAAA,QACX,SAAA,EAAW;AAAA,UACT,sCAAsC,IAAI,CAAA;AAAA;AAAA;AAAA;AAAA;AAAA,kBAAA,EAKhC,IAAA,CAAK,UAAU,CAAA,kBAAA,EAAqB,MAAM;AAAA,EAAA;AAAA,SAEtD;AAAA,QACA,aAAA,EAAe;AAAA;AACjB,KACF;AAAA,EACF,CAAA;AAAA,EAEA,WAAA,CAAY,IAAA,EAAM,MAAA,EAA+B,SAAA,EAA2B;AAC1E,IAAA,MAAM,KAAA,GAAQC,iCAAA;AAAA,MACZ,KAAK,SAAA,CAAU,IAAA,CAAK,SAAS,IAAA,CAAK,QAAA,EAAU,KAAK,SAAS;AAAA,KAC5D;AACA,IAAA,MAAM,KAAA,GAAQ,aAAA,CAAc,MAAA,CAAO,SAAA,EAAW,WAAW,CAAA;AAIzD,IAAA,OAAO;AAAA,MACLJ,cAAA;AAAA,oBAAA,EACgB,KAAK;AAAA,0BAAA,EACC,MAAA,CAAO,MAAM,CAAA,kBAAA,EAAqB,MAAA,CAAO,OAAO;AAAA,MAAA,CAAA;AAAA,MAEtEA,cAAA;AAAA,oBAAA,EACgB,KAAK,CAAA;AAAA,gBAAA,EACT,MAAA,CAAO,MAAM,CAAA,EAAA,EAAK,MAAA,CAAO,OAAO,KAAK,SAAS,CAAA,EAAA,EAAK,SAAS,CAAA,EAAA,EAAK,KAAK,CAAA;AAAA,MAAA;AAAA,KAEpF;AAAA,EACF,CAAA;AAAA,EAEA,WAAA,CAAY,MAAM,MAAA,EAA+C;AAC/D,IAAA,MAAM,KAAA,GAAQI,iCAAA;AAAA,MACZ,KAAK,SAAA,CAAU,IAAA,CAAK,SAAS,IAAA,CAAK,QAAA,EAAU,KAAK,SAAS;AAAA,KAC5D;AACA,IAAA,OAAO;AAAA,MACLJ,cAAA;AAAA,oBAAA,EACgB,KAAK;AAAA,0BAAA,EACC,MAAA,CAAO,MAAM,CAAA,kBAAA,EAAqB,MAAA,CAAO,OAAO;AAAA,MAAA;AAAA,KAExE;AAAA,EACF,CAAA;AAAA,EAEA,WAAA,CAAY,MAAM,MAAA,EAAiC;AACjD,IAAA,MAAM,KAAA,GAAQI,iCAAA;AAAA,MACZ,KAAK,SAAA,CAAU,IAAA,CAAK,SAAS,IAAA,CAAK,QAAA,EAAU,KAAK,SAAS;AAAA,KAC5D;AACA,IAAA,MAAM,eAAA,GAAkBJ,iBAAM,KAAK,CAAA,YAAA,CAAA;AACnC,IAAA,MAAM,QAAA,GAAW,kBAAA;AAAA,MACf,eAAA;AAAA,MACA,MAAA,CAAO,cAAA;AAAA,MACP,MAAA,CAAO;AAAA,KACT;AACA,IAAA,MAAM,KAAA,GAAQK,uCAAA,CAAsB,QAAA,EAAU,MAAA,CAAO,MAAM,CAAA;AAE3D,IAAA,IAAI,YAAA,CAAa,IAAI,CAAA,EAAG;AAMtB,MAAA,MAAM,KAAA,GAAQ,aAAA,CAAc,MAAA,CAAO,cAAA,EAAgB,gBAAgB,CAAA;AACnE,MAAA,MAAMC,WAAAA,GAAoB;AAAA,QACxBN,cAAA,CAAA,EAAM,KAAK,CAAA,mBAAA,EAAsB,KAAK,CAAA,CAAA;AAAA,QACtCA,cAAA,CAAA,IAAA,EAAU,OAAO,KAAK,CAAA,CAAA;AAAA,QACtBA,cAAA,CAAA,EAAM,KAAK,CAAA,cAAA,EAAiB,MAAA,CAAO,OAAO,CAAA;AAAA,OAC5C;AACA,MAAA,IAAI,MAAA,CAAO,aAAa,MAAA,EAAW;AACjC,QAAAM,WAAAA,CAAW,IAAA;AAAA,UACTC,yCAAA,CAAwB,QAAA,EAAU,MAAA,CAAO,MAAA,EAAQ,OAAO,QAAQ;AAAA,SAClE;AAAA,MACF;AACA,MAAA,OAAOP,cAAA;AAAA,eAAA,EACI,KAAK,0BAA0B,KAAK,CAAA;AAAA,aAAA,EACtC,KAAK;AAAA,cAAA,EACJA,cAAA,CAAI,IAAA,CAAKM,WAAAA,EAAYN,cAAA,CAAA,KAAA,CAAU,CAAC;AAAA;AAAA,MAAA,CAAA;AAAA,IAG5C;AAGA,IAAA,MAAM,aAAoB,CAACA,cAAA,CAAA,EAAM,KAAK,CAAA,cAAA,EAAiB,MAAA,CAAO,OAAO,CAAA,CAAE,CAAA;AACvE,IAAA,IAAI,MAAA,CAAO,aAAa,MAAA,EAAW;AACjC,MAAA,UAAA,CAAW,IAAA;AAAA,QACTO,yCAAA,CAAwB,QAAA,EAAU,MAAA,CAAO,MAAA,EAAQ,OAAO,QAAQ;AAAA,OAClE;AAAA,IACF;AACA,IAAA,OAAOP,cAAA;AAAA,aAAA,EACI,KAAK,0BAA0B,KAAK,CAAA;AAAA,WAAA,EACtC,KAAK;AAAA,YAAA,EACJA,cAAA,CAAI,IAAA,CAAK,UAAA,EAAYA,cAAA,CAAA,KAAA,CAAU,CAAC;AAAA,eAAA,EAC7B,QAAQ,CAAA;AAAA,YAAA,EACX,OAAO,KAAK;AAAA,IAAA,CAAA;AAAA,EAExB,CAAA;AAAA,EAEA,kBAAA,CAAmB,eAAA,EAAiB,cAAA,EAAgB,MAAA,EAAQ;AAC1D,IAAA,OAAO,kBAAA,CAAmB,eAAA,EAAiB,cAAA,EAAgB,MAAM,CAAA;AAAA,EACnE,CAAA;AAAA;AAAA;AAAA,EAIA,gBAAA,GAAoC;AAClC,IAAA,OAAO,MAAA;AAAA,EACT,CAAA;AAAA,EAEA,cAAA,GAAkC;AAChC,IAAA,OAAO,MAAA;AAAA,EACT,CAAA;AAAA,EAEA,iBAAiB,IAAA,EAAyB;AACxC,IAAA,MAAM,KAAA,GAAQE,iCAAA;AAAA,MACZ,KAAK,SAAA,CAAU,IAAA,CAAK,SAAS,IAAA,CAAK,QAAA,EAAU,KAAK,SAAS;AAAA,KAC5D;AAEA,IAAA,OAAO,CAAC,CAAA,qBAAA,EAAwB,KAAK,CAAA,CAAE,CAAA;AAAA,EACzC;AACF","file":"chunk-4K6IHUW6.cjs","sourcesContent":["/**\n * sqlite-vec strategy — real `vec0` KNN for local SQLite.\n *\n * This upgrades `createLocalSqliteBackend` (better-sqlite3 + the sqlite-vec\n * extension) from the legacy brute-force-on-a-shared-table model to genuine\n * approximate-nearest-neighbor search on per-`(nodeKind, fieldPath)` storage.\n *\n * ## Storage choice: a `vec0` virtual table per field\n *\n * Each slot is backed by its own `vec0` virtual table:\n *\n * ```sql\n * CREATE VIRTUAL TABLE tg_vec_<kind>_<field> USING vec0(\n *   node_id TEXT PRIMARY KEY,\n *   graph_id TEXT PARTITION KEY,\n *   +created_at TEXT,\n *   +updated_at TEXT,\n *   embedding float[<N>] distance_metric=<cosine|l2>\n * );\n * ```\n *\n * Verified against the installed `sqlite-vec` `v0.1.9`:\n *\n * - **vec0 over a plain `vec_f32` BLOB column.** The shipped option (a) — a\n *   `vec0` virtual table — is chosen over option (b) — a regular table whose\n *   `vec_f32` BLOB column is brute-forced — because vec0 is the *only* form\n *   that gives local SQLite a real ANN index (today's code only ever\n *   brute-forces). vec0 indexes inline (the virtual table *is* the index), so\n *   there is no separate `CREATE INDEX` step.\n * - **KNN with partition-correct filtering.** `WHERE embedding MATCH vec_f32(q)\n *   AND k = ? AND graph_id = ?` returns the nearest `k`, ordered by `distance`,\n *   scoped to one graph — a near vector in another `graph_id` partition does\n *   not leak. This is strictly better than libSQL's table-global `vector_top_k`\n *   (which over-fetches for multi-graph correctness): vec0's `PARTITION KEY`\n *   makes the per-graph KNN exact.\n * - **One table serves both read paths.** The same `embedding` column answers\n *   the compiler's brute-force `distanceExpression` (`vec_distance_cosine` /\n *   `vec_distance_l2` over the column, `ORDER BY … LIMIT`, no `MATCH`), so the\n *   `.where(field.similarTo(...))` CTE path needs no second structure.\n * - **`k` and `LIMIT` are mutually exclusive** in a vec0 `MATCH` query, so the\n *   ANN path uses `k = <limit>` with no `LIMIT`, and the brute-force path uses\n *   `LIMIT <limit>` with no `MATCH`.\n * - **No `ON CONFLICT` / `INSERT OR REPLACE`** on a vec0 primary key, so upsert\n *   is `DELETE` + `INSERT` (atomic under the caller's outer transaction),\n *   mirroring the FTS5 strategy.\n *\n * Metric support is `cosine` + `l2` (no `inner_product` — sqlite-vec has no\n * `vec_distance_ip`), advertised as data on `capabilities`. The table's\n * `distance_metric` is fixed at creation from the slot's metric, alongside its\n * fixed dimension `N`.\n *\n * The libSQL and pgvector siblings implement the same `VectorStrategy`\n * contract for their engines.\n */\nimport { type SQL, sql } from \"drizzle-orm\";\n\nimport { quotedTableName } from \"../../../backend/drizzle/operations/shared\";\nimport { type StrategyTableContribution } from \"../../../backend/table-contribution\";\nimport {\n  type DeleteEmbeddingParams,\n  type UpsertEmbeddingParams,\n  type VectorCapabilities,\n  type VectorMetric,\n  type VectorSearchParams,\n} from \"../../../backend/types\";\nimport {\n  assertFiniteEmbedding,\n  quoteIdentifier,\n  VECTOR_CONTRIBUTION_PREFIX,\n  vectorMinScoreCondition,\n  vectorPhysicalName,\n  vectorScoreExpression,\n  type VectorSlot,\n  type VectorStrategy,\n} from \"../vector-strategy\";\n\n/** Physical-name prefix for the per-field `vec0` virtual table. */\nconst TABLE_PREFIX = \"tg_vec\";\n\n/**\n * sqlite-vec stores each dimension as a 32-bit float; `float[N]` accepts\n * large `N`, but we advertise the same conservative ceiling the legacy\n * SQLite path used so capability discovery is stable across the swap.\n */\nconst SQLITE_VEC_MAX_DIMENSIONS = 8000;\n\nconst SQLITE_VEC_CAPABILITIES: VectorCapabilities = {\n  supported: true,\n  metrics: [\"cosine\", \"l2\"],\n  // vec0 always indexes inline, so the portable ANN intent (\"hnsw\", the\n  // `embedding()` default) is honored by real KNN; \"none\" stays brute-force.\n  // Both map to the same vec0 table — only `buildSearch` differs.\n  indexTypes: [\"hnsw\", \"none\"],\n  maxDimensions: SQLITE_VEC_MAX_DIMENSIONS,\n};\n\n/** Whether a slot's declared index type maps to vec0's `MATCH … k =` KNN. */\nfunction usesAnnIndex(slot: VectorSlot): boolean {\n  return slot.indexType === \"hnsw\";\n}\n\n/** vec0 `distance_metric` token accepted in the virtual-table definition. */\nfunction sqliteVecMetricToken(metric: VectorMetric): string {\n  switch (metric) {\n    case \"cosine\": {\n      return \"cosine\";\n    }\n    case \"l2\": {\n      return \"l2\";\n    }\n    case \"inner_product\": {\n      throw new Error(\n        \"sqlite-vec does not support inner_product. Use 'cosine' or 'l2'.\",\n      );\n    }\n    default: {\n      const _exhaustive: never = metric;\n      throw new Error(`Unsupported vector metric: ${String(_exhaustive)}`);\n    }\n  }\n}\n\n/**\n * Formats an embedding as the bound JSON argument to `vec_f32(...)`. The\n * value rides as a parameter (not interpolated) — `vec_f32` parses the JSON\n * array into the packed float blob. Finiteness is validated first so a\n * NaN/Infinity names its index before JSON.stringify masks it as `null`.\n */\nfunction vecF32Literal(embedding: readonly number[], name: string): SQL {\n  assertFiniteEmbedding(embedding, name);\n  return sql`vec_f32(${JSON.stringify(embedding)})`;\n}\n\nfunction distanceExpression(\n  embeddingColumn: SQL,\n  queryEmbedding: readonly number[],\n  metric: VectorMetric,\n): SQL {\n  const query = vecF32Literal(queryEmbedding, \"queryEmbedding\");\n  switch (metric) {\n    case \"cosine\": {\n      return sql`vec_distance_cosine(${embeddingColumn}, ${query})`;\n    }\n    case \"l2\": {\n      return sql`vec_distance_l2(${embeddingColumn}, ${query})`;\n    }\n    case \"inner_product\": {\n      throw new Error(\n        \"sqlite-vec does not support inner_product. Use 'cosine' or 'l2'.\",\n      );\n    }\n    default: {\n      const _exhaustive: never = metric;\n      throw new Error(`Unsupported vector metric: ${String(_exhaustive)}`);\n    }\n  }\n}\n\nexport const sqliteVecStrategy: VectorStrategy = {\n  name: \"sqlite-vec\",\n  capabilities: SQLITE_VEC_CAPABILITIES,\n\n  tableName(graphId, nodeKind, fieldPath) {\n    return vectorPhysicalName(TABLE_PREFIX, graphId, nodeKind, fieldPath);\n  },\n\n  ownedTables(slot): readonly StrategyTableContribution[] {\n    const table = this.tableName(slot.graphId, slot.nodeKind, slot.fieldPath);\n    const name = quoteIdentifier(table);\n    const metric = sqliteVecMetricToken(slot.metric);\n\n    // vec0 virtual tables cannot be modeled as a Drizzle table, so this is a\n    // raw-ddl contribution — emitted verbatim and invisible to drizzle-kit.\n    // `graph_id` is a PARTITION KEY so per-graph KNN is exact; `created_at` /\n    // `updated_at` are auxiliary (`+`) columns (stored, not filtered on). The\n    // virtual table *is* the index — no separate CREATE INDEX is emitted.\n    return [\n      {\n        logicalName: `${VECTOR_CONTRIBUTION_PREFIX}:${slot.nodeKind}.${slot.fieldPath}`,\n        owner: \"sqlite-vec\",\n        tableName: table,\n        createDdl: [\n          `CREATE VIRTUAL TABLE IF NOT EXISTS ${name} USING vec0(\n  node_id TEXT PRIMARY KEY,\n  graph_id TEXT PARTITION KEY,\n  +created_at TEXT,\n  +updated_at TEXT,\n  embedding float[${slot.dimensions}] distance_metric=${metric}\n);`,\n        ],\n        runtimeEnsure: true,\n      },\n    ];\n  },\n\n  buildUpsert(slot, params: UpsertEmbeddingParams, timestamp): readonly SQL[] {\n    const table = quotedTableName(\n      this.tableName(slot.graphId, slot.nodeKind, slot.fieldPath),\n    );\n    const value = vecF32Literal(params.embedding, \"embedding\");\n    // vec0 rejects ON CONFLICT / INSERT OR REPLACE on its primary key —\n    // emulate upsert with DELETE + INSERT, atomic under the caller's\n    // outer transaction (mirrors the FTS5 fulltext strategy).\n    return [\n      sql`\n        DELETE FROM ${table}\n        WHERE \"node_id\" = ${params.nodeId} AND \"graph_id\" = ${params.graphId}\n      `,\n      sql`\n        INSERT INTO ${table} (\"node_id\", \"graph_id\", \"created_at\", \"updated_at\", \"embedding\")\n        VALUES (${params.nodeId}, ${params.graphId}, ${timestamp}, ${timestamp}, ${value})\n      `,\n    ];\n  },\n\n  buildDelete(slot, params: DeleteEmbeddingParams): readonly SQL[] {\n    const table = quotedTableName(\n      this.tableName(slot.graphId, slot.nodeKind, slot.fieldPath),\n    );\n    return [\n      sql`\n        DELETE FROM ${table}\n        WHERE \"node_id\" = ${params.nodeId} AND \"graph_id\" = ${params.graphId}\n      `,\n    ];\n  },\n\n  buildSearch(slot, params: VectorSearchParams): SQL {\n    const table = quotedTableName(\n      this.tableName(slot.graphId, slot.nodeKind, slot.fieldPath),\n    );\n    const embeddingColumn = sql`${table}.\"embedding\"`;\n    const distance = distanceExpression(\n      embeddingColumn,\n      params.queryEmbedding,\n      params.metric,\n    );\n    const score = vectorScoreExpression(distance, params.metric);\n\n    if (usesAnnIndex(slot)) {\n      // vec0 KNN. `k` and `LIMIT` are mutually exclusive in a MATCH query, so\n      // bound `k = limit` and omit LIMIT. `graph_id` filters by the partition\n      // key (exact, no recall loss). Order by vec0's own `distance` column —\n      // the recomputed `distance` expression is reused only for score/minScore\n      // so the math matches the brute-force path and the shared helpers.\n      const query = vecF32Literal(params.queryEmbedding, \"queryEmbedding\");\n      const conditions: SQL[] = [\n        sql`${table}.\"embedding\" MATCH ${query}`,\n        sql`k = ${params.limit}`,\n        sql`${table}.\"graph_id\" = ${params.graphId}`,\n      ];\n      if (params.minScore !== undefined) {\n        conditions.push(\n          vectorMinScoreCondition(distance, params.metric, params.minScore),\n        );\n      }\n      return sql`\n        SELECT ${table}.\"node_id\" AS node_id, ${score} AS score\n        FROM ${table}\n        WHERE ${sql.join(conditions, sql` AND `)}\n        ORDER BY distance ASC\n      `;\n    }\n\n    // Brute-force scan: no MATCH, so LIMIT is allowed (and required).\n    const conditions: SQL[] = [sql`${table}.\"graph_id\" = ${params.graphId}`];\n    if (params.minScore !== undefined) {\n      conditions.push(\n        vectorMinScoreCondition(distance, params.metric, params.minScore),\n      );\n    }\n    return sql`\n      SELECT ${table}.\"node_id\" AS node_id, ${score} AS score\n      FROM ${table}\n      WHERE ${sql.join(conditions, sql` AND `)}\n      ORDER BY ${distance} ASC\n      LIMIT ${params.limit}\n    `;\n  },\n\n  distanceExpression(embeddingColumn, queryEmbedding, metric) {\n    return distanceExpression(embeddingColumn, queryEmbedding, metric);\n  },\n\n  // vec0 indexes inline (the virtual table is the index), so there is no\n  // standalone ANN index to create or drop — both return undefined.\n  buildCreateIndex(): SQL | undefined {\n    return undefined;\n  },\n\n  buildDropIndex(): SQL | undefined {\n    return undefined;\n  },\n\n  buildDropStorage(slot): readonly string[] {\n    const table = quoteIdentifier(\n      this.tableName(slot.graphId, slot.nodeKind, slot.fieldPath),\n    );\n    // Dropping the vec0 virtual table removes its backing shadow tables too.\n    return [`DROP TABLE IF EXISTS ${table}`];\n  },\n};\n"]}

package/dist/{chunk-ZTREJOGB.js → chunk-7KLVFWXR.js}

@@ -401,7 +401,6 @@ var postgresDialect = {
   capabilities: {
     standardQueryStrategy: "cte_project",
     recursiveQueryStrategy: "recursive_cte",
-    setOperationStrategy: "standard_parenthesized",
     materializeIntermediateTraversalCtes: false,
     emitNotMaterializedHint: true,
     forceRecursiveWorktableOuterJoinOrder: false,
@@ -484,6 +483,12 @@ var postgresDialect = {
     return sql`${column} ILIKE ${pattern}`;
   },
   // ============================================================
+  // Set Operations
+  // ============================================================
+  wrapSetOperationOperand(inner) {
+    return sql`(${inner})`;
+  },
+  // ============================================================
   // Recursive CTE Path Operations
   // ============================================================
   initializePath(nodeId) {
@@ -519,36 +524,9 @@ var postgresDialect = {
   // ============================================================
   // Vector Operations
   // ============================================================
+  // Compile-time gate for `field.similarTo(...)`; the active
+  // `VectorStrategy` (pgvector) owns the distance SQL.
   supportsVectors: true,
-  vectorDistance(column, embedding, metric) {
-    const formatted = this.formatEmbedding(embedding);
-    switch (metric) {
-      case "cosine": {
-        return sql`(${column} <=> ${formatted})`;
-      }
-      case "l2": {
-        return sql`(${column} <-> ${formatted})`;
-      }
-      case "inner_product": {
-        return sql`(${column} <#> ${formatted})`;
-      }
-      default: {
-        const _exhaustive = metric;
-        throw new Error("Unsupported vector metric: " + String(_exhaustive));
-      }
-    }
-  },
-  formatEmbedding(embedding) {
-    for (const [index, value] of embedding.entries()) {
-      if (typeof value !== "number" || !Number.isFinite(value)) {
-        throw new TypeError(
-          `embedding[${index}] must be a finite number, got: ${value}`
-        );
-      }
-    }
-    const asString = `[${embedding.join(",")}]`;
-    return sql`${asString}::vector`;
-  },
   // ============================================================
   // Fulltext Operations
   // ============================================================
@@ -580,7 +558,6 @@ var sqliteDialect = {
   capabilities: {
     standardQueryStrategy: "cte_project",
     recursiveQueryStrategy: "recursive_cte",
-    setOperationStrategy: "sqlite_compound",
     materializeIntermediateTraversalCtes: true,
     emitNotMaterializedHint: false,
     forceRecursiveWorktableOuterJoinOrder: true,
@@ -665,6 +642,12 @@ var sqliteDialect = {
     return sql`LOWER(${column}) LIKE LOWER(${pattern})`;
   },
   // ============================================================
+  // Set Operations
+  // ============================================================
+  wrapSetOperationOperand(inner) {
+    return sql`SELECT * FROM (${inner})`;
+  },
+  // ============================================================
   // Recursive CTE Path Operations
   // ============================================================
   initializePath(nodeId) {
@@ -703,43 +686,105 @@ var sqliteDialect = {
   // ============================================================
   // Vector Operations
   // ============================================================
+  // Compile-time gate for `field.similarTo(...)`; the active
+  // `VectorStrategy` (sqlite-vec / libSQL-native) owns the distance SQL.
   supportsVectors: true,
-  vectorDistance(column, embedding, metric) {
-    const formatted = this.formatEmbedding(embedding);
-    switch (metric) {
-      case "cosine": {
-        return sql`vec_distance_cosine(${column}, ${formatted})`;
-      }
-      case "l2": {
-        return sql`vec_distance_l2(${column}, ${formatted})`;
-      }
-      case "inner_product": {
-        throw new Error(
-          "Inner product distance is not supported by sqlite-vec. Use 'cosine' or 'l2' metrics instead."
-        );
-      }
-      default: {
-        const _exhaustive = metric;
-        throw new Error("Unsupported vector metric: " + String(_exhaustive));
-      }
-    }
-  },
-  formatEmbedding(embedding) {
-    for (const [index, value] of embedding.entries()) {
-      if (typeof value !== "number" || !Number.isFinite(value)) {
-        throw new TypeError(
-          `embedding[${index}] must be a finite number, got: ${value}`
-        );
-      }
-    }
-    const asJson = JSON.stringify(embedding);
-    return sql`vec_f32(${asJson})`;
-  },
   // ============================================================
   // Fulltext Operations
   // ============================================================
   fulltext: fts5Strategy
 };
+var VECTOR_CONTRIBUTION_PREFIX = "vector";
+function buildVectorCapabilities(strategy) {
+  return strategy.capabilities;
+}
+function vectorScoreExpression(distanceExpression, metric) {
+  switch (metric) {
+    case "cosine": {
+      return sql`(1 - (${distanceExpression}))`;
+    }
+    case "l2":
+    case "inner_product": {
+      return distanceExpression;
+    }
+    default: {
+      const _exhaustive = metric;
+      throw new Error(`Unsupported vector metric: ${String(_exhaustive)}`);
+    }
+  }
+}
+function vectorMinScoreCondition(distanceExpression, metric, minScore) {
+  assertVectorMinScore(minScore, metric);
+  switch (metric) {
+    case "cosine": {
+      return sql`${distanceExpression} <= ${1 - minScore}`;
+    }
+    case "l2": {
+      return sql`${distanceExpression} <= ${minScore}`;
+    }
+    case "inner_product": {
+      return sql`${distanceExpression} <= ${-minScore}`;
+    }
+    default: {
+      const _exhaustive = metric;
+      throw new Error(`Unsupported vector metric: ${String(_exhaustive)}`);
+    }
+  }
+}
+function assertFiniteEmbedding(embedding, name) {
+  for (const [index, value] of embedding.entries()) {
+    if (typeof value !== "number" || !Number.isFinite(value)) {
+      throw new TypeError(
+        `${name}[${index}] must be a finite number, got: ${value}`
+      );
+    }
+  }
+}
+function assertVectorSearchLimit(limit) {
+  if (!Number.isInteger(limit) || limit <= 0) {
+    throw new RangeError(
+      `vectorSearch limit must be a positive integer, got: ${limit}`
+    );
+  }
+}
+function assertVectorMinScore(minScore, metric, label = "minScore") {
+  if (!Number.isFinite(minScore)) {
+    throw new RangeError(`${label} must be a finite number, got: ${minScore}`);
+  }
+  if (metric === "cosine" && (minScore < -1 || minScore > 1)) {
+    throw new RangeError(
+      `${label} for the cosine metric must be between -1 and 1, got: ${minScore}`
+    );
+  }
+}
+var MAX_VECTOR_IDENTIFIER_LENGTH = 63;
+function sanitizeIdentifierPart(value) {
+  return value.toLowerCase().replaceAll(/[^a-z0-9_]/g, "_");
+}
+function shortHash(input) {
+  let h1 = 3735928559;
+  let h2 = 1103547991;
+  for (let index = 0; index < input.length; index++) {
+    const ch = input.codePointAt(index);
+    h1 = Math.imul(h1 ^ ch, 2654435761);
+    h2 = Math.imul(h2 ^ ch, 1597334677);
+  }
+  h1 = Math.imul(h1 ^ h1 >>> 16, 2246822507) ^ Math.imul(h2 ^ h2 >>> 13, 3266489909);
+  h2 = Math.imul(h2 ^ h2 >>> 16, 2246822507) ^ Math.imul(h1 ^ h1 >>> 13, 3266489909);
+  const hi = (h2 >>> 0).toString(16).padStart(8, "0");
+  const lo = (h1 >>> 0).toString(16).padStart(8, "0");
+  return `${hi}${lo}`.slice(0, 8);
+}
+function vectorPhysicalName(prefix, graphId, nodeKind, fieldPath) {
+  const readable = `${prefix}_${sanitizeIdentifierPart(graphId)}_${sanitizeIdentifierPart(nodeKind)}_${sanitizeIdentifierPart(fieldPath)}`;
+  const hash = shortHash(JSON.stringify([graphId, nodeKind, fieldPath]));
+  const full = `${readable}_${hash}`;
+  if (full.length <= MAX_VECTOR_IDENTIFIER_LENGTH) return full;
+  return `${readable.slice(0, MAX_VECTOR_IDENTIFIER_LENGTH - 1 - hash.length)}_${hash}`;
+}
+function quoteIdentifier2(name) {
+  return `"${name.replaceAll('"', '""')}"`;
+}
 
 // src/query/dialect/index.ts
 var DIALECT_ADAPTERS = {
@@ -983,7 +1028,7 @@ function generateEdgeIndexDDL(index, dialect, options = {}) {
 function generateTableIndexDDL(index, dialect, tableName, options) {
   const ifNotExists = options.ifNotExists ?? true;
   const propsColumn = sql.raw('"props"');
-  const systemColumn = (column) => sql.raw(quoteIdentifier2(column));
+  const systemColumn = (column) => sql.raw(quoteIdentifier3(column));
   const keys = index.entity === "node" ? compileNodeIndexKeys(index, dialect, propsColumn, systemColumn).keys : compileEdgeIndexKeys(index, dialect, propsColumn, systemColumn).keys;
   const whereSql = index.where ? sqlToInlineString(
     compileIndexWhere(
@@ -1001,9 +1046,9 @@ function generateTableIndexDDL(index, dialect, tableName, options) {
   const concurrent = options.concurrent === true && dialect === "postgres" ? "CONCURRENTLY " : "";
   const ifNotExistsSql = ifNotExists ? "IF NOT EXISTS " : "";
   const whereClause = whereSql ? ` WHERE ${whereSql}` : "";
-  return `CREATE ${unique}INDEX ${concurrent}${ifNotExistsSql}${quoteIdentifier2(index.name)} ON ${quoteIdentifier2(tableName)} (${keySql})${whereClause};`;
+  return `CREATE ${unique}INDEX ${concurrent}${ifNotExistsSql}${quoteIdentifier3(index.name)} ON ${quoteIdentifier3(tableName)} (${keySql})${whereClause};`;
 }
-function quoteIdentifier2(identifier) {
+function quoteIdentifier3(identifier) {
   return `"${identifier.replaceAll('"', '""')}"`;
 }
 function sqlToInlineString(object, dialect) {
@@ -1364,6 +1409,6 @@ function notWhere(predicate) {
   };
 }
 
-export { BASE_CONTRIBUTION_OWNER, andWhere, buildFulltextCapabilities, buildPostgresEdgeIndexBuilders, buildPostgresNodeIndexBuilders, buildSqliteEdgeIndexBuilders, buildSqliteNodeIndexBuilders, edgeColumnList, fts5Strategy, generateEdgeIndexDDL, generateIndexDDL, generateNodeIndexDDL, getDialect, nodeColumnList, notWhere, orWhere, quotedColumn, quotedTableName, sqlNull, toDeclaredIndex, toDeclaredIndexes, tsvectorStrategy };
-//# sourceMappingURL=chunk-ZTREJOGB.js.map
-//# sourceMappingURL=chunk-ZTREJOGB.js.map
+export { BASE_CONTRIBUTION_OWNER, VECTOR_CONTRIBUTION_PREFIX, andWhere, assertFiniteEmbedding, assertVectorMinScore, assertVectorSearchLimit, buildFulltextCapabilities, buildPostgresEdgeIndexBuilders, buildPostgresNodeIndexBuilders, buildSqliteEdgeIndexBuilders, buildSqliteNodeIndexBuilders, buildVectorCapabilities, edgeColumnList, fts5Strategy, generateEdgeIndexDDL, generateIndexDDL, generateNodeIndexDDL, getDialect, nodeColumnList, notWhere, orWhere, quoteIdentifier2 as quoteIdentifier, quotedColumn, quotedTableName, sqlNull, toDeclaredIndex, toDeclaredIndexes, tsvectorStrategy, vectorMinScoreCondition, vectorPhysicalName, vectorScoreExpression };
+//# sourceMappingURL=chunk-7KLVFWXR.js.map
+//# sourceMappingURL=chunk-7KLVFWXR.js.map