codesift-mcp 0.1.0 → 0.2.0

package/dist/retrieval/retrieval-constants.d.ts

@@ -0,0 +1,27 @@
+/** Approximate characters per token for budget estimation.
+ * 3 is conservative (accounts for JSON overhead: keys, quotes, indentation).
+ * Real ratio is ~3.5 for JSON-serialized code, but 3 prevents budget overruns. */
+export declare const CHARS_PER_TOKEN = 3;
+/** Reciprocal Rank Fusion smoothing constant. */
+export declare const RRF_K = 60;
+/** Lines of proximity to consider chunks adjacent (merge threshold). */
+export declare const ADJACENCY_GAP = 5;
+/** Padding width for line numbers in formatted output. */
+export declare const LINE_NUMBER_PAD = 6;
+/** Word count threshold below which queries are not decomposed. */
+export declare const QUERY_DECOMPOSE_THRESHOLD = 8;
+/** Lower bound of the split-window (fraction of word count). */
+export declare const SPLIT_WINDOW_LO = 0.35;
+/** Upper bound of the split-window (fraction of word count). */
+export declare const SPLIT_WINDOW_HI = 0.65;
+/** Maximum sub-queries per codebaseRetrieval call. */
+export declare const MAX_QUERIES = 20;
+/** Minimum remaining token budget to include a truncated result. */
+export declare const MIN_TRUNCATION_TOKENS = 100;
+/** Default top-K results for semantic/hybrid queries. */
+export declare const DEFAULT_TOP_K = 10;
+/** Default character limit for symbol source truncation. */
+export declare const DEFAULT_SOURCE_CHARS = 200;
+/** Timeout in ms for embedding API calls (Voyage, OpenAI, Ollama). */
+export declare const EMBED_TIMEOUT_MS = 30000;
+//# sourceMappingURL=retrieval-constants.d.ts.map

package/dist/retrieval/retrieval-constants.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"retrieval-constants.d.ts","sourceRoot":"","sources":["../../src/retrieval/retrieval-constants.ts"],"names":[],"mappings":"AAAA;;kFAEkF;AAClF,eAAO,MAAM,eAAe,IAAI,CAAC;AAEjC,iDAAiD;AACjD,eAAO,MAAM,KAAK,KAAK,CAAC;AAExB,wEAAwE;AACxE,eAAO,MAAM,aAAa,IAAI,CAAC;AAE/B,0DAA0D;AAC1D,eAAO,MAAM,eAAe,IAAI,CAAC;AAEjC,mEAAmE;AACnE,eAAO,MAAM,yBAAyB,IAAI,CAAC;AAE3C,gEAAgE;AAChE,eAAO,MAAM,eAAe,OAAO,CAAC;AAEpC,gEAAgE;AAChE,eAAO,MAAM,eAAe,OAAO,CAAC;AAEpC,sDAAsD;AACtD,eAAO,MAAM,WAAW,KAAK,CAAC;AAE9B,oEAAoE;AACpE,eAAO,MAAM,qBAAqB,MAAM,CAAC;AAEzC,yDAAyD;AACzD,eAAO,MAAM,aAAa,KAAK,CAAC;AAEhC,4DAA4D;AAC5D,eAAO,MAAM,oBAAoB,MAAM,CAAC;AAExC,sEAAsE;AACtE,eAAO,MAAM,gBAAgB,QAAS,CAAC"}

package/dist/retrieval/retrieval-constants.js

@@ -0,0 +1,27 @@
+/** Approximate characters per token for budget estimation.
+ * 3 is conservative (accounts for JSON overhead: keys, quotes, indentation).
+ * Real ratio is ~3.5 for JSON-serialized code, but 3 prevents budget overruns. */
+export const CHARS_PER_TOKEN = 3;
+/** Reciprocal Rank Fusion smoothing constant. */
+export const RRF_K = 60;
+/** Lines of proximity to consider chunks adjacent (merge threshold). */
+export const ADJACENCY_GAP = 5;
+/** Padding width for line numbers in formatted output. */
+export const LINE_NUMBER_PAD = 6;
+/** Word count threshold below which queries are not decomposed. */
+export const QUERY_DECOMPOSE_THRESHOLD = 8;
+/** Lower bound of the split-window (fraction of word count). */
+export const SPLIT_WINDOW_LO = 0.35;
+/** Upper bound of the split-window (fraction of word count). */
+export const SPLIT_WINDOW_HI = 0.65;
+/** Maximum sub-queries per codebaseRetrieval call. */
+export const MAX_QUERIES = 20;
+/** Minimum remaining token budget to include a truncated result. */
+export const MIN_TRUNCATION_TOKENS = 100;
+/** Default top-K results for semantic/hybrid queries. */
+export const DEFAULT_TOP_K = 10;
+/** Default character limit for symbol source truncation. */
+export const DEFAULT_SOURCE_CHARS = 200;
+/** Timeout in ms for embedding API calls (Voyage, OpenAI, Ollama). */
+export const EMBED_TIMEOUT_MS = 30_000;
+//# sourceMappingURL=retrieval-constants.js.map

package/dist/retrieval/retrieval-constants.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"retrieval-constants.js","sourceRoot":"","sources":["../../src/retrieval/retrieval-constants.ts"],"names":[],"mappings":"AAAA;;kFAEkF;AAClF,MAAM,CAAC,MAAM,eAAe,GAAG,CAAC,CAAC;AAEjC,iDAAiD;AACjD,MAAM,CAAC,MAAM,KAAK,GAAG,EAAE,CAAC;AAExB,wEAAwE;AACxE,MAAM,CAAC,MAAM,aAAa,GAAG,CAAC,CAAC;AAE/B,0DAA0D;AAC1D,MAAM,CAAC,MAAM,eAAe,GAAG,CAAC,CAAC;AAEjC,mEAAmE;AACnE,MAAM,CAAC,MAAM,yBAAyB,GAAG,CAAC,CAAC;AAE3C,gEAAgE;AAChE,MAAM,CAAC,MAAM,eAAe,GAAG,IAAI,CAAC;AAEpC,gEAAgE;AAChE,MAAM,CAAC,MAAM,eAAe,GAAG,IAAI,CAAC;AAEpC,sDAAsD;AACtD,MAAM,CAAC,MAAM,WAAW,GAAG,EAAE,CAAC;AAE9B,oEAAoE;AACpE,MAAM,CAAC,MAAM,qBAAqB,GAAG,GAAG,CAAC;AAEzC,yDAAyD;AACzD,MAAM,CAAC,MAAM,aAAa,GAAG,EAAE,CAAC;AAEhC,4DAA4D;AAC5D,MAAM,CAAC,MAAM,oBAAoB,GAAG,GAAG,CAAC;AAExC,sEAAsE;AACtE,MAAM,CAAC,MAAM,gBAAgB,GAAG,MAAM,CAAC"}

package/dist/retrieval/retrieval-schemas.d.ts

@@ -0,0 +1,107 @@
+import { z } from "zod";
+export declare const SubQuerySchema: z.ZodDiscriminatedUnion<[z.ZodObject<{
+    type: z.ZodLiteral<"symbols">;
+    query: z.ZodString;
+    kind: z.ZodOptional<z.ZodEnum<{
+        function: "function";
+        method: "method";
+        class: "class";
+        interface: "interface";
+        type: "type";
+        variable: "variable";
+        constant: "constant";
+        field: "field";
+        enum: "enum";
+        namespace: "namespace";
+        module: "module";
+        section: "section";
+        metadata: "metadata";
+        test_suite: "test_suite";
+        test_case: "test_case";
+        test_hook: "test_hook";
+        default_export: "default_export";
+        conversation_turn: "conversation_turn";
+        conversation_summary: "conversation_summary";
+        unknown: "unknown";
+    }>>;
+    file_pattern: z.ZodOptional<z.ZodString>;
+    top_k: z.ZodOptional<z.ZodNumber>;
+    source_chars: z.ZodOptional<z.ZodNumber>;
+    rerank: z.ZodOptional<z.ZodBoolean>;
+}, z.core.$strip>, z.ZodObject<{
+    type: z.ZodLiteral<"text">;
+    query: z.ZodString;
+    regex: z.ZodOptional<z.ZodBoolean>;
+    context_lines: z.ZodOptional<z.ZodNumber>;
+    file_pattern: z.ZodOptional<z.ZodString>;
+}, z.core.$strip>, z.ZodObject<{
+    type: z.ZodLiteral<"file_tree">;
+    path: z.ZodOptional<z.ZodString>;
+    path_prefix: z.ZodOptional<z.ZodString>;
+    name_pattern: z.ZodOptional<z.ZodString>;
+    depth: z.ZodOptional<z.ZodNumber>;
+    compact: z.ZodOptional<z.ZodBoolean>;
+    min_symbols: z.ZodOptional<z.ZodNumber>;
+}, z.core.$strip>, z.ZodObject<{
+    type: z.ZodLiteral<"outline">;
+    file_path: z.ZodString;
+}, z.core.$strip>, z.ZodObject<{
+    type: z.ZodLiteral<"references">;
+    symbol_name: z.ZodString;
+}, z.core.$strip>, z.ZodObject<{
+    type: z.ZodLiteral<"call_chain">;
+    symbol_name: z.ZodString;
+    direction: z.ZodOptional<z.ZodEnum<{
+        callers: "callers";
+        callees: "callees";
+    }>>;
+    depth: z.ZodOptional<z.ZodNumber>;
+    include_source: z.ZodOptional<z.ZodBoolean>;
+}, z.core.$strip>, z.ZodObject<{
+    type: z.ZodLiteral<"impact">;
+    since: z.ZodString;
+    depth: z.ZodOptional<z.ZodNumber>;
+    until: z.ZodOptional<z.ZodString>;
+    include_source: z.ZodOptional<z.ZodBoolean>;
+}, z.core.$strip>, z.ZodObject<{
+    type: z.ZodLiteral<"context">;
+    query: z.ZodString;
+    max_tokens: z.ZodOptional<z.ZodNumber>;
+}, z.core.$strip>, z.ZodObject<{
+    type: z.ZodLiteral<"knowledge_map">;
+    focus: z.ZodOptional<z.ZodString>;
+    depth: z.ZodOptional<z.ZodNumber>;
+}, z.core.$strip>, z.ZodObject<{
+    type: z.ZodLiteral<"semantic">;
+    query: z.ZodString;
+    top_k: z.ZodOptional<z.ZodNumber>;
+    file_filter: z.ZodOptional<z.ZodString>;
+    exclude_tests: z.ZodOptional<z.ZodBoolean>;
+    source_chars: z.ZodOptional<z.ZodNumber>;
+    rerank: z.ZodOptional<z.ZodBoolean>;
+}, z.core.$strip>, z.ZodObject<{
+    type: z.ZodLiteral<"hybrid">;
+    query: z.ZodString;
+    top_k: z.ZodOptional<z.ZodNumber>;
+    file_filter: z.ZodOptional<z.ZodString>;
+    exclude_tests: z.ZodOptional<z.ZodBoolean>;
+    rerank: z.ZodOptional<z.ZodBoolean>;
+}, z.core.$strip>, z.ZodObject<{
+    type: z.ZodLiteral<"conversation">;
+    query: z.ZodString;
+    project: z.ZodOptional<z.ZodString>;
+    limit: z.ZodDefault<z.ZodOptional<z.ZodNumber>>;
+}, z.core.$strip>], "type">;
+export type SubQuery = z.infer<typeof SubQuerySchema>;
+export type SubQueryResult = {
+    type: string;
+    data: unknown;
+    tokens: number;
+};
+export interface CodebaseRetrievalResult {
+    results: SubQueryResult[];
+    total_tokens: number;
+    truncated: boolean;
+    query_count: number;
+}
+//# sourceMappingURL=retrieval-schemas.d.ts.map

package/dist/retrieval/retrieval-schemas.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"retrieval-schemas.d.ts","sourceRoot":"","sources":["../../src/retrieval/retrieval-schemas.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,CAAC,EAAE,MAAM,KAAK,CAAC;AAqGxB,eAAO,MAAM,cAAc;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;2BAazB,CAAC;AAEH,MAAM,MAAM,QAAQ,GAAG,CAAC,CAAC,KAAK,CAAC,OAAO,cAAc,CAAC,CAAC;AAEtD,MAAM,MAAM,cAAc,GAAG;IAC3B,IAAI,EAAE,MAAM,CAAC;IACb,IAAI,EAAE,OAAO,CAAC;IACd,MAAM,EAAE,MAAM,CAAC;CAChB,CAAC;AAEF,MAAM,WAAW,uBAAuB;IACtC,OAAO,EAAE,cAAc,EAAE,CAAC;IAC1B,YAAY,EAAE,MAAM,CAAC;IACrB,SAAS,EAAE,OAAO,CAAC;IACnB,WAAW,EAAE,MAAM,CAAC;CACrB"}

package/dist/retrieval/retrieval-schemas.js

@@ -0,0 +1,102 @@
+import { z } from "zod";
+const SymbolKindSchema = z.enum([
+    "function", "method", "class", "interface", "type", "variable",
+    "constant", "field", "enum", "namespace", "module", "section",
+    "metadata", "test_suite", "test_case", "test_hook", "default_export",
+    "conversation_turn", "conversation_summary", "unknown",
+]);
+const SymbolsQuerySchema = z.object({
+    type: z.literal("symbols"),
+    query: z.string(),
+    kind: SymbolKindSchema.optional(),
+    file_pattern: z.string().optional(),
+    top_k: z.number().int().positive().optional(),
+    source_chars: z.number().int().nonnegative().optional(),
+    rerank: z.boolean().optional(),
+});
+const TextQuerySchema = z.object({
+    type: z.literal("text"),
+    query: z.string(),
+    regex: z.boolean().optional(),
+    context_lines: z.number().int().nonnegative().optional(),
+    file_pattern: z.string().optional(),
+});
+const FileTreeQuerySchema = z.object({
+    type: z.literal("file_tree"),
+    path: z.string().optional(),
+    path_prefix: z.string().optional(),
+    name_pattern: z.string().optional(),
+    depth: z.number().int().positive().optional(),
+    compact: z.boolean().optional(),
+    min_symbols: z.number().int().nonnegative().optional(),
+});
+const OutlineQuerySchema = z.object({
+    type: z.literal("outline"),
+    file_path: z.string(),
+});
+const ReferencesQuerySchema = z.object({
+    type: z.literal("references"),
+    symbol_name: z.string(),
+});
+const CallChainQuerySchema = z.object({
+    type: z.literal("call_chain"),
+    symbol_name: z.string(),
+    direction: z.enum(["callers", "callees"]).optional(),
+    depth: z.number().int().positive().optional(),
+    include_source: z.boolean().optional(),
+});
+const ImpactQuerySchema = z.object({
+    type: z.literal("impact"),
+    since: z.string(),
+    depth: z.number().int().positive().optional(),
+    until: z.string().optional(),
+    include_source: z.boolean().optional(),
+});
+const ContextQuerySchema = z.object({
+    type: z.literal("context"),
+    query: z.string(),
+    max_tokens: z.number().int().positive().optional(),
+});
+const KnowledgeMapQuerySchema = z.object({
+    type: z.literal("knowledge_map"),
+    focus: z.string().optional(),
+    depth: z.number().int().positive().optional(),
+});
+const SemanticQuerySchema = z.object({
+    type: z.literal("semantic"),
+    query: z.string(),
+    top_k: z.number().int().positive().optional(),
+    file_filter: z.string().optional(),
+    exclude_tests: z.boolean().optional(),
+    source_chars: z.number().int().nonnegative().optional(),
+    rerank: z.boolean().optional(),
+});
+const HybridQuerySchema = z.object({
+    type: z.literal("hybrid"),
+    query: z.string(),
+    top_k: z.number().int().positive().optional(),
+    file_filter: z.string().optional(),
+    exclude_tests: z.boolean().optional(),
+    rerank: z.boolean().optional(),
+});
+const ConversationQuerySchema = z.object({
+    type: z.literal("conversation"),
+    query: z.string(),
+    project: z.string().optional(),
+    limit: z.number().int().positive().optional().default(5),
+});
+export const SubQuerySchema = z.discriminatedUnion("type", [
+    SymbolsQuerySchema,
+    TextQuerySchema,
+    FileTreeQuerySchema,
+    OutlineQuerySchema,
+    ReferencesQuerySchema,
+    CallChainQuerySchema,
+    ImpactQuerySchema,
+    ContextQuerySchema,
+    KnowledgeMapQuerySchema,
+    SemanticQuerySchema,
+    HybridQuerySchema,
+    ConversationQuerySchema,
+]);
+//# sourceMappingURL=retrieval-schemas.js.map

package/dist/retrieval/retrieval-schemas.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"retrieval-schemas.js","sourceRoot":"","sources":["../../src/retrieval/retrieval-schemas.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,CAAC,EAAE,MAAM,KAAK,CAAC;AAExB,MAAM,gBAAgB,GAAG,CAAC,CAAC,IAAI,CAAC;IAC9B,UAAU,EAAE,QAAQ,EAAE,OAAO,EAAE,WAAW,EAAE,MAAM,EAAE,UAAU;IAC9D,UAAU,EAAE,OAAO,EAAE,MAAM,EAAE,WAAW,EAAE,QAAQ,EAAE,SAAS;IAC7D,UAAU,EAAE,YAAY,EAAE,WAAW,EAAE,WAAW,EAAE,gBAAgB;IACpE,mBAAmB,EAAE,sBAAsB,EAAE,SAAS;CACvD,CAAC,CAAC;AAEH,MAAM,kBAAkB,GAAG,CAAC,CAAC,MAAM,CAAC;IAClC,IAAI,EAAE,CAAC,CAAC,OAAO,CAAC,SAAS,CAAC;IAC1B,KAAK,EAAE,CAAC,CAAC,MAAM,EAAE;IACjB,IAAI,EAAE,gBAAgB,CAAC,QAAQ,EAAE;IACjC,YAAY,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,QAAQ,EAAE;IACnC,KAAK,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,GAAG,EAAE,CAAC,QAAQ,EAAE,CAAC,QAAQ,EAAE;IAC7C,YAAY,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,GAAG,EAAE,CAAC,WAAW,EAAE,CAAC,QAAQ,EAAE;IACvD,MAAM,EAAE,CAAC,CAAC,OAAO,EAAE,CAAC,QAAQ,EAAE;CAC/B,CAAC,CAAC;AAEH,MAAM,eAAe,GAAG,CAAC,CAAC,MAAM,CAAC;IAC/B,IAAI,EAAE,CAAC,CAAC,OAAO,CAAC,MAAM,CAAC;IACvB,KAAK,EAAE,CAAC,CAAC,MAAM,EAAE;IACjB,KAAK,EAAE,CAAC,CAAC,OAAO,EAAE,CAAC,QAAQ,EAAE;IAC7B,aAAa,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,GAAG,EAAE,CAAC,WAAW,EAAE,CAAC,QAAQ,EAAE;IACxD,YAAY,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,QAAQ,EAAE;CACpC,CAAC,CAAC;AAEH,MAAM,mBAAmB,GAAG,CAAC,CAAC,MAAM,CAAC;IACnC,IAAI,EAAE,CAAC,CAAC,OAAO,CAAC,WAAW,CAAC;IAC5B,IAAI,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,QAAQ,EAAE;IAC3B,WAAW,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,QAAQ,EAAE;IAClC,YAAY,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,QAAQ,EAAE;IACnC,KAAK,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,GAAG,EAAE,CAAC,QAAQ,EAAE,CAAC,QAAQ,EAAE;IAC7C,OAAO,EAAE,CAAC,CAAC,OAAO,EAAE,CAAC,QAAQ,EAAE;IAC/B,WAAW,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,GAAG,EAAE,CAAC,WAAW,EAAE,CAAC,QAAQ,EAAE;CACvD,CAAC,CAAC;AAEH,MAAM,kBAAkB,GAAG,CAAC,CAAC,MAAM,CAAC;IAClC,IAAI,EAAE,CAAC,CAAC,OAAO,CAAC,SAAS,CAAC;IAC1B,SAAS,EAAE,CAAC,CAAC,MAAM,EAAE;CACtB,CAAC,CAAC;AAEH,MAAM,qBAAqB,GAAG,CAAC,CAAC,MAAM,CAAC;IACrC,IAAI,EAAE,CAAC,CAAC,OAAO,CAAC,YAAY,CAAC;IAC7B,WAAW,EAAE,CAAC,CAAC,MAAM,EAAE;CACxB,CAAC,CAAC;AAEH,MAAM,oBAAoB,GAAG,CAAC,CAAC,MAAM,CAAC;IACpC,IAAI,EAAE,CAAC,CAAC,OAAO,CAAC,YAAY,CAAC;IAC7B,WAAW,EAAE,CAAC,CAAC,MAAM,EAAE;IACvB,SAAS,EAAE,CAAC,CAAC,IAAI,CAAC,CAAC,SAAS,EAAE,SAAS,CAAC,CAAC,CAAC,QAAQ,EAAE;IACpD,KAAK,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,GAAG,EAAE,CAAC,QAAQ,EAAE,CAAC,QAAQ,EAAE;IAC7C,cAAc,EAAE,CAAC,CAAC,OAAO,EAAE,CAAC,QAAQ,EAAE;CACvC,CAAC,CAAC;AAEH,MAAM,iBAAiB,GAAG,CAAC,CAAC,MAAM,CAAC;IACjC,IAAI,EAAE,CAAC,CAAC,OAAO,CAAC,QAAQ,CAAC;IACzB,KAAK,EAAE,CAAC,CAAC,MAAM,EAAE;IACjB,KAAK,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,GAAG,EAAE,CAAC,QAAQ,EAAE,CAAC,QAAQ,EAAE;IAC7C,KAAK,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,QAAQ,EAAE;IAC5B,cAAc,EAAE,CAAC,CAAC,OAAO,EAAE,CAAC,QAAQ,EAAE;CACvC,CAAC,CAAC;AAEH,MAAM,kBAAkB,GAAG,CAAC,CAAC,MAAM,CAAC;IAClC,IAAI,EAAE,CAAC,CAAC,OAAO,CAAC,SAAS,CAAC;IAC1B,KAAK,EAAE,CAAC,CAAC,MAAM,EAAE;IACjB,UAAU,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,GAAG,EAAE,CAAC,QAAQ,EAAE,CAAC,QAAQ,EAAE;CACnD,CAAC,CAAC;AAEH,MAAM,uBAAuB,GAAG,CAAC,CAAC,MAAM,CAAC;IACvC,IAAI,EAAE,CAAC,CAAC,OAAO,CAAC,eAAe,CAAC;IAChC,KAAK,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,QAAQ,EAAE;IAC5B,KAAK,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,GAAG,EAAE,CAAC,QAAQ,EAAE,CAAC,QAAQ,EAAE;CAC9C,CAAC,CAAC;AAEH,MAAM,mBAAmB,GAAG,CAAC,CAAC,MAAM,CAAC;IACnC,IAAI,EAAE,CAAC,CAAC,OAAO,CAAC,UAAU,CAAC;IAC3B,KAAK,EAAE,CAAC,CAAC,MAAM,EAAE;IACjB,KAAK,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,GAAG,EAAE,CAAC,QAAQ,EAAE,CAAC,QAAQ,EAAE;IAC7C,WAAW,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,QAAQ,EAAE;IAClC,aAAa,EAAE,CAAC,CAAC,OAAO,EAAE,CAAC,QAAQ,EAAE;IACrC,YAAY,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,GAAG,EAAE,CAAC,WAAW,EAAE,CAAC,QAAQ,EAAE;IACvD,MAAM,EAAE,CAAC,CAAC,OAAO,EAAE,CAAC,QAAQ,EAAE;CAC/B,CAAC,CAAC;AAEH,MAAM,iBAAiB,GAAG,CAAC,CAAC,MAAM,CAAC;IACjC,IAAI,EAAE,CAAC,CAAC,OAAO,CAAC,QAAQ,CAAC;IACzB,KAAK,EAAE,CAAC,CAAC,MAAM,EAAE;IACjB,KAAK,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,GAAG,EAAE,CAAC,QAAQ,EAAE,CAAC,QAAQ,EAAE;IAC7C,WAAW,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,QAAQ,EAAE;IAClC,aAAa,EAAE,CAAC,CAAC,OAAO,EAAE,CAAC,QAAQ,EAAE;IACrC,MAAM,EAAE,CAAC,CAAC,OAAO,EAAE,CAAC,QAAQ,EAAE;CAC/B,CAAC,CAAC;AAEH,MAAM,uBAAuB,GAAG,CAAC,CAAC,MAAM,CAAC;IACvC,IAAI,EAAE,CAAC,CAAC,OAAO,CAAC,cAAc,CAAC;IAC/B,KAAK,EAAE,CAAC,CAAC,MAAM,EAAE;IACjB,OAAO,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,QAAQ,EAAE;IAC9B,KAAK,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,GAAG,EAAE,CAAC,QAAQ,EAAE,CAAC,QAAQ,EAAE,CAAC,OAAO,CAAC,CAAC,CAAC;CACzD,CAAC,CAAC;AAEH,MAAM,CAAC,MAAM,cAAc,GAAG,CAAC,CAAC,kBAAkB,CAAC,MAAM,EAAE;IACzD,kBAAkB;IAClB,eAAe;IACf,mBAAmB;IACnB,kBAAkB;IAClB,qBAAqB;IACrB,oBAAoB;IACpB,iBAAiB;IACjB,kBAAkB;IAClB,uBAAuB;IACvB,mBAAmB;IACnB,iBAAiB;IACjB,uBAAuB;CACxB,CAAC,CAAC"}

package/dist/retrieval/retrieval-utils.d.ts

@@ -0,0 +1,40 @@
+import type { CodeChunk } from "../types.js";
+/**
+ * Estimate token count from a string. ~4 chars per token.
+ */
+export declare function estimateTokens(text: string): number;
+/**
+ * Filter embedding entries by file path substring and/or test file exclusion.
+ * Uses a lookup map to resolve the file path for each embedding ID.
+ */
+export declare function filterEmbeddingsByFile(embeddings: Map<string, Float32Array>, fileLookup: Map<string, string | undefined>, fileFilter: string | undefined, excludeTests: boolean): Map<string, Float32Array>;
+/**
+ * Compute RRF scores from multiple embedding query vectors against filtered embeddings.
+ * Each vector produces a ranked list; scores are accumulated via RRF formula.
+ */
+export declare function computeRRFScores(vecs: number[][], filteredEmbeddings: Map<string, Float32Array>, cosSim: (a: Float32Array, b: Float32Array) => number): Map<string, number>;
+export type ChunkEntry = {
+    startLine: number;
+    endLine: number;
+    text: string;
+};
+/**
+ * Group top chunk IDs by file, merge overlapping/adjacent chunks, and format
+ * as numbered plain text sections.
+ */
+export declare function formatChunksAsText(topIds: string[], chunks: Map<string, CodeChunk>, excludeTests: boolean): string;
+/**
+ * Split a long query into sub-queries at natural connectors for RRF merging.
+ */
+export declare function decomposeQuery(query: string): string[];
+/**
+ * Truncate symbol source to a character limit, preserving the rest of the symbol object.
+ */
+export declare function truncateSymbolSource<T extends {
+    source?: string;
+}>(sym: T, limit: number): T;
+/**
+ * Race a promise against a timeout. Rejects with a descriptive error on timeout.
+ */
+export declare function withTimeout<T>(promise: Promise<T>, ms: number, label: string): Promise<T>;
+//# sourceMappingURL=retrieval-utils.d.ts.map

package/dist/retrieval/retrieval-utils.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"retrieval-utils.d.ts","sourceRoot":"","sources":["../../src/retrieval/retrieval-utils.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,SAAS,EAAE,MAAM,aAAa,CAAC;AAY7C;;GAEG;AACH,wBAAgB,cAAc,CAAC,IAAI,EAAE,MAAM,GAAG,MAAM,CAEnD;AAED;;;GAGG;AACH,wBAAgB,sBAAsB,CACpC,UAAU,EAAE,GAAG,CAAC,MAAM,EAAE,YAAY,CAAC,EACrC,UAAU,EAAE,GAAG,CAAC,MAAM,EAAE,MAAM,GAAG,SAAS,CAAC,EAC3C,UAAU,EAAE,MAAM,GAAG,SAAS,EAC9B,YAAY,EAAE,OAAO,GACpB,GAAG,CAAC,MAAM,EAAE,YAAY,CAAC,CAS3B;AAED;;;GAGG;AACH,wBAAgB,gBAAgB,CAC9B,IAAI,EAAE,MAAM,EAAE,EAAE,EAChB,kBAAkB,EAAE,GAAG,CAAC,MAAM,EAAE,YAAY,CAAC,EAC7C,MAAM,EAAE,CAAC,CAAC,EAAE,YAAY,EAAE,CAAC,EAAE,YAAY,KAAK,MAAM,GACnD,GAAG,CAAC,MAAM,EAAE,MAAM,CAAC,CAiBrB;AAED,MAAM,MAAM,UAAU,GAAG;IAAE,SAAS,EAAE,MAAM,CAAC;IAAC,OAAO,EAAE,MAAM,CAAC;IAAC,IAAI,EAAE,MAAM,CAAA;CAAE,CAAC;AAE9E;;;GAGG;AACH,wBAAgB,kBAAkB,CAChC,MAAM,EAAE,MAAM,EAAE,EAChB,MAAM,EAAE,GAAG,CAAC,MAAM,EAAE,SAAS,CAAC,EAC9B,YAAY,EAAE,OAAO,GACpB,MAAM,CAyCR;AAED;;GAEG;AACH,wBAAgB,cAAc,CAAC,KAAK,EAAE,MAAM,GAAG,MAAM,EAAE,CAkBtD;AAED;;GAEG;AACH,wBAAgB,oBAAoB,CAAC,CAAC,SAAS;IAAE,MAAM,CAAC,EAAE,MAAM,CAAA;CAAE,EAChE,GAAG,EAAE,CAAC,EACN,KAAK,EAAE,MAAM,GACZ,CAAC,CAKH;AAED;;GAEG;AACH,wBAAgB,WAAW,CAAC,CAAC,EAC3B,OAAO,EAAE,OAAO,CAAC,CAAC,CAAC,EACnB,EAAE,EAAE,MAAM,EACV,KAAK,EAAE,MAAM,GACZ,OAAO,CAAC,CAAC,CAAC,CAQZ"}

package/dist/retrieval/retrieval-utils.js

@@ -0,0 +1,139 @@
+import { isTestFile } from "../utils/test-file.js";
+import { CHARS_PER_TOKEN, RRF_K, ADJACENCY_GAP, LINE_NUMBER_PAD, QUERY_DECOMPOSE_THRESHOLD, SPLIT_WINDOW_LO, SPLIT_WINDOW_HI, } from "./retrieval-constants.js";
+/**
+ * Estimate token count from a string. ~4 chars per token.
+ */
+export function estimateTokens(text) {
+    return Math.ceil(text.length / CHARS_PER_TOKEN);
+}
+/**
+ * Filter embedding entries by file path substring and/or test file exclusion.
+ * Uses a lookup map to resolve the file path for each embedding ID.
+ */
+export function filterEmbeddingsByFile(embeddings, fileLookup, fileFilter, excludeTests) {
+    if (!fileFilter && !excludeTests)
+        return embeddings;
+    return new Map([...embeddings.entries()].filter(([id]) => {
+        const file = fileLookup.get(id);
+        if (!file)
+            return false;
+        if (fileFilter && !file.includes(fileFilter))
+            return false;
+        if (excludeTests && isTestFile(file))
+            return false;
+        return true;
+    }));
+}
+/**
+ * Compute RRF scores from multiple embedding query vectors against filtered embeddings.
+ * Each vector produces a ranked list; scores are accumulated via RRF formula.
+ */
+export function computeRRFScores(vecs, filteredEmbeddings, cosSim) {
+    const rrfScores = new Map();
+    for (const vec of vecs) {
+        if (!vec)
+            continue;
+        const qEmbed = new Float32Array(vec);
+        const subScores = [];
+        for (const [id, chunkVec] of filteredEmbeddings) {
+            if (chunkVec.length === qEmbed.length) {
+                subScores.push({ id, score: cosSim(qEmbed, chunkVec) });
+            }
+        }
+        subScores.sort((a, b) => b.score - a.score);
+        subScores.forEach((s, rank) => {
+            rrfScores.set(s.id, (rrfScores.get(s.id) ?? 0) + 1 / (RRF_K + rank + 1));
+        });
+    }
+    return rrfScores;
+}
+/**
+ * Group top chunk IDs by file, merge overlapping/adjacent chunks, and format
+ * as numbered plain text sections.
+ */
+export function formatChunksAsText(topIds, chunks, excludeTests) {
+    const byFile = new Map();
+    for (const id of topIds) {
+        const chunk = chunks.get(id);
+        if (!chunk)
+            continue;
+        if (excludeTests && isTestFile(chunk.file))
+            continue;
+        const existing = byFile.get(chunk.file) ?? [];
+        existing.push({ startLine: chunk.startLine, endLine: chunk.endLine, text: chunk.text });
+        byFile.set(chunk.file, existing);
+    }
+    const sections = ["The following code sections were retrieved:"];
+    for (const [file, fileChunks] of byFile) {
+        fileChunks.sort((a, b) => a.startLine - b.startLine);
+        const merged = [];
+        for (const chunk of fileChunks) {
+            const last = merged[merged.length - 1];
+            if (last && chunk.startLine <= last.endLine + ADJACENCY_GAP) {
+                if (chunk.endLine > last.endLine) {
+                    const overlapLines = last.endLine - chunk.startLine + 1;
+                    const newLines = chunk.text.split("\n").slice(overlapLines);
+                    last.text = last.text + "\n" + newLines.join("\n");
+                    last.endLine = chunk.endLine;
+                }
+            }
+            else {
+                merged.push({ startLine: chunk.startLine, endLine: chunk.endLine, text: chunk.text });
+            }
+        }
+        sections.push(`Path: ${file}`);
+        for (const chunk of merged) {
+            const lines = chunk.text.split("\n");
+            const numbered = lines.map((line, i) => {
+                const lineNo = String(chunk.startLine + i).padStart(LINE_NUMBER_PAD, " ");
+                return `${lineNo}\t${line}`;
+            }).join("\n");
+            sections.push(numbered);
+        }
+        sections.push("...");
+    }
+    return sections.join("\n");
+}
+/**
+ * Split a long query into sub-queries at natural connectors for RRF merging.
+ */
+export function decomposeQuery(query) {
+    const words = query.split(/\s+/).filter(Boolean);
+    if (words.length <= QUERY_DECOMPOSE_THRESHOLD)
+        return [query];
+    const splitWords = new Set(["and", "or", "from", "to", "with", "using", "for", "via", "then"]);
+    const lo = Math.floor(words.length * SPLIT_WINDOW_LO);
+    const hi = Math.floor(words.length * SPLIT_WINDOW_HI);
+    for (let i = lo; i <= hi; i++) {
+        if (splitWords.has((words[i] ?? "").toLowerCase())) {
+            const a = words.slice(0, i).join(" ");
+            const b = words.slice(i + 1).join(" ");
+            if (a.trim() && b.trim())
+                return [a, b];
+        }
+    }
+    const mid = Math.floor(words.length / 2);
+    return [words.slice(0, mid).join(" "), words.slice(mid).join(" ")];
+}
+/**
+ * Truncate symbol source to a character limit, preserving the rest of the symbol object.
+ */
+export function truncateSymbolSource(sym, limit) {
+    if (limit > 0 && sym.source && sym.source.length > limit) {
+        return { ...sym, source: sym.source.slice(0, limit) };
+    }
+    return sym;
+}
+/**
+ * Race a promise against a timeout. Rejects with a descriptive error on timeout.
+ */
+export function withTimeout(promise, ms, label) {
+    let timer;
+    return Promise.race([
+        promise.finally(() => clearTimeout(timer)),
+        new Promise((_resolve, reject) => {
+            timer = setTimeout(() => reject(new Error(`${label} timed out after ${ms}ms`)), ms);
+        }),
+    ]);
+}
+//# sourceMappingURL=retrieval-utils.js.map

package/dist/retrieval/retrieval-utils.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"retrieval-utils.js","sourceRoot":"","sources":["../../src/retrieval/retrieval-utils.ts"],"names":[],"mappings":"AACA,OAAO,EAAE,UAAU,EAAE,MAAM,uBAAuB,CAAC;AACnD,OAAO,EACL,eAAe,EACf,KAAK,EACL,aAAa,EACb,eAAe,EACf,yBAAyB,EACzB,eAAe,EACf,eAAe,GAChB,MAAM,0BAA0B,CAAC;AAElC;;GAEG;AACH,MAAM,UAAU,cAAc,CAAC,IAAY;IACzC,OAAO,IAAI,CAAC,IAAI,CAAC,IAAI,CAAC,MAAM,GAAG,eAAe,CAAC,CAAC;AAClD,CAAC;AAED;;;GAGG;AACH,MAAM,UAAU,sBAAsB,CACpC,UAAqC,EACrC,UAA2C,EAC3C,UAA8B,EAC9B,YAAqB;IAErB,IAAI,CAAC,UAAU,IAAI,CAAC,YAAY;QAAE,OAAO,UAAU,CAAC;IACpD,OAAO,IAAI,GAAG,CAAC,CAAC,GAAG,UAAU,CAAC,OAAO,EAAE,CAAC,CAAC,MAAM,CAAC,CAAC,CAAC,EAAE,CAAC,EAAE,EAAE;QACvD,MAAM,IAAI,GAAG,UAAU,CAAC,GAAG,CAAC,EAAE,CAAC,CAAC;QAChC,IAAI,CAAC,IAAI;YAAE,OAAO,KAAK,CAAC;QACxB,IAAI,UAAU,IAAI,CAAC,IAAI,CAAC,QAAQ,CAAC,UAAU,CAAC;YAAE,OAAO,KAAK,CAAC;QAC3D,IAAI,YAAY,IAAI,UAAU,CAAC,IAAI,CAAC;YAAE,OAAO,KAAK,CAAC;QACnD,OAAO,IAAI,CAAC;IACd,CAAC,CAAC,CAAC,CAAC;AACN,CAAC;AAED;;;GAGG;AACH,MAAM,UAAU,gBAAgB,CAC9B,IAAgB,EAChB,kBAA6C,EAC7C,MAAoD;IAEpD,MAAM,SAAS,GAAG,IAAI,GAAG,EAAkB,CAAC;IAC5C,KAAK,MAAM,GAAG,IAAI,IAAI,EAAE,CAAC;QACvB,IAAI,CAAC,GAAG;YAAE,SAAS;QACnB,MAAM,MAAM,GAAG,IAAI,YAAY,CAAC,GAAG,CAAC,CAAC;QACrC,MAAM,SAAS,GAAyC,EAAE,CAAC;QAC3D,KAAK,MAAM,CAAC,EAAE,EAAE,QAAQ,CAAC,IAAI,kBAAkB,EAAE,CAAC;YAChD,IAAI,QAAQ,CAAC,MAAM,KAAK,MAAM,CAAC,MAAM,EAAE,CAAC;gBACtC,SAAS,CAAC,IAAI,CAAC,EAAE,EAAE,EAAE,KAAK,EAAE,MAAM,CAAC,MAAM,EAAE,QAAQ,CAAC,EAAE,CAAC,CAAC;YAC1D,CAAC;QACH,CAAC;QACD,SAAS,CAAC,IAAI,CAAC,CAAC,CAAC,EAAE,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,KAAK,GAAG,CAAC,CAAC,KAAK,CAAC,CAAC;QAC5C,SAAS,CAAC,OAAO,CAAC,CAAC,CAAC,EAAE,IAAI,EAAE,EAAE;YAC5B,SAAS,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,SAAS,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,CAAC,IAAI,CAAC,CAAC,GAAG,CAAC,GAAG,CAAC,KAAK,GAAG,IAAI,GAAG,CAAC,CAAC,CAAC,CAAC;QAC3E,CAAC,CAAC,CAAC;IACL,CAAC;IACD,OAAO,SAAS,CAAC;AACnB,CAAC;AAID;;;GAGG;AACH,MAAM,UAAU,kBAAkB,CAChC,MAAgB,EAChB,MAA8B,EAC9B,YAAqB;IAErB,MAAM,MAAM,GAAG,IAAI,GAAG,EAAwB,CAAC;IAC/C,KAAK,MAAM,EAAE,IAAI,MAAM,EAAE,CAAC;QACxB,MAAM,KAAK,GAAG,MAAM,CAAC,GAAG,CAAC,EAAE,CAAC,CAAC;QAC7B,IAAI,CAAC,KAAK;YAAE,SAAS;QACrB,IAAI,YAAY,IAAI,UAAU,CAAC,KAAK,CAAC,IAAI,CAAC;YAAE,SAAS;QACrD,MAAM,QAAQ,GAAG,MAAM,CAAC,GAAG,CAAC,KAAK,CAAC,IAAI,CAAC,IAAI,EAAE,CAAC;QAC9C,QAAQ,CAAC,IAAI,CAAC,EAAE,SAAS,EAAE,KAAK,CAAC,SAAS,EAAE,OAAO,EAAE,KAAK,CAAC,OAAO,EAAE,IAAI,EAAE,KAAK,CAAC,IAAI,EAAE,CAAC,CAAC;QACxF,MAAM,CAAC,GAAG,CAAC,KAAK,CAAC,IAAI,EAAE,QAAQ,CAAC,CAAC;IACnC,CAAC;IAED,MAAM,QAAQ,GAAa,CAAC,6CAA6C,CAAC,CAAC;IAC3E,KAAK,MAAM,CAAC,IAAI,EAAE,UAAU,CAAC,IAAI,MAAM,EAAE,CAAC;QACxC,UAAU,CAAC,IAAI,CAAC,CAAC,CAAC,EAAE,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,SAAS,GAAG,CAAC,CAAC,SAAS,CAAC,CAAC;QACrD,MAAM,MAAM,GAAiB,EAAE,CAAC;QAChC,KAAK,MAAM,KAAK,IAAI,UAAU,EAAE,CAAC;YAC/B,MAAM,IAAI,GAAG,MAAM,CAAC,MAAM,CAAC,MAAM,GAAG,CAAC,CAAC,CAAC;YACvC,IAAI,IAAI,IAAI,KAAK,CAAC,SAAS,IAAI,IAAI,CAAC,OAAO,GAAG,aAAa,EAAE,CAAC;gBAC5D,IAAI,KAAK,CAAC,OAAO,GAAG,IAAI,CAAC,OAAO,EAAE,CAAC;oBACjC,MAAM,YAAY,GAAG,IAAI,CAAC,OAAO,GAAG,KAAK,CAAC,SAAS,GAAG,CAAC,CAAC;oBACxD,MAAM,QAAQ,GAAG,KAAK,CAAC,IAAI,CAAC,KAAK,CAAC,IAAI,CAAC,CAAC,KAAK,CAAC,YAAY,CAAC,CAAC;oBAC5D,IAAI,CAAC,IAAI,GAAG,IAAI,CAAC,IAAI,GAAG,IAAI,GAAG,QAAQ,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC;oBACnD,IAAI,CAAC,OAAO,GAAG,KAAK,CAAC,OAAO,CAAC;gBAC/B,CAAC;YACH,CAAC;iBAAM,CAAC;gBACN,MAAM,CAAC,IAAI,CAAC,EAAE,SAAS,EAAE,KAAK,CAAC,SAAS,EAAE,OAAO,EAAE,KAAK,CAAC,OAAO,EAAE,IAAI,EAAE,KAAK,CAAC,IAAI,EAAE,CAAC,CAAC;YACxF,CAAC;QACH,CAAC;QACD,QAAQ,CAAC,IAAI,CAAC,SAAS,IAAI,EAAE,CAAC,CAAC;QAC/B,KAAK,MAAM,KAAK,IAAI,MAAM,EAAE,CAAC;YAC3B,MAAM,KAAK,GAAG,KAAK,CAAC,IAAI,CAAC,KAAK,CAAC,IAAI,CAAC,CAAC;YACrC,MAAM,QAAQ,GAAG,KAAK,CAAC,GAAG,CAAC,CAAC,IAAI,EAAE,CAAC,EAAE,EAAE;gBACrC,MAAM,MAAM,GAAG,MAAM,CAAC,KAAK,CAAC,SAAS,GAAG,CAAC,CAAC,CAAC,QAAQ,CAAC,eAAe,EAAE,GAAG,CAAC,CAAC;gBAC1E,OAAO,GAAG,MAAM,KAAK,IAAI,EAAE,CAAC;YAC9B,CAAC,CAAC,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC;YACd,QAAQ,CAAC,IAAI,CAAC,QAAQ,CAAC,CAAC;QAC1B,CAAC;QACD,QAAQ,CAAC,IAAI,CAAC,KAAK,CAAC,CAAC;IACvB,CAAC;IAED,OAAO,QAAQ,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC;AAC7B,CAAC;AAED;;GAEG;AACH,MAAM,UAAU,cAAc,CAAC,KAAa;IAC1C,MAAM,KAAK,GAAG,KAAK,CAAC,KAAK,CAAC,KAAK,CAAC,CAAC,MAAM,CAAC,OAAO,CAAC,CAAC;IACjD,IAAI,KAAK,CAAC,MAAM,IAAI,yBAAyB;QAAE,OAAO,CAAC,KAAK,CAAC,CAAC;IAE9D,MAAM,UAAU,GAAG,IAAI,GAAG,CAAC,CAAC,KAAK,EAAE,IAAI,EAAE,MAAM,EAAE,IAAI,EAAE,MAAM,EAAE,OAAO,EAAE,KAAK,EAAE,KAAK,EAAE,MAAM,CAAC,CAAC,CAAC;IAC/F,MAAM,EAAE,GAAG,IAAI,CAAC,KAAK,CAAC,KAAK,CAAC,MAAM,GAAG,eAAe,CAAC,CAAC;IACtD,MAAM,EAAE,GAAG,IAAI,CAAC,KAAK,CAAC,KAAK,CAAC,MAAM,GAAG,eAAe,CAAC,CAAC;IAEtD,KAAK,IAAI,CAAC,GAAG,EAAE,EAAE,CAAC,IAAI,EAAE,EAAE,CAAC,EAAE,EAAE,CAAC;QAC9B,IAAI,UAAU,CAAC,GAAG,CAAC,CAAC,KAAK,CAAC,CAAC,CAAC,IAAI,EAAE,CAAC,CAAC,WAAW,EAAE,CAAC,EAAE,CAAC;YACnD,MAAM,CAAC,GAAG,KAAK,CAAC,KAAK,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,IAAI,CAAC,GAAG,CAAC,CAAC;YACtC,MAAM,CAAC,GAAG,KAAK,CAAC,KAAK,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,IAAI,CAAC,GAAG,CAAC,CAAC;YACvC,IAAI,CAAC,CAAC,IAAI,EAAE,IAAI,CAAC,CAAC,IAAI,EAAE;gBAAE,OAAO,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC;QAC1C,CAAC;IACH,CAAC;IAED,MAAM,GAAG,GAAG,IAAI,CAAC,KAAK,CAAC,KAAK,CAAC,MAAM,GAAG,CAAC,CAAC,CAAC;IACzC,OAAO,CAAC,KAAK,CAAC,KAAK,CAAC,CAAC,EAAE,GAAG,CAAC,CAAC,IAAI,CAAC,GAAG,CAAC,EAAE,KAAK,CAAC,KAAK,CAAC,GAAG,CAAC,CAAC,IAAI,CAAC,GAAG,CAAC,CAAC,CAAC;AACrE,CAAC;AAED;;GAEG;AACH,MAAM,UAAU,oBAAoB,CAClC,GAAM,EACN,KAAa;IAEb,IAAI,KAAK,GAAG,CAAC,IAAI,GAAG,CAAC,MAAM,IAAI,GAAG,CAAC,MAAM,CAAC,MAAM,GAAG,KAAK,EAAE,CAAC;QACzD,OAAO,EAAE,GAAG,GAAG,EAAE,MAAM,EAAE,GAAG,CAAC,MAAM,CAAC,KAAK,CAAC,CAAC,EAAE,KAAK,CAAC,EAAE,CAAC;IACxD,CAAC;IACD,OAAO,GAAG,CAAC;AACb,CAAC;AAED;;GAEG;AACH,MAAM,UAAU,WAAW,CACzB,OAAmB,EACnB,EAAU,EACV,KAAa;IAEb,IAAI,KAAoC,CAAC;IACzC,OAAO,OAAO,CAAC,IAAI,CAAC;QAClB,OAAO,CAAC,OAAO,CAAC,GAAG,EAAE,CAAC,YAAY,CAAC,KAAK,CAAC,CAAC;QAC1C,IAAI,OAAO,CAAQ,CAAC,QAAQ,EAAE,MAAM,EAAE,EAAE;YACtC,KAAK,GAAG,UAAU,CAAC,GAAG,EAAE,CAAC,MAAM,CAAC,IAAI,KAAK,CAAC,GAAG,KAAK,oBAAoB,EAAE,IAAI,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC;QACtF,CAAC,CAAC;KACH,CAAC,CAAC;AACL,CAAC"}

package/dist/retrieval/semantic-handlers.d.ts

@@ -0,0 +1,8 @@
+import type { SubQuery, SubQueryResult } from "./retrieval-schemas.js";
+export declare function handleSemanticQuery(repo: string, query: Extract<SubQuery, {
+    type: "semantic";
+}>): Promise<SubQueryResult>;
+export declare function handleHybridQuery(repo: string, query: Extract<SubQuery, {
+    type: "hybrid";
+}>): Promise<SubQueryResult>;
+//# sourceMappingURL=semantic-handlers.d.ts.map

package/dist/retrieval/semantic-handlers.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"semantic-handlers.d.ts","sourceRoot":"","sources":["../../src/retrieval/semantic-handlers.ts"],"names":[],"mappings":"AAEA,OAAO,KAAK,EAAE,QAAQ,EAAE,cAAc,EAAE,MAAM,wBAAwB,CAAC;AA8FvE,wBAAsB,mBAAmB,CACvC,IAAI,EAAE,MAAM,EACZ,KAAK,EAAE,OAAO,CAAC,QAAQ,EAAE;IAAE,IAAI,EAAE,UAAU,CAAA;CAAE,CAAC,GAC7C,OAAO,CAAC,cAAc,CAAC,CA8CzB;AAMD,wBAAsB,iBAAiB,CACrC,IAAI,EAAE,MAAM,EACZ,KAAK,EAAE,OAAO,CAAC,QAAQ,EAAE;IAAE,IAAI,EAAE,QAAQ,CAAA;CAAE,CAAC,GAC3C,OAAO,CAAC,cAAc,CAAC,CA2DzB"}

package/dist/retrieval/semantic-handlers.js

@@ -0,0 +1,152 @@
+import { isTestFile } from "../utils/test-file.js";
+import { estimateTokens, filterEmbeddingsByFile, computeRRFScores, formatChunksAsText, decomposeQuery, truncateSymbolSource, withTimeout, } from "./retrieval-utils.js";
+import { DEFAULT_TOP_K, DEFAULT_SOURCE_CHARS, EMBED_TIMEOUT_MS, RRF_K } from "./retrieval-constants.js";
+function formatSemanticResults(results) {
+    if (results.length === 0)
+        return "(no results)";
+    return results.map((r) => {
+        const s = r.symbol;
+        const header = `${s.file}:${s.start_line} ${s.kind} ${s.name}${s.signature ? ` ${s.signature}` : ""}`;
+        return s.source ? `${header}\n${s.source}` : header;
+    }).join("\n\n");
+}
+async function loadSemanticContext(repo, query) {
+    const { createEmbeddingProvider, cosineSimilarity } = await import("../search/semantic.js");
+    const { loadConfig: getConfig } = await import("../config.js");
+    const { getRepo } = await import("../storage/registry.js");
+    const { loadChunks, loadChunkEmbeddings, getChunkPath, getChunkEmbeddingPath } = await import("../storage/chunk-store.js");
+    const config = getConfig();
+    if (!config.embeddingProvider) {
+        throw new Error("No embedding provider configured. Set CODESIFT_VOYAGE_API_KEY, CODESIFT_OPENAI_API_KEY, or CODESIFT_OLLAMA_URL.");
+    }
+    const topK = query.top_k ?? DEFAULT_TOP_K;
+    const fileFilter = query.file_filter;
+    const excludeTests = query.exclude_tests ?? true;
+    const provider = createEmbeddingProvider(config.embeddingProvider, config);
+    const subQueryTexts = decomposeQuery(query.query);
+    const vecs = await withTimeout(provider.embed(subQueryTexts), EMBED_TIMEOUT_MS, "Embedding API");
+    const repoMeta = await getRepo(config.registryPath, repo);
+    let chunks = null;
+    let chunkEmbeddings = null;
+    if (repoMeta) {
+        [chunks, chunkEmbeddings] = await Promise.all([
+            loadChunks(getChunkPath(repoMeta.index_path)),
+            loadChunkEmbeddings(getChunkEmbeddingPath(repoMeta.index_path)),
+        ]);
+    }
+    const chunkFileLookup = chunks
+        ? new Map([...chunks.entries()].map(([id, c]) => [id, c.file]))
+        : new Map();
+    const sourceEmbeddings = chunkEmbeddings ?? new Map();
+    const filteredEmbeddings = filterEmbeddingsByFile(sourceEmbeddings, chunkFileLookup, fileFilter, excludeTests);
+    return {
+        config, provider, vecs, repoMeta,
+        chunks, chunkEmbeddings, chunkFileLookup, filteredEmbeddings,
+        topK, fileFilter, excludeTests, cosineSimilarity,
+    };
+}
+// ---------------------------------------------------------------------------
+// Semantic query handler
+// ---------------------------------------------------------------------------
+export async function handleSemanticQuery(repo, query) {
+    const ctx = await loadSemanticContext(repo, query);
+    // Chunk-level semantic search (preferred path)
+    if (ctx.chunks && ctx.chunkEmbeddings) {
+        const rrfScores = computeRRFScores(ctx.vecs, ctx.filteredEmbeddings, ctx.cosineSimilarity);
+        const topIds = [...rrfScores.entries()]
+            .sort((a, b) => b[1] - a[1])
+            .slice(0, ctx.topK)
+            .map(([id]) => id);
+        let finalIds = topIds;
+        if (query.rerank) {
+            const { rerankChunkIds } = await import("../search/reranker.js");
+            finalIds = await rerankChunkIds(query.query, topIds, ctx.chunks);
+        }
+        const text = formatChunksAsText(finalIds, ctx.chunks, false);
+        return { type: "semantic", data: text, tokens: estimateTokens(text) };
+    }
+    // Fall back to symbol-level semantic search
+    const { getCodeIndex, getEmbeddingCache } = await import("../tools/index-tools.js");
+    const { searchSemantic } = await import("../search/semantic.js");
+    const index = await getCodeIndex(repo);
+    if (!index)
+        throw new Error(`Repository "${repo}" not found`);
+    const embeddings = await getEmbeddingCache(repo);
+    if (!embeddings) {
+        throw new Error(`No embeddings for "${repo}". Run index_folder with an embedding provider configured.`);
+    }
+    const sourceLimit = query.source_chars ?? DEFAULT_SOURCE_CHARS;
+    const symbolMap = new Map(index.symbols.map((s) => [s.id, s]));
+    const symFileLookup = new Map([...symbolMap.entries()].map(([id, s]) => [id, s.file]));
+    const filteredEmbeddings = filterEmbeddingsByFile(embeddings, symFileLookup, ctx.fileFilter, ctx.excludeTests);
+    const primaryVec = ctx.vecs[0];
+    if (!primaryVec)
+        throw new Error("Embedding provider returned no vector");
+    const results = searchSemantic(new Float32Array(primaryVec), filteredEmbeddings, symbolMap, ctx.topK);
+    // Truncate source then format as text (avoid double JSON serialization)
+    const truncated = results.map((r) => ({ ...r, symbol: truncateSymbolSource(r.symbol, sourceLimit) }));
+    const text = formatSemanticResults(truncated);
+    return { type: "semantic", data: text, tokens: estimateTokens(text) };
+}
+// ---------------------------------------------------------------------------
+// Hybrid query handler (semantic + text, RRF-merged)
+// ---------------------------------------------------------------------------
+export async function handleHybridQuery(repo, query) {
+    const { searchText } = await import("../tools/search-tools.js");
+    // Run text search in parallel with semantic context loading (embed + chunk load)
+    const [ctx, textMatches] = await Promise.all([
+        loadSemanticContext(repo, query),
+        searchText(repo, query.query, { file_pattern: query.file_filter }).catch(() => []),
+    ]);
+    if (!ctx.repoMeta)
+        throw new Error(`Repository "${repo}" not found`);
+    if (!ctx.chunks || !ctx.chunkEmbeddings)
+        throw new Error(`No chunk index for "${repo}"`);
+    // 1. Semantic RRF contributions
+    const rrfScores = computeRRFScores(ctx.vecs, ctx.filteredEmbeddings, ctx.cosineSimilarity);
+    // 2. Text match RRF contributions — map match line -> covering chunk
+    const fileToChunks = new Map();
+    for (const [id, chunk] of ctx.chunks) {
+        const list = fileToChunks.get(chunk.file) ?? [];
+        list.push({ id, startLine: chunk.startLine, endLine: chunk.endLine });
+        fileToChunks.set(chunk.file, list);
+    }
+    for (const list of fileToChunks.values()) {
+        list.sort((a, b) => a.startLine - b.startLine);
+    }
+    for (let rank = 0; rank < textMatches.length; rank++) {
+        const match = textMatches[rank];
+        if (!match)
+            continue;
+        if (ctx.excludeTests && isTestFile(match.file))
+            continue;
+        const list = fileToChunks.get(match.file) ?? [];
+        // Binary search for covering chunk (list is sorted by startLine)
+        let lo = 0, hi = list.length - 1;
+        while (lo <= hi) {
+            const mid = (lo + hi) >> 1;
+            const chunk = list[mid];
+            if (chunk.endLine < match.line) {
+                lo = mid + 1;
+            }
+            else if (chunk.startLine > match.line) {
+                hi = mid - 1;
+            }
+            else {
+                rrfScores.set(chunk.id, (rrfScores.get(chunk.id) ?? 0) + 1 / (RRF_K + rank + 1));
+                break;
+            }
+        }
+    }
+    let topIds = [...rrfScores.entries()]
+        .sort((a, b) => b[1] - a[1])
+        .slice(0, ctx.topK)
+        .map(([id]) => id);
+    if (query.rerank) {
+        const { rerankChunkIds } = await import("../search/reranker.js");
+        topIds = await rerankChunkIds(query.query, topIds, ctx.chunks);
+    }
+    const hybridText = formatChunksAsText(topIds, ctx.chunks, ctx.excludeTests);
+    return { type: "hybrid", data: hybridText, tokens: estimateTokens(hybridText) };
+}
+//# sourceMappingURL=semantic-handlers.js.map