raggrep 0.15.0 → 0.17.0

package/dist/domain/ports/embedding.d.ts

@@ -10,12 +10,22 @@ import type { Logger } from "./logger";
  * Available embedding model names
  */
 export type EmbeddingModelName = "all-MiniLM-L6-v2" | "all-MiniLM-L12-v2" | "bge-small-en-v1.5" | "paraphrase-MiniLM-L3-v2" | "nomic-embed-text-v1.5";
+/**
+ * Which Transformers.js distribution backs local embedding inference.
+ * Used by the composition root / factory to pick an adapter implementation.
+ */
+export type EmbeddingRuntime = "xenova" | "huggingface";
 /**
  * Configuration for embedding provider
  */
 export interface EmbeddingConfig {
     /** Model name to use */
     model: EmbeddingModelName;
+    /**
+     * Embedding runtime (`@xenova/transformers` vs `@huggingface/transformers`).
+     * Defaults to `huggingface` when omitted.
+     */
+    runtime?: EmbeddingRuntime;
     /** Whether to show progress during model loading (deprecated, use logger instead) */
     showProgress?: boolean;
     /** Logger for reporting download progress */

package/dist/domain/ports/index.d.ts

@@ -5,7 +5,7 @@
  * These are implemented by infrastructure adapters.
  */
 export type { FileSystem, FileStats } from "./filesystem";
-export type { EmbeddingProvider, EmbeddingConfig, EmbeddingModelName, } from "./embedding";
+export type { EmbeddingProvider, EmbeddingConfig, EmbeddingModelName, EmbeddingRuntime, } from "./embedding";
 export type { IndexStorage } from "./storage";
 export type { Logger, ProgressInfo, LoggerFactory } from "./logger";
 export type { IParser, IGrammarManager, ParsedChunk, ParseResult, ParserConfig, ParserLanguage, GrammarStatus, } from "./parser";

package/dist/domain/services/chunkContext.d.ts

@@ -0,0 +1,76 @@
+/**
+ * Chunk Context Preparation Service
+ *
+ * Provides a unified utility for preparing chunk content with path context
+ * for embedding. This ensures consistent behavior across all indexing modules.
+ *
+ * The path context helps embeddings understand the structural location of code,
+ * improving search relevance for queries that reference file paths or domains.
+ */
+/**
+ * Options for preparing a chunk for embedding.
+ */
+export interface ChunkContextOptions {
+    /** Relative file path (from project root) */
+    filepath: string;
+    /** The chunk content to embed */
+    content: string;
+    /** Optional name for the chunk (e.g., function name, heading) */
+    name?: string;
+    /** Optional documentation comment (e.g., JSDoc, docstring) */
+    docComment?: string;
+}
+/**
+ * Prepare chunk content for embedding by injecting path context.
+ *
+ * This function should be used by ALL indexing modules to ensure consistent
+ * path context injection. The path context is prepended to the content to
+ * give the embedding model awareness of the file's location in the project.
+ *
+ * Format: `[domain layer segment1 segment2] [name:] [docComment] content`
+ *
+ * @param options - Chunk context options
+ * @returns Content string ready for embedding
+ *
+ * @example
+ * ```typescript
+ * const embeddingContent = prepareChunkForEmbedding({
+ *   filepath: 'services/dynamodb/streams/handler.ts',
+ *   content: 'export function processStream() { ... }',
+ *   name: 'processStream',
+ * });
+ * // Returns: "[dynamodb service streams] processStream: export function processStream() { ... }"
+ * ```
+ */
+export declare function prepareChunkForEmbedding(options: ChunkContextOptions): string;
+/**
+ * Extract path keywords for inclusion in FileSummary.
+ *
+ * This provides a deduplicated list of keywords extracted from the file path,
+ * suitable for BM25 and other keyword-based search mechanisms.
+ *
+ * @param filepath - Relative file path
+ * @returns Array of deduplicated lowercase keywords
+ *
+ * @example
+ * ```typescript
+ * const keywords = extractPathKeywordsForFileSummary('services/dynamodb/streams/handler.ts');
+ * // Returns: ['services', 'dynamodb', 'streams', 'handler']
+ * ```
+ */
+export declare function extractPathKeywordsForFileSummary(filepath: string): string[];
+/**
+ * Get path context data for inclusion in FileSummary.
+ *
+ * This returns the parsed path context that can be stored in FileSummary
+ * for later use during search scoring.
+ *
+ * @param filepath - Relative file path
+ * @returns Path context object
+ */
+export declare function getPathContextForFileSummary(filepath: string): {
+    segments: string[];
+    layer?: string;
+    domain?: string;
+    depth: number;
+};

package/dist/domain/services/index.d.ts

@@ -18,3 +18,4 @@ export { introspectFile, findNearestReadme, introspectionToKeywords, detectScope
 export { validateConfig, formatValidationIssues, type ValidationIssue, type ValidationResult, } from "./configValidator";
 export { calculatePhraseMatch, hasExactPhrase, calculateTokenCoverage, tokenizeForMatching, PHRASE_MATCH_CONSTANTS, type PhraseMatchResult, } from "./phraseMatch";
 export { isIdentifierQuery, extractSearchLiteral, findOccurrences, searchFiles, extractIdentifiersFromContent, isSearchableContent, } from "./simpleSearch";
+export { prepareChunkForEmbedding, extractPathKeywordsForFileSummary, getPathContextForFileSummary, type ChunkContextOptions, } from "./chunkContext";