@grepr/cli 1.1.4 → 1.3.0-94fe60c

package/build/dist/lib/time-utils.js

@@ -73,7 +73,7 @@ export function parseSinceOption(since) {
             const sinceTime = new Date(now.getTime() - milliseconds);
             return sinceTime.toISOString();
         }
-        catch (error) {
+        catch {
             throw new Error(`Invalid duration format: ${since}. Use ISO 8601 duration format like PT5H, P1D, PT30M, P1Y6M, P2W`);
         }
     }

package/build/dist/lib/time-utils.js.map

@@ -1 +1 @@
-{"version":3,"file":"time-utils.js","sourceRoot":"","sources":["../../../src/main/typescript/lib/time-utils.ts"],"names":[],"mappings":"AAAA;;GAEG;AAEH,OAAO,KAAK,MAAM,OAAO,CAAC;AAC1B,OAAO,QAAQ,MAAM,0BAA0B,CAAC;AAEhD,uDAAuD;AACvD,KAAK,CAAC,MAAM,CAAC,QAAQ,CAAC,CAAC;AAEvB;;;;;;;;;;;;;GAaG;AACH,MAAM,UAAU,2BAA2B,CAAC,QAAgB;IAC1D,0CAA0C;IAC1C,IAAI,CAAC,QAAQ,CAAC,KAAK,CAAC,IAAI,CAAC,EAAE,CAAC;QAC1B,MAAM,IAAI,KAAK,CAAC,4BAA4B,QAAQ,2EAA2E,CAAC,CAAC;IACnI,CAAC;IAED,IAAI,CAAC;QACH,2CAA2C;QAC3C,MAAM,MAAM,GAAG,KAAK,CAAC,QAAQ,CAAC,QAAQ,CAAC,CAAC;QAExC,0BAA0B;QAC1B,MAAM,iBAAiB,GAAG,MAAM,CAAC,cAAc,EAAE,CAAC;QAElD,4CAA4C;QAC5C,IAAI,iBAAiB,KAAK,CAAC,EAAE,CAAC;YAC5B,kEAAkE;YAClE,IAAI,QAAQ,CAAC,KAAK,CAAC,2BAA2B,CAAC,EAAE,CAAC;gBAChD,MAAM,IAAI,KAAK,CAAC,uCAAuC,QAAQ,EAAE,CAAC,CAAC;YACrE,CAAC;YACD,0EAA0E;YAC1E,MAAM,IAAI,KAAK,CAAC,4BAA4B,QAAQ,2EAA2E,CAAC,CAAC;QACnI,CAAC;QAED,IAAI,CAAC,QAAQ,CAAC,iBAAiB,CAAC,IAAI,iBAAiB,GAAG,CAAC,EAAE,CAAC;YAC1D,MAAM,IAAI,KAAK,CAAC,qBAAqB,QAAQ,EAAE,CAAC,CAAC;QACnD,CAAC;QAED,OAAO,iBAAiB,CAAC;IAC3B,CAAC;IAAC,OAAO,KAAK,EAAE,CAAC;QACf,gDAAgD;QAChD,IAAI,KAAK,YAAY,KAAK,IAAI,KAAK,CAAC,OAAO,CAAC,QAAQ,CAAC,oCAAoC,CAAC,EAAE,CAAC;YAC3F,MAAM,KAAK,CAAC;QACd,CAAC;QACD,oCAAoC;QACpC,MAAM,IAAI,KAAK,CAAC,4BAA4B,QAAQ,2EAA2E,CAAC,CAAC;IACnI,CAAC;AACH,CAAC;AAED;;;;;;;GAOG;AACH,MAAM,UAAU,gBAAgB,CAAC,KAAa;IAC5C,8CAA8C;IAC9C,IAAI,KAAK,CAAC,KAAK,CAAC,sCAAsC,CAAC,EAAE,CAAC;QACxD,OAAO,KAAK,CAAC;IACf,CAAC;IAED,qDAAqD;IACrD,IAAI,KAAK,CAAC,KAAK,CAAC,IAAI,CAAC,EAAE,CAAC;QACtB,IAAI,CAAC;YACH,MAAM,GAAG,GAAG,IAAI,IAAI,EAAE,CAAC;YACvB,MAAM,YAAY,GAAG,2BAA2B,CAAC,KAAK,CAAC,CAAC;YACxD,MAAM,SAAS,GAAG,IAAI,IAAI,CAAC,GAAG,CAAC,OAAO,EAAE,GAAG,YAAY,CAAC,CAAC;YACzD,OAAO,SAAS,CAAC,WAAW,EAAE,CAAC;QACjC,CAAC;QAAC,OAAO,KAAK,EAAE,CAAC;YACf,MAAM,IAAI,KAAK,CAAC,4BAA4B,KAAK,kEAAkE,CAAC,CAAC;QACvH,CAAC;IACH,CAAC;IAED,iEAAiE;IACjE,MAAM,MAAM,GAAG,IAAI,CAAC,KAAK,CAAC,KAAK,CAAC,CAAC;IACjC,IAAI,KAAK,CAAC,MAAM,CAAC,EAAE,CAAC;QAClB,MAAM,IAAI,KAAK,CAAC,wBAAwB,KAAK,iEAAiE,CAAC,CAAC;IAClH,CAAC;IAED,OAAO,IAAI,IAAI,CAAC,MAAM,CAAC,CAAC,WAAW,EAAE,CAAC;AACxC,CAAC"}
+{"version":3,"file":"time-utils.js","sourceRoot":"","sources":["../../../src/main/typescript/lib/time-utils.ts"],"names":[],"mappings":"AAAA;;GAEG;AAEH,OAAO,KAAK,MAAM,OAAO,CAAC;AAC1B,OAAO,QAAQ,MAAM,0BAA0B,CAAC;AAEhD,uDAAuD;AACvD,KAAK,CAAC,MAAM,CAAC,QAAQ,CAAC,CAAC;AAEvB;;;;;;;;;;;;;GAaG;AACH,MAAM,UAAU,2BAA2B,CAAC,QAAgB;IAC1D,0CAA0C;IAC1C,IAAI,CAAC,QAAQ,CAAC,KAAK,CAAC,IAAI,CAAC,EAAE,CAAC;QAC1B,MAAM,IAAI,KAAK,CAAC,4BAA4B,QAAQ,2EAA2E,CAAC,CAAC;IACnI,CAAC;IAED,IAAI,CAAC;QACH,2CAA2C;QAC3C,MAAM,MAAM,GAAG,KAAK,CAAC,QAAQ,CAAC,QAAQ,CAAC,CAAC;QAExC,0BAA0B;QAC1B,MAAM,iBAAiB,GAAG,MAAM,CAAC,cAAc,EAAE,CAAC;QAElD,4CAA4C;QAC5C,IAAI,iBAAiB,KAAK,CAAC,EAAE,CAAC;YAC5B,kEAAkE;YAClE,IAAI,QAAQ,CAAC,KAAK,CAAC,2BAA2B,CAAC,EAAE,CAAC;gBAChD,MAAM,IAAI,KAAK,CAAC,uCAAuC,QAAQ,EAAE,CAAC,CAAC;YACrE,CAAC;YACD,0EAA0E;YAC1E,MAAM,IAAI,KAAK,CAAC,4BAA4B,QAAQ,2EAA2E,CAAC,CAAC;QACnI,CAAC;QAED,IAAI,CAAC,QAAQ,CAAC,iBAAiB,CAAC,IAAI,iBAAiB,GAAG,CAAC,EAAE,CAAC;YAC1D,MAAM,IAAI,KAAK,CAAC,qBAAqB,QAAQ,EAAE,CAAC,CAAC;QACnD,CAAC;QAED,OAAO,iBAAiB,CAAC;IAC3B,CAAC;IAAC,OAAO,KAAK,EAAE,CAAC;QACf,gDAAgD;QAChD,IAAI,KAAK,YAAY,KAAK,IAAI,KAAK,CAAC,OAAO,CAAC,QAAQ,CAAC,oCAAoC,CAAC,EAAE,CAAC;YAC3F,MAAM,KAAK,CAAC;QACd,CAAC;QACD,oCAAoC;QACpC,MAAM,IAAI,KAAK,CAAC,4BAA4B,QAAQ,2EAA2E,CAAC,CAAC;IACnI,CAAC;AACH,CAAC;AAED;;;;;;;GAOG;AACH,MAAM,UAAU,gBAAgB,CAAC,KAAa;IAC5C,8CAA8C;IAC9C,IAAI,KAAK,CAAC,KAAK,CAAC,sCAAsC,CAAC,EAAE,CAAC;QACxD,OAAO,KAAK,CAAC;IACf,CAAC;IAED,qDAAqD;IACrD,IAAI,KAAK,CAAC,KAAK,CAAC,IAAI,CAAC,EAAE,CAAC;QACtB,IAAI,CAAC;YACH,MAAM,GAAG,GAAG,IAAI,IAAI,EAAE,CAAC;YACvB,MAAM,YAAY,GAAG,2BAA2B,CAAC,KAAK,CAAC,CAAC;YACxD,MAAM,SAAS,GAAG,IAAI,IAAI,CAAC,GAAG,CAAC,OAAO,EAAE,GAAG,YAAY,CAAC,CAAC;YACzD,OAAO,SAAS,CAAC,WAAW,EAAE,CAAC;QACjC,CAAC;QAAC,MAAM,CAAC;YACP,MAAM,IAAI,KAAK,CAAC,4BAA4B,KAAK,kEAAkE,CAAC,CAAC;QACvH,CAAC;IACH,CAAC;IAED,iEAAiE;IACjE,MAAM,MAAM,GAAG,IAAI,CAAC,KAAK,CAAC,KAAK,CAAC,CAAC;IACjC,IAAI,KAAK,CAAC,MAAM,CAAC,EAAE,CAAC;QAClB,MAAM,IAAI,KAAK,CAAC,wBAAwB,KAAK,iEAAiE,CAAC,CAAC;IAClH,CAAC;IAED,OAAO,IAAI,IAAI,CAAC,MAAM,CAAC,CAAC,WAAW,EAAE,CAAC;AACxC,CAAC"}

package/build/dist/lib/transformers-embeddings.d.ts

@@ -0,0 +1,76 @@
+import { EmbeddingsModel, EmbeddingsResponse } from 'vectra';
+/**
+ * Embeddings provider for Vectra using Transformers.js.
+ *
+ * This class implements Vectra's EmbeddingsModel interface using Transformers.js,
+ * a pure JavaScript implementation of transformer models. This approach provides:
+ * - Zero external dependencies (no Docker, no server startup)
+ * - Automatic model caching (~90MB downloaded once, reused thereafter)
+ * - Cross-platform compatibility (works on any Node.js environment)
+ *
+ * The model generates dense vector embeddings for text, enabling semantic similarity
+ * search. The embeddings are generated using mean pooling and normalization, which
+ * is standard for sentence similarity tasks.
+ *
+ * Design decision: We chose Transformers.js over alternatives like Ollama because:
+ * 1. No runtime server dependency (Ollama requires Docker container)
+ * 2. Faster startup time (no container initialization)
+ * 3. Smaller distribution size (model cached by npm, not bundled)
+ * 4. Better integration with Node.js ecosystem
+ */
+export declare class TransformersEmbeddings implements EmbeddingsModel {
+    private pipeline;
+    /**
+     * Maximum number of tokens the model can process per input.
+     * This is exposed publicly so Vectra can use it for chunking decisions.
+     */
+    readonly maxTokens: number;
+    private readonly modelName;
+    /**
+     * Creates a new TransformersEmbeddings instance.
+     *
+     * @param modelName - HuggingFace model identifier (e.g., 'Xenova/all-MiniLM-L6-v2')
+     *                    Default model produces 384-dimensional embeddings and is ~90MB.
+     *                    See https://huggingface.co/models?library=transformers.js&pipeline_tag=feature-extraction
+     *                    for other compatible models.
+     * @param maxTokens - Maximum tokens per input. Should match model's configuration.
+     *                    Used by Vectra for automatic document chunking.
+     */
+    constructor(modelName?: string, maxTokens?: number);
+    /**
+     * Initializes the embedding model pipeline.
+     *
+     * This method is lazy - it only loads the model when first called. The model
+     * is downloaded from HuggingFace on first run and cached locally by Transformers.js
+     * in ~/.cache/huggingface/ (or platform equivalent).
+     *
+     * Subsequent calls to this method are no-ops if the pipeline is already initialized.
+     *
+     * @throws Error if model download or initialization fails (e.g., network issues,
+     *         unsupported model architecture)
+     */
+    initialize(): Promise<void>;
+    /**
+     * Generates vector embeddings for one or more text inputs.
+     *
+     * This is the main method called by Vectra to generate embeddings for documents
+     * and queries. The implementation:
+     * 1. Initializes the pipeline if not already initialized (lazy loading)
+     * 2. Normalizes input to always be an array
+     * 3. Generates embeddings for each input sequentially
+     * 4. Returns embeddings as a 2D array of numbers
+     *
+     * Important: We use mean pooling and normalization for the embeddings. This is
+     * standard practice for sentence similarity tasks:
+     * - Mean pooling: Averages token embeddings to get a single vector per sentence
+     * - Normalization: Scales vectors to unit length, enabling cosine similarity
+     *                   via simple dot product
+     *
+     * @param inputs - Single text string or array of strings to embed
+     * @returns Promise resolving to EmbeddingsResponse with either:
+     *          - success status and array of embedding vectors
+     *          - error status and error message
+     */
+    createEmbeddings(inputs: string | string[]): Promise<EmbeddingsResponse>;
+}
+//# sourceMappingURL=transformers-embeddings.d.ts.map

package/build/dist/lib/transformers-embeddings.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"transformers-embeddings.d.ts","sourceRoot":"","sources":["../../../src/main/typescript/lib/transformers-embeddings.ts"],"names":[],"mappings":"AACA,OAAO,EAAE,eAAe,EAAE,kBAAkB,EAAE,MAAM,QAAQ,CAAC;AAE7D;;;;;;;;;;;;;;;;;;GAkBG;AACH,qBAAa,sBAAuB,YAAW,eAAe;IAC5D,OAAO,CAAC,QAAQ,CAA0C;IAE1D;;;OAGG;IACH,SAAgB,SAAS,EAAE,MAAM,CAAC;IAElC,OAAO,CAAC,QAAQ,CAAC,SAAS,CAAS;IAEnC;;;;;;;;;OASG;gBACS,SAAS,SAA4B,EAAE,SAAS,SAAM;IAKlE;;;;;;;;;;;OAWG;IACG,UAAU,IAAI,OAAO,CAAC,IAAI,CAAC;IAMjC;;;;;;;;;;;;;;;;;;;;OAoBG;IACG,gBAAgB,CAAC,MAAM,EAAE,MAAM,GAAG,MAAM,EAAE,GAAG,OAAO,CAAC,kBAAkB,CAAC;CA8B/E"}

package/build/dist/lib/transformers-embeddings.js

@@ -0,0 +1,109 @@
+import { pipeline } from '@xenova/transformers';
+/**
+ * Embeddings provider for Vectra using Transformers.js.
+ *
+ * This class implements Vectra's EmbeddingsModel interface using Transformers.js,
+ * a pure JavaScript implementation of transformer models. This approach provides:
+ * - Zero external dependencies (no Docker, no server startup)
+ * - Automatic model caching (~90MB downloaded once, reused thereafter)
+ * - Cross-platform compatibility (works on any Node.js environment)
+ *
+ * The model generates dense vector embeddings for text, enabling semantic similarity
+ * search. The embeddings are generated using mean pooling and normalization, which
+ * is standard for sentence similarity tasks.
+ *
+ * Design decision: We chose Transformers.js over alternatives like Ollama because:
+ * 1. No runtime server dependency (Ollama requires Docker container)
+ * 2. Faster startup time (no container initialization)
+ * 3. Smaller distribution size (model cached by npm, not bundled)
+ * 4. Better integration with Node.js ecosystem
+ */
+export class TransformersEmbeddings {
+    pipeline = null;
+    /**
+     * Maximum number of tokens the model can process per input.
+     * This is exposed publicly so Vectra can use it for chunking decisions.
+     */
+    maxTokens;
+    modelName;
+    /**
+     * Creates a new TransformersEmbeddings instance.
+     *
+     * @param modelName - HuggingFace model identifier (e.g., 'Xenova/all-MiniLM-L6-v2')
+     *                    Default model produces 384-dimensional embeddings and is ~90MB.
+     *                    See https://huggingface.co/models?library=transformers.js&pipeline_tag=feature-extraction
+     *                    for other compatible models.
+     * @param maxTokens - Maximum tokens per input. Should match model's configuration.
+     *                    Used by Vectra for automatic document chunking.
+     */
+    constructor(modelName = 'Xenova/all-MiniLM-L6-v2', maxTokens = 512) {
+        this.modelName = modelName;
+        this.maxTokens = maxTokens;
+    }
+    /**
+     * Initializes the embedding model pipeline.
+     *
+     * This method is lazy - it only loads the model when first called. The model
+     * is downloaded from HuggingFace on first run and cached locally by Transformers.js
+     * in ~/.cache/huggingface/ (or platform equivalent).
+     *
+     * Subsequent calls to this method are no-ops if the pipeline is already initialized.
+     *
+     * @throws Error if model download or initialization fails (e.g., network issues,
+     *         unsupported model architecture)
+     */
+    async initialize() {
+        if (!this.pipeline) {
+            this.pipeline = await pipeline('feature-extraction', this.modelName);
+        }
+    }
+    /**
+     * Generates vector embeddings for one or more text inputs.
+     *
+     * This is the main method called by Vectra to generate embeddings for documents
+     * and queries. The implementation:
+     * 1. Initializes the pipeline if not already initialized (lazy loading)
+     * 2. Normalizes input to always be an array
+     * 3. Generates embeddings for each input sequentially
+     * 4. Returns embeddings as a 2D array of numbers
+     *
+     * Important: We use mean pooling and normalization for the embeddings. This is
+     * standard practice for sentence similarity tasks:
+     * - Mean pooling: Averages token embeddings to get a single vector per sentence
+     * - Normalization: Scales vectors to unit length, enabling cosine similarity
+     *                   via simple dot product
+     *
+     * @param inputs - Single text string or array of strings to embed
+     * @returns Promise resolving to EmbeddingsResponse with either:
+     *          - success status and array of embedding vectors
+     *          - error status and error message
+     */
+    async createEmbeddings(inputs) {
+        try {
+            if (!this.pipeline) {
+                await this.initialize();
+            }
+            if (!this.pipeline) {
+                throw new Error('Pipeline not initialized');
+            }
+            const inputArray = Array.isArray(inputs) ? inputs : [inputs];
+            const embeddings = [];
+            for (const text of inputArray) {
+                const output = await this.pipeline(text, { pooling: 'mean', normalize: true });
+                const embedding = Array.from(output.data);
+                embeddings.push(embedding);
+            }
+            return {
+                status: 'success',
+                output: embeddings
+            };
+        }
+        catch (error) {
+            return {
+                status: 'error',
+                message: error instanceof Error ? error.message : 'Unknown error generating embeddings'
+            };
+        }
+    }
+}
+//# sourceMappingURL=transformers-embeddings.js.map

package/build/dist/lib/transformers-embeddings.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"transformers-embeddings.js","sourceRoot":"","sources":["../../../src/main/typescript/lib/transformers-embeddings.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,QAAQ,EAA6B,MAAM,sBAAsB,CAAC;AAG3E;;;;;;;;;;;;;;;;;;GAkBG;AACH,MAAM,OAAO,sBAAsB;IACzB,QAAQ,GAAqC,IAAI,CAAC;IAE1D;;;OAGG;IACa,SAAS,CAAS;IAEjB,SAAS,CAAS;IAEnC;;;;;;;;;OASG;IACH,YAAY,SAAS,GAAG,yBAAyB,EAAE,SAAS,GAAG,GAAG;QAChE,IAAI,CAAC,SAAS,GAAG,SAAS,CAAC;QAC3B,IAAI,CAAC,SAAS,GAAG,SAAS,CAAC;IAC7B,CAAC;IAED;;;;;;;;;;;OAWG;IACH,KAAK,CAAC,UAAU;QACd,IAAI,CAAC,IAAI,CAAC,QAAQ,EAAE,CAAC;YACnB,IAAI,CAAC,QAAQ,GAAG,MAAM,QAAQ,CAAC,oBAAoB,EAAE,IAAI,CAAC,SAAS,CAAC,CAAC;QACvE,CAAC;IACH,CAAC;IAED;;;;;;;;;;;;;;;;;;;;OAoBG;IACH,KAAK,CAAC,gBAAgB,CAAC,MAAyB;QAC9C,IAAI,CAAC;YACH,IAAI,CAAC,IAAI,CAAC,QAAQ,EAAE,CAAC;gBACnB,MAAM,IAAI,CAAC,UAAU,EAAE,CAAC;YAC1B,CAAC;YAED,IAAI,CAAC,IAAI,CAAC,QAAQ,EAAE,CAAC;gBACnB,MAAM,IAAI,KAAK,CAAC,0BAA0B,CAAC,CAAC;YAC9C,CAAC;YAED,MAAM,UAAU,GAAG,KAAK,CAAC,OAAO,CAAC,MAAM,CAAC,CAAC,CAAC,CAAC,MAAM,CAAC,CAAC,CAAC,CAAC,MAAM,CAAC,CAAC;YAC7D,MAAM,UAAU,GAAe,EAAE,CAAC;YAElC,KAAK,MAAM,IAAI,IAAI,UAAU,EAAE,CAAC;gBAC9B,MAAM,MAAM,GAAG,MAAM,IAAI,CAAC,QAAQ,CAAC,IAAI,EAAE,EAAE,OAAO,EAAE,MAAM,EAAE,SAAS,EAAE,IAAI,EAAE,CAAC,CAAC;gBAC/E,MAAM,SAAS,GAAG,KAAK,CAAC,IAAI,CAAC,MAAM,CAAC,IAAoB,CAAC,CAAC;gBAC1D,UAAU,CAAC,IAAI,CAAC,SAAS,CAAC,CAAC;YAC7B,CAAC;YAED,OAAO;gBACL,MAAM,EAAE,SAAS;gBACjB,MAAM,EAAE,UAAU;aACnB,CAAC;QACJ,CAAC;QAAC,OAAO,KAAK,EAAE,CAAC;YACf,OAAO;gBACL,MAAM,EAAE,OAAO;gBACf,OAAO,EAAE,KAAK,YAAY,KAAK,CAAC,CAAC,CAAC,KAAK,CAAC,OAAO,CAAC,CAAC,CAAC,qCAAqC;aACxF,CAAC;QACJ,CAAC;IACH,CAAC;CACF"}