@objectstack/spec 0.2.0 → 0.3.1

package/dist/ai/rag-pipeline.zod.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"rag-pipeline.zod.d.ts","sourceRoot":"","sources":["../../src/ai/rag-pipeline.zod.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,CAAC,EAAE,MAAM,KAAK,CAAC;AAExB;;;;;GAKG;AAEH;;GAEG;AACH,eAAO,MAAM,yBAAyB,iIAWpC,CAAC;AAEH;;GAEG;AACH,eAAO,MAAM,oBAAoB;;;;;;;;;;;;;;;;;;;;;;;;EAQ/B,CAAC;AAEH;;GAEG;AACH,eAAO,MAAM,sBAAsB;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;IAyBjC,CAAC;AAEH;;GAEG;AACH,eAAO,MAAM,sBAAsB;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;EAWjC,CAAC;AAEH;;GAEG;AACH,eAAO,MAAM,mBAAmB;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;EAO9B,CAAC;AAEH;;GAEG;AACH,eAAO,MAAM,uBAAuB;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;IAuBlC,CAAC;AAEH;;GAEG;AACH,eAAO,MAAM,qBAAqB;;;;;;;;;;;;;;;EAKhC,CAAC;AAEH;;GAEG;AACH,eAAO,MAAM,uBAAuB;;;;IAKlC,iBAAiB;;;;IAKjB,oBAAoB;;;IAIpB,kBAAkB;;;;;;;;;;;;;;;;;;;;;;;;;;;;EAIlB,CAAC;AAEH;;GAEG;AACH,eAAO,MAAM,0BAA0B;;IAGrC,aAAa;;IAGb,iBAAiB;;IAGjB,iBAAiB;;;;IAKjB,sBAAsB;;;IAItB,oBAAoB;;;;;;;;;;;;;;;;;;;;;;EAEpB,CAAC;AAEH;;GAEG;AACH,eAAO,MAAM,uBAAuB;IAClC,eAAe;;;;IAKf,iBAAiB;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;QAjDjB,iBAAiB;;;;QAKjB,oBAAoB;;;QAIpB,kBAAkB;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;IA+ClB,uBAAuB;;;QAnCvB,aAAa;;QAGb,iBAAiB;;QAGjB,iBAAiB;;;;QAKjB,sBAAsB;;;QAItB,oBAAoB;;;;;;;;;;;;;;;;;;;;;;;IAuBpB,yBAAyB;;;IAIzB,yBAAyB;;IAQzB,cAAc;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;EAId,CAAC;AAEH;;GAEG;AACH,eAAO,MAAM,qBAAqB;;;IAIhC,wBAAwB;;;IAIxB,cAAc;;;;;;;;;;;IAMd,cAAc;;;;;;;;;;;;;;;;;;;;;;;;;EAGd,CAAC;AAEH;;GAEG;AACH,eAAO,MAAM,sBAAsB;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;EAWjC,CAAC;AAEH;;GAEG;AACH,eAAO,MAAM,uBAAuB;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;EAUlC,CAAC;AAGH,MAAM,MAAM,mBAAmB,GAAG,CAAC,CAAC,KAAK,CAAC,OAAO,yBAAyB,CAAC,CAAC;AAC5E,MAAM,MAAM,cAAc,GAAG,CAAC,CAAC,KAAK,CAAC,OAAO,oBAAoB,CAAC,CAAC;AAClE,MAAM,MAAM,gBAAgB,GAAG,CAAC,CAAC,KAAK,CAAC,OAAO,sBAAsB,CAAC,CAAC;AACtE,MAAM,MAAM,gBAAgB,GAAG,CAAC,CAAC,KAAK,CAAC,OAAO,sBAAsB,CAAC,CAAC;AACtE,MAAM,MAAM,aAAa,GAAG,CAAC,CAAC,KAAK,CAAC,OAAO,mBAAmB,CAAC,CAAC;AAChE,MAAM,MAAM,iBAAiB,GAAG,CAAC,CAAC,KAAK,CAAC,OAAO,uBAAuB,CAAC,CAAC;AACxE,MAAM,MAAM,eAAe,GAAG,CAAC,CAAC,KAAK,CAAC,OAAO,qBAAqB,CAAC,CAAC;AACpE,MAAM,MAAM,iBAAiB,GAAG,CAAC,CAAC,KAAK,CAAC,OAAO,uBAAuB,CAAC,CAAC;AACxE,MAAM,MAAM,oBAAoB,GAAG,CAAC,CAAC,KAAK,CAAC,OAAO,0BAA0B,CAAC,CAAC;AAC9E,MAAM,MAAM,iBAAiB,GAAG,CAAC,CAAC,KAAK,CAAC,OAAO,uBAAuB,CAAC,CAAC;AACxE,MAAM,MAAM,eAAe,GAAG,CAAC,CAAC,KAAK,CAAC,OAAO,qBAAqB,CAAC,CAAC;AACpE,MAAM,MAAM,gBAAgB,GAAG,CAAC,CAAC,KAAK,CAAC,OAAO,sBAAsB,CAAC,CAAC;AACtE,MAAM,MAAM,iBAAiB,GAAG,CAAC,CAAC,KAAK,CAAC,OAAO,uBAAuB,CAAC,CAAC"}

package/dist/ai/rag-pipeline.zod.js

@@ -0,0 +1,244 @@
+"use strict";
+Object.defineProperty(exports, "__esModule", { value: true });
+exports.RAGPipelineStatusSchema = exports.RAGQueryResponseSchema = exports.RAGQueryRequestSchema = exports.RAGPipelineConfigSchema = exports.DocumentLoaderConfigSchema = exports.VectorStoreConfigSchema = exports.RerankingConfigSchema = exports.RetrievalStrategySchema = exports.DocumentChunkSchema = exports.DocumentMetadataSchema = exports.ChunkingStrategySchema = exports.EmbeddingModelSchema = exports.VectorStoreProviderSchema = void 0;
+const zod_1 = require("zod");
+/**
+ * RAG (Retrieval-Augmented Generation) Pipeline Protocol
+ *
+ * Defines schemas for building context-aware AI assistants using RAG techniques.
+ * Enables vector search, document chunking, embeddings, and retrieval configuration.
+ */
+/**
+ * Vector Store Provider
+ */
+exports.VectorStoreProviderSchema = zod_1.z.enum([
+    'pinecone',
+    'weaviate',
+    'qdrant',
+    'milvus',
+    'chroma',
+    'pgvector',
+    'redis',
+    'opensearch',
+    'elasticsearch',
+    'custom',
+]);
+/**
+ * Embedding Model
+ */
+exports.EmbeddingModelSchema = zod_1.z.object({
+    provider: zod_1.z.enum(['openai', 'cohere', 'huggingface', 'azure_openai', 'local', 'custom']),
+    model: zod_1.z.string().describe('Model name (e.g., "text-embedding-3-large")'),
+    dimensions: zod_1.z.number().int().positive().describe('Embedding vector dimensions'),
+    maxTokens: zod_1.z.number().int().positive().optional().describe('Maximum tokens per embedding'),
+    batchSize: zod_1.z.number().int().positive().optional().default(100).describe('Batch size for embedding'),
+    endpoint: zod_1.z.string().url().optional().describe('Custom endpoint URL'),
+    apiKey: zod_1.z.string().optional().describe('API key or reference to secret'),
+});
+/**
+ * Text Chunking Strategy
+ */
+exports.ChunkingStrategySchema = zod_1.z.discriminatedUnion('type', [
+    zod_1.z.object({
+        type: zod_1.z.literal('fixed'),
+        chunkSize: zod_1.z.number().int().positive().describe('Fixed chunk size in tokens/chars'),
+        chunkOverlap: zod_1.z.number().int().min(0).default(0).describe('Overlap between chunks'),
+        unit: zod_1.z.enum(['tokens', 'characters']).default('tokens'),
+    }),
+    zod_1.z.object({
+        type: zod_1.z.literal('semantic'),
+        model: zod_1.z.string().optional().describe('Model for semantic chunking'),
+        minChunkSize: zod_1.z.number().int().positive().default(100),
+        maxChunkSize: zod_1.z.number().int().positive().default(1000),
+    }),
+    zod_1.z.object({
+        type: zod_1.z.literal('recursive'),
+        separators: zod_1.z.array(zod_1.z.string()).default(['\n\n', '\n', ' ', '']),
+        chunkSize: zod_1.z.number().int().positive(),
+        chunkOverlap: zod_1.z.number().int().min(0).default(0),
+    }),
+    zod_1.z.object({
+        type: zod_1.z.literal('markdown'),
+        maxChunkSize: zod_1.z.number().int().positive().default(1000),
+        respectHeaders: zod_1.z.boolean().default(true).describe('Keep headers with content'),
+        respectCodeBlocks: zod_1.z.boolean().default(true).describe('Keep code blocks intact'),
+    }),
+]);
+/**
+ * Document Metadata Schema
+ */
+exports.DocumentMetadataSchema = zod_1.z.object({
+    source: zod_1.z.string().describe('Document source (file path, URL, etc.)'),
+    sourceType: zod_1.z.enum(['file', 'url', 'api', 'database', 'custom']).optional(),
+    title: zod_1.z.string().optional(),
+    author: zod_1.z.string().optional(),
+    createdAt: zod_1.z.string().optional().describe('ISO timestamp'),
+    updatedAt: zod_1.z.string().optional().describe('ISO timestamp'),
+    tags: zod_1.z.array(zod_1.z.string()).optional(),
+    category: zod_1.z.string().optional(),
+    language: zod_1.z.string().optional().describe('Document language (ISO 639-1 code)'),
+    custom: zod_1.z.record(zod_1.z.any()).optional().describe('Custom metadata fields'),
+});
+/**
+ * Document Chunk
+ */
+exports.DocumentChunkSchema = zod_1.z.object({
+    id: zod_1.z.string().describe('Unique chunk identifier'),
+    content: zod_1.z.string().describe('Chunk text content'),
+    embedding: zod_1.z.array(zod_1.z.number()).optional().describe('Embedding vector'),
+    metadata: exports.DocumentMetadataSchema,
+    chunkIndex: zod_1.z.number().int().min(0).describe('Chunk position in document'),
+    tokens: zod_1.z.number().int().optional().describe('Token count'),
+});
+/**
+ * Retrieval Strategy
+ */
+exports.RetrievalStrategySchema = zod_1.z.discriminatedUnion('type', [
+    zod_1.z.object({
+        type: zod_1.z.literal('similarity'),
+        topK: zod_1.z.number().int().positive().default(5).describe('Number of results to retrieve'),
+        scoreThreshold: zod_1.z.number().min(0).max(1).optional().describe('Minimum similarity score'),
+    }),
+    zod_1.z.object({
+        type: zod_1.z.literal('mmr'),
+        topK: zod_1.z.number().int().positive().default(5),
+        fetchK: zod_1.z.number().int().positive().default(20).describe('Initial fetch size'),
+        lambda: zod_1.z.number().min(0).max(1).default(0.5).describe('Diversity vs relevance (0=diverse, 1=relevant)'),
+    }),
+    zod_1.z.object({
+        type: zod_1.z.literal('hybrid'),
+        topK: zod_1.z.number().int().positive().default(5),
+        vectorWeight: zod_1.z.number().min(0).max(1).default(0.7).describe('Weight for vector search'),
+        keywordWeight: zod_1.z.number().min(0).max(1).default(0.3).describe('Weight for keyword search'),
+    }),
+    zod_1.z.object({
+        type: zod_1.z.literal('parent_document'),
+        topK: zod_1.z.number().int().positive().default(5),
+        retrieveParent: zod_1.z.boolean().default(true).describe('Retrieve full parent document'),
+    }),
+]);
+/**
+ * Reranking Configuration
+ */
+exports.RerankingConfigSchema = zod_1.z.object({
+    enabled: zod_1.z.boolean().default(false),
+    model: zod_1.z.string().optional().describe('Reranking model name'),
+    provider: zod_1.z.enum(['cohere', 'huggingface', 'custom']).optional(),
+    topK: zod_1.z.number().int().positive().default(3).describe('Final number of results after reranking'),
+});
+/**
+ * Vector Store Configuration
+ */
+exports.VectorStoreConfigSchema = zod_1.z.object({
+    provider: exports.VectorStoreProviderSchema,
+    indexName: zod_1.z.string().describe('Index/collection name'),
+    namespace: zod_1.z.string().optional().describe('Namespace for multi-tenancy'),
+    /** Connection */
+    host: zod_1.z.string().optional().describe('Vector store host'),
+    port: zod_1.z.number().int().optional().describe('Vector store port'),
+    apiKey: zod_1.z.string().optional().describe('API key or reference to secret'),
+    /** Configuration */
+    dimensions: zod_1.z.number().int().positive().describe('Vector dimensions'),
+    metric: zod_1.z.enum(['cosine', 'euclidean', 'dotproduct']).optional().default('cosine'),
+    /** Performance */
+    batchSize: zod_1.z.number().int().positive().optional().default(100),
+    connectionPoolSize: zod_1.z.number().int().positive().optional().default(10),
+    timeout: zod_1.z.number().int().positive().optional().default(30000).describe('Timeout in milliseconds'),
+});
+/**
+ * Document Loader Configuration
+ */
+exports.DocumentLoaderConfigSchema = zod_1.z.object({
+    type: zod_1.z.enum(['file', 'directory', 'url', 'api', 'database', 'custom']),
+    /** Source */
+    source: zod_1.z.string().describe('Source path, URL, or identifier'),
+    /** File Types */
+    fileTypes: zod_1.z.array(zod_1.z.string()).optional().describe('Accepted file extensions (e.g., [".pdf", ".md"])'),
+    /** Processing */
+    recursive: zod_1.z.boolean().optional().default(false).describe('Process directories recursively'),
+    maxFileSize: zod_1.z.number().int().optional().describe('Maximum file size in bytes'),
+    excludePatterns: zod_1.z.array(zod_1.z.string()).optional().describe('Patterns to exclude'),
+    /** Text Extraction */
+    extractImages: zod_1.z.boolean().optional().default(false).describe('Extract text from images (OCR)'),
+    extractTables: zod_1.z.boolean().optional().default(false).describe('Extract and format tables'),
+    /** Custom Loader */
+    loaderConfig: zod_1.z.record(zod_1.z.any()).optional().describe('Custom loader-specific config'),
+});
+/**
+ * RAG Pipeline Configuration
+ */
+exports.RAGPipelineConfigSchema = zod_1.z.object({
+    /** Identity */
+    name: zod_1.z.string().regex(/^[a-z_][a-z0-9_]*$/).describe('Pipeline name (snake_case)'),
+    label: zod_1.z.string().describe('Display name'),
+    description: zod_1.z.string().optional(),
+    /** Components */
+    embedding: exports.EmbeddingModelSchema,
+    vectorStore: exports.VectorStoreConfigSchema,
+    chunking: exports.ChunkingStrategySchema,
+    retrieval: exports.RetrievalStrategySchema,
+    reranking: exports.RerankingConfigSchema.optional(),
+    /** Document Loading */
+    loaders: zod_1.z.array(exports.DocumentLoaderConfigSchema).optional().describe('Document loaders'),
+    /** Context Management */
+    maxContextTokens: zod_1.z.number().int().positive().default(4000).describe('Maximum tokens in context'),
+    contextWindow: zod_1.z.number().int().positive().optional().describe('LLM context window size'),
+    /** Metadata Filtering */
+    metadataFilters: zod_1.z.record(zod_1.z.union([
+        zod_1.z.string(),
+        zod_1.z.number(),
+        zod_1.z.boolean(),
+        zod_1.z.array(zod_1.z.union([zod_1.z.string(), zod_1.z.number()])),
+    ])).optional().describe('Filters for retrieval (e.g., {category: "docs", status: "published"})'),
+    /** Caching */
+    enableCache: zod_1.z.boolean().default(true),
+    cacheTTL: zod_1.z.number().int().positive().default(3600).describe('Cache TTL in seconds'),
+    cacheInvalidationStrategy: zod_1.z.enum(['time_based', 'manual', 'on_update']).default('time_based').optional(),
+});
+/**
+ * RAG Query Request
+ */
+exports.RAGQueryRequestSchema = zod_1.z.object({
+    query: zod_1.z.string().describe('User query'),
+    pipelineName: zod_1.z.string().describe('Pipeline to use'),
+    /** Override defaults */
+    topK: zod_1.z.number().int().positive().optional(),
+    metadataFilters: zod_1.z.record(zod_1.z.any()).optional(),
+    /** Context */
+    conversationHistory: zod_1.z.array(zod_1.z.object({
+        role: zod_1.z.enum(['user', 'assistant', 'system']),
+        content: zod_1.z.string(),
+    })).optional(),
+    /** Options */
+    includeMetadata: zod_1.z.boolean().default(true),
+    includeSources: zod_1.z.boolean().default(true),
+});
+/**
+ * RAG Query Response
+ */
+exports.RAGQueryResponseSchema = zod_1.z.object({
+    query: zod_1.z.string(),
+    results: zod_1.z.array(zod_1.z.object({
+        content: zod_1.z.string(),
+        score: zod_1.z.number(),
+        metadata: exports.DocumentMetadataSchema.optional(),
+        chunkId: zod_1.z.string().optional(),
+    })),
+    context: zod_1.z.string().describe('Assembled context for LLM'),
+    tokensUsed: zod_1.z.number().int().optional(),
+    retrievalTime: zod_1.z.number().optional().describe('Retrieval time in milliseconds'),
+});
+/**
+ * RAG Pipeline Status
+ */
+exports.RAGPipelineStatusSchema = zod_1.z.object({
+    name: zod_1.z.string(),
+    status: zod_1.z.enum(['active', 'indexing', 'error', 'disabled']),
+    documentsIndexed: zod_1.z.number().int().min(0),
+    lastIndexed: zod_1.z.string().optional().describe('ISO timestamp'),
+    errorMessage: zod_1.z.string().optional(),
+    health: zod_1.z.object({
+        vectorStore: zod_1.z.enum(['healthy', 'unhealthy', 'unknown']),
+        embeddingService: zod_1.z.enum(['healthy', 'unhealthy', 'unknown']),
+    }).optional(),
+});