@msbayindir/context-rag 1.0.0-beta.1

package/package.json

@@ -0,0 +1,91 @@
+{
+  "name": "@msbayindir/context-rag",
+  "version": "1.0.0-beta.1",
+  "type": "module",
+  "description": "A powerful, multimodal RAG engine with contextual retrieval, auto-prompt discovery, and PostgreSQL-native vector search",
+  "main": "dist/index.js",
+  "module": "dist/index.mjs",
+  "types": "dist/index.d.ts",
+  "exports": {
+    ".": {
+      "types": "./dist/index.d.ts",
+      "require": "./dist/index.js",
+      "import": "./dist/index.mjs"
+    },
+    "./prisma": {
+      "require": "./dist/database/prisma/schema.prisma",
+      "import": "./dist/database/prisma/schema.prisma"
+    }
+  },
+  "bin": {
+    "context-rag": "./dist/bin/cli.js"
+  },
+  "files": [
+    "dist",
+    "prisma",
+    "README.md",
+    "LICENSE"
+  ],
+  "scripts": {
+    "build": "tsup",
+    "dev": "tsup --watch",
+    "test": "vitest",
+    "test:coverage": "vitest --coverage",
+    "lint": "eslint src --ext .ts",
+    "lint:fix": "eslint src --ext .ts --fix",
+    "format": "prettier --write \"src/**/*.ts\"",
+    "typecheck": "tsc --noEmit",
+    "prepublishOnly": "npm run build"
+  },
+  "keywords": [
+    "rag",
+    "retrieval-augmented-generation",
+    "vector-search",
+    "pgvector",
+    "postgresql",
+    "gemini",
+    "ai",
+    "llm",
+    "pdf",
+    "multimodal",
+    "contextual-retrieval",
+    "embedding"
+  ],
+  "author": "Muhammed Bayindir",
+  "license": "MIT",
+  "repository": {
+    "type": "git",
+    "url": "https://github.com/msbayindir/ContextRAG.git"
+  },
+  "bugs": {
+    "url": "https://github.com/msbayindir/ContextRAG/issues"
+  },
+  "homepage": "https://github.com/msbayindir/ContextRAG#readme",
+  "engines": {
+    "node": ">=18.0.0"
+  },
+  "dependencies": {
+    "@google/generative-ai": "^0.21.0",
+    "@prisma/client": "^6.2.1",
+    "commander": "^12.1.0",
+    "p-limit": "^7.2.0",
+    "pdf-parse": "^1.1.1",
+    "zod": "^3.24.1"
+  },
+  "devDependencies": {
+    "@eslint/js": "^9.39.2",
+    "@types/node": "^22.10.7",
+    "@types/pdf-parse": "^1.1.4",
+    "@typescript-eslint/eslint-plugin": "^8.21.0",
+    "@typescript-eslint/parser": "^8.21.0",
+    "eslint": "^9.18.0",
+    "prettier": "^3.4.2",
+    "prisma": "^6.2.1",
+    "tsup": "^8.3.5",
+    "typescript": "^5.7.3",
+    "vitest": "^2.1.8"
+  },
+  "peerDependencies": {
+    "@prisma/client": ">=5.0.0"
+  }
+}

package/prisma/schema.prisma

@@ -0,0 +1,225 @@
+// This is a reference schema for Context-RAG
+// Users should copy these models to their own schema.prisma
+// or use `npx context-rag init` to add them automatically
+
+generator client {
+  provider        = "prisma-client-js"
+  previewFeatures = ["postgresqlExtensions"]
+}
+
+datasource db {
+  provider   = "postgresql"
+  url        = env("DATABASE_URL")
+  extensions = [vector]
+}
+
+// ============================================
+// Context-RAG Models
+// ============================================
+
+/// Stores prompt configurations for different document types
+/// Each document type can have multiple versions for A/B testing
+model ContextRagPromptConfig {
+  id String @id @default(uuid())
+
+  /// Document type identifier (e.g., "Medical", "Legal", "Technical")
+  documentType String @map("document_type")
+
+  /// Human-readable name for this configuration
+  name String
+
+  /// System prompt used for AI processing
+  systemPrompt String @map("system_prompt") @db.Text
+
+  /// Chunking strategy configuration as JSON
+  chunkStrategy Json @map("chunk_strategy")
+
+  /// Version number for this document type (incremental)
+  version Int @default(1)
+
+  /// Whether this configuration is active
+  isActive Boolean @default(true) @map("is_active")
+
+  /// Whether this is the default config for the document type
+  isDefault Boolean @default(false) @map("is_default")
+
+  /// Who created this: 'discovery' | 'manual' | user ID
+  createdBy String? @map("created_by")
+
+  /// Reason for changes / changelog entry
+  changeLog String? @map("change_log")
+
+  createdAt DateTime @default(now()) @map("created_at")
+  updatedAt DateTime @updatedAt @map("updated_at")
+
+  /// Related chunks created with this config
+  chunks ContextRagChunk[]
+
+  @@unique([documentType, version])
+  @@index([documentType, isActive])
+  @@map("context_rag_prompt_configs")
+}
+
+/// Stores vector chunks for semantic search
+model ContextRagChunk {
+  id String @id @default(uuid())
+
+  /// Reference to the prompt config used to create this chunk
+  promptConfigId String                 @map("prompt_config_id")
+  promptConfig   ContextRagPromptConfig @relation(fields: [promptConfigId], references: [id], onDelete: Cascade)
+
+  /// Reference to the source document
+  documentId String @map("document_id")
+
+  /// Sequential index of this chunk within the document
+  chunkIndex Int @map("chunk_index")
+
+  /// Type of content: TEXT, TABLE, LIST, CODE, HEADING, IMAGE_REF, QUOTE, MIXED
+  chunkType String @map("chunk_type")
+
+  /// Plain text content optimized for vector search (original)
+  searchContent String @map("search_content") @db.Text
+
+  /// Enriched content: context + searchContent (used for embedding when RAG enhancement enabled)
+  enrichedContent String? @map("enriched_content") @db.Text
+
+  /// AI-generated context text only (for debugging/display)
+  contextText String? @map("context_text") @db.Text
+
+  /// Vector embedding (768 dimensions for Gemini)
+  searchVector Unsupported("vector(768)") @map("search_vector")
+
+  /// Rich Markdown content for display
+  displayContent String @map("display_content") @db.Text
+
+  /// Starting page number in source document
+  sourcePageStart Int @map("source_page_start")
+
+  /// Ending page number in source document
+  sourcePageEnd Int @map("source_page_end")
+
+  /// Confidence score from AI processing (0.0 - 1.0)
+  confidenceScore Float @default(0.5) @map("confidence_score")
+
+  /// Additional metadata as JSON
+  /// { page, pageRange, type, confidence, tokens, keywords, ... }
+  metadata Json
+
+  createdAt DateTime @default(now()) @map("created_at")
+
+  @@index([promptConfigId])
+  @@index([documentId])
+  @@index([chunkType])
+  @@index([confidenceScore])
+  @@map("context_rag_chunks")
+}
+
+/// Tracks document processing state
+model ContextRagDocument {
+  id String @id @default(uuid())
+
+  /// Original filename
+  filename String
+
+  /// SHA-256 hash for deduplication
+  fileHash String @map("file_hash")
+
+  /// File size in bytes
+  fileSize Int @map("file_size")
+
+  /// Total page count
+  pageCount Int @map("page_count")
+
+  /// Document type (matches PromptConfig.documentType)
+  documentType String? @map("document_type")
+
+  /// Experiment identifier for A/B testing models
+  /// e.g., "exp_flash_v1", "exp_pro_v2"
+  experimentId String? @map("experiment_id")
+
+  /// AI model used for processing
+  /// e.g., "gemini-3-flash-preview", "gemini-3-pro-preview"
+  modelName String? @map("model_name")
+
+  /// Model configuration as JSON
+  /// { temperature, maxOutputTokens, ... }
+  modelConfig Json? @map("model_config")
+
+  /// Processing status: PENDING, DISCOVERING, AWAITING_APPROVAL, PROCESSING, COMPLETED, FAILED, PARTIAL
+  status String @default("PENDING")
+
+  /// Reference to prompt config used for processing
+  promptConfigId String? @map("prompt_config_id")
+
+  /// Batch processing statistics
+  totalBatches     Int @default(0) @map("total_batches")
+  completedBatches Int @default(0) @map("completed_batches")
+  failedBatches    Int @default(0) @map("failed_batches")
+
+  /// Total token usage as JSON { input, output, total }
+  tokenUsage Json? @map("token_usage")
+
+  /// Total processing time in milliseconds
+  processingMs Int? @map("processing_ms")
+
+  /// Error message if failed
+  errorMessage String? @map("error_message")
+
+  createdAt   DateTime  @default(now()) @map("created_at")
+  completedAt DateTime? @map("completed_at")
+
+  /// Related batch jobs
+  batches ContextRagBatch[]
+
+  /// Unique constraint: same file + same experiment = one record
+  /// Allows same PDF to be processed with different experiments
+  @@unique([fileHash, experimentId])
+  @@index([status])
+  @@index([fileHash])
+  @@index([documentType])
+  @@index([experimentId])
+  @@map("context_rag_documents")
+}
+
+/// Tracks individual batch processing jobs
+model ContextRagBatch {
+  id String @id @default(uuid())
+
+  /// Reference to parent document
+  documentId String             @map("document_id")
+  document   ContextRagDocument @relation(fields: [documentId], references: [id], onDelete: Cascade)
+
+  /// Sequential batch index (0-based)
+  batchIndex Int @map("batch_index")
+
+  /// Starting page of this batch
+  pageStart Int @map("page_start")
+
+  /// Ending page of this batch
+  pageEnd Int @map("page_end")
+
+  /// Batch status: PENDING, PROCESSING, RETRYING, COMPLETED, FAILED
+  status String @default("PENDING")
+
+  /// Number of retry attempts
+  retryCount Int @default(0) @map("retry_count")
+
+  /// Last error message if failed
+  lastError String? @map("last_error")
+
+  /// Token usage for this batch as JSON
+  tokenUsage Json? @map("token_usage")
+
+  /// Processing time in milliseconds
+  processingMs Int? @map("processing_ms")
+
+  startedAt   DateTime? @map("started_at")
+  completedAt DateTime? @map("completed_at")
+  createdAt   DateTime  @default(now()) @map("created_at")
+  updatedAt   DateTime  @updatedAt @map("updated_at")
+
+  @@unique([documentId, batchIndex])
+  @@index([documentId, status])
+  @@index([status])
+  @@map("context_rag_batches")
+}