Octocode MCP

# AI Agent Frameworks & RAG Resources > Production-ready AI agent frameworks, RAG systems, and vector databases for Node.js/TypeScript ** Purpose:** Essential AI resources for building intelligent agents with embeddings, RAG, and rerankers **🌐 Focus:** TypeScript-native frameworks (Mastra, LlamaIndexTS, LangChainJS, Vercel AI SDK) **⚙️ Runtime:** Node.js/TypeScript with production-ready vector databases **📅 Updated:** October 14, 2025 --- ## Quick Reference ### Production Stack (2025) - **Agent Framework:** `mastra-ai/mastra` - TypeScript-native with built-in RAG + observability - **RAG Framework:** `run-llama/LlamaIndexTS` - 100+ data connectors, production-ready - **Vector DB:** `qdrant/qdrant` - Fastest with best Node.js SDK - **Embeddings:** `xenova/transformers.js` - Run models in Node.js/browser - **Reranking:** Cohere Re-rank API - 20-30% accuracy boost - **Observability:** `langfuse/langfuse` - LLM engineering platform ### TypeScript Agent Frameworks (Start Here) - **Best Overall:** `mastra-ai/mastra` (17K+ ⭐) - Complete framework with RAG + observability - **Best RAG:** `run-llama/LlamaIndexTS` (3K+ ⭐) - Comprehensive data framework - **Most Popular:** `langchain-ai/langchainjs` (15K+ ⭐) - Massive ecosystem - **Best for Next.js:** `vercel/ai` (14K+ ⭐) - Streaming + React components ### Vector Databases - **Best Performance:** `qdrant/qdrant` (23K+ ⭐) - Fastest with advanced filtering - **Enterprise Scale:** `milvus-io/milvus` (33K+ ⭐) - Billions of vectors - **Easiest to Use:** `chroma-core/chroma` (18K+ ⭐) - Best DX, rapid prototyping - **With PostgreSQL:** `pgvector/pgvector` (16K+ ⭐) - Keep vectors with relational data ### Embeddings & Reranking - **Node.js Embeddings:** `xenova/transformers.js` (17K+ ⭐) - Run 100+ models in JavaScript - **Reranking (Essential):** Cohere Re-rank API - Improves RAG accuracy 20-30% - **API Services:** OpenAI `text-embedding-3-small`, Cohere Embed v3, Voyage AI ### RAG Techniques (Learn Here) - **Most Comprehensive:** `NirDiamant/RAG_Techniques` (22K+ ⭐) - Complete guide to advanced RAG - **Advanced Patterns:** `NirDiamant/Controllable-RAG-Agent` (1.4K+ ⭐) - Graph-based agent RAG - **Practical Cookbook:** `athina-ai/rag-cookbooks` (2.3K+ ⭐) - Working implementations - **Evaluation:** `explodinggradients/ragas` (9K+ ⭐) - RAG assessment framework ### Multi-Agent Systems - **Complete Platform:** `langgenius/dify` (116K+ ⭐) - Visual workflows + RAG + monitoring - **Team-Based:** `crewAIInc/crewAI` (39K+ ⭐) - Role-based collaborative agents - **Visual Builder:** `FlowiseAI/Flowise` (45K+ ⭐) - Drag-and-drop on LangChain ### Observability - **Best Overall:** `langfuse/langfuse` (17K+ ⭐) - Open-source LLM engineering platform - **Built-in:** Mastra, VoltAgent - Frameworks with observability included --- ## Table of Contents 1. [TypeScript Agent Frameworks](#typescript-agent-frameworks) 2. [Vector Databases](#vector-databases-production-ready) 3. [Embeddings & Reranking](#embeddings--reranking) 4. [RAG Frameworks & Techniques](#rag-frameworks--techniques) 5. [Multi-Agent Systems](#multi-agent-systems) 6. [LangChain Ecosystem](#langchain-ecosystem) 7. [Observability & Evaluation](#observability--evaluation) --- ## TypeScript Agent Frameworks ### Production-Ready TypeScript Frameworks **⭐ mastra-ai/mastra** (17,368+ stars) ⚡ ESSENTIAL - Complete TypeScript AI agent framework with built-in RAG, observability, and assistants - 🔗 https://github.com/mastra-ai/mastra - **Use Case:** THE modern TypeScript framework for AI agents (2025) - **Why Essential:** TypeScript-native, built-in RAG + observability, supports all LLMs - **Features:** Assistants, RAG, workflows, agents, observability, streaming - **Best For:** Production TypeScript AI apps, Next.js integration, modern DX **⭐ VoltAgent/voltagent** (3,579+ stars) ⚡ HIGHLY RECOMMENDED - Open Source TypeScript AI Agent Framework with built-in LLM Observability - 🔗 https://github.com/VoltAgent/voltagent - **Use Case:** TypeScript agent framework with first-class observability - **Why Recommended:** Production-ready with monitoring built-in - **Best For:** TypeScript agents needing observability out of the box **⭐ run-llama/LlamaIndexTS** (2,912+ stars) ⚡ ESSENTIAL - TypeScript/JavaScript data framework for LLM applications - 🔗 https://github.com/run-llama/LlamaIndexTS - **Use Case:** Comprehensive RAG framework for Node.js/TypeScript - **Why Essential:** Best TypeScript alternative to Python LlamaIndex - **Features:** 100+ data connectors, indexing, retrieval, query engines - **Best For:** Production RAG systems in Node.js, data-heavy applications **⭐ langchain-ai/langchainjs** (15,400+ stars) ⚡ ESSENTIAL - Official LangChain JavaScript/TypeScript implementation - 🔗 https://github.com/langchain-ai/langchainjs - **Use Case:** Industry-standard LLM framework with comprehensive ecosystem - **Why Essential:** Most popular, massive ecosystem, production-ready - **Features:** Document loaders, vector stores, retrievers, chains, agents - **Best For:** Complex workflows, extensive integrations, mature ecosystem **⭐ vercel/ai** (13,800+ stars) ⚡ HIGHLY RECOMMENDED - Vercel AI SDK - build AI-powered products with React, Next.js, Vue, Svelte - 🔗 https://github.com/vercel/ai - **Use Case:** Modern RAG in Next.js/React with streaming and UI components - **Why Recommended:** Best DX for full-stack TypeScript RAG applications - **Features:** Streaming responses, RAG utilities, UI components, edge runtime - **Best For:** Next.js apps, streaming RAG, modern web applications --- ## Vector Databases (Production Ready) ### Top 4 Vector Databases for Node.js **⭐ qdrant/qdrant** (22,800+ stars) ⚡ ESSENTIAL - Vector database written in Rust with advanced filtering and excellent Node.js SDK - 🔗 https://github.com/qdrant/qdrant - **Use Case:** #1 choice for production RAG with TypeScript/Node.js - **Why Essential:** Fastest vector search, best Node.js SDK, advanced filtering - **Best For:** Production RAG systems, semantic search with metadata **⭐ milvus-io/milvus** (32,900+ stars) ⚡ ESSENTIAL - Industry-standard vector database for billion-scale similarity search - 🔗 https://github.com/milvus-io/milvus - **Use Case:** Enterprise-scale vector search with cloud-native architecture - **Why Essential:** Most mature, handles billions of vectors, strong Node.js client - **Best For:** Large-scale RAG, recommendation engines, enterprise systems **⭐ chroma-core/chroma** (18,200+ stars) ⚡ HIGHLY RECOMMENDED - AI-native embedding database with best developer experience - 🔗 https://github.com/chroma-core/chroma - **Use Case:** Fastest to prototype, perfect for LangChain/LlamaIndex integration - **Why Recommended:** Minimal setup, great DX, excellent LangChain support - **Best For:** Prototypes, small-medium datasets, rapid development **⭐ pgvector/pgvector** (16,100+ stars) ⚡ HIGHLY RECOMMENDED - Vector similarity search for PostgreSQL - 🔗 https://github.com/pgvector/pgvector - **Use Case:** Keep embeddings with relational data in Postgres - **Why Recommended:** Best for hybrid data models, no separate vector DB needed - **Best For:** Existing Postgres apps, cost-effective RAG ### Managed Services & Utilities **Pinecone** - Fully managed vector database (SaaS) - TypeScript client: `pinecone-io/pinecone-ts-client` - **Best For:** Zero ops, fast time-to-market **⭐ zilliztech/GPTCache** (7,600+ stars) - Semantic cache for LLMs - 95%+ cost reduction - 🔗 https://github.com/zilliztech/GPTCache - **Best For:** Cost optimization for production LLM apps --- ## Embeddings & Reranking ### Embedding Solutions **⭐ xenova/transformers.js** (16,500+ stars) ⚡ ESSENTIAL - Run transformer models directly in Node.js/browser - no Python needed - 🔗 https://github.com/xenova/transformers.js - **Use Case:** THE solution for embeddings in Node.js/TypeScript - **Why Essential:** Run 100+ models in JavaScript, production-ready - **Popular Models:** `all-MiniLM-L6-v2` (fast), `bge-small-en-v1.5` (quality) - **Best For:** Node.js backends, edge computing, self-hosted embeddings **Embedding APIs** ⚡ HIGHLY RECOMMENDED - **OpenAI:** `text-embedding-3-small` (fast/cheap), `text-embedding-3-large` (quality) - **Cohere:** Embed v3 (multilingual, compression support) - **Voyage AI:** Domain-specific (code, finance, legal) - **Best For:** Production systems, managed infrastructure, consistent quality ### Reranking (Critical for RAG Quality) **⭐ Cohere Re-rank API** ⚡ ESSENTIAL - Best-in-class reranking - 20-30% accuracy improvement - **Use Case:** Rerank top-k retrieved documents before sending to LLM - **Why Essential:** Dramatically improves RAG accuracy with minimal effort - **Best For:** All production RAG systems **⭐ FlagOpen/FlagEmbedding** (8,700+ stars) (Python) - BAAI's embedding and re-ranking models (bge-reranker) - 🔗 https://github.com/FlagOpen/FlagEmbedding - **Use Case:** Self-hosted re-ranking models - **Node.js Alternative:** Use Cohere Re-rank API or deploy via inference server --- ## RAG Frameworks & Techniques ### RAG Frameworks (See [TypeScript Agent Frameworks](#typescript-agent-frameworks) for Mastra, LlamaIndexTS, LangChain, Vercel AI SDK) ### Advanced RAG Techniques (Must-Read) **⭐ NirDiamant/RAG_Techniques** (22,355+ stars) ⚡ ESSENTIAL - Comprehensive repository of advanced RAG techniques and implementations - 🔗 https://github.com/NirDiamant/RAG_Techniques - **Use Case:** THE complete guide to cutting-edge RAG patterns - **Why Essential:** Most comprehensive RAG techniques collection (2025) - **Techniques:** Hybrid search, re-ranking, query expansion, fusion, citation tracking - **Best For:** Learning advanced RAG, implementing SOTA patterns **⭐ NirDiamant/Controllable-RAG-Agent** (1,448+ stars) ⚡ HIGHLY RECOMMENDED - Advanced RAG with graph-based agent architecture and controllable retrieval - 🔗 https://github.com/NirDiamant/Controllable-RAG-Agent - **Use Case:** State-of-the-art RAG with agent-controlled retrieval - **Why Recommended:** Implements latest RAG research (2024-2025) - **Features:** Adaptive retrieval, query decomposition, fusion strategies - **Best For:** Complex RAG systems, multi-step reasoning **⭐ athina-ai/rag-cookbooks** (2,280+ stars) ⚡ HIGHLY RECOMMENDED - Advanced RAG techniques cookbook with practical implementations - 🔗 https://github.com/athina-ai/rag-cookbooks - **Use Case:** Hands-on RAG cookbook with working examples - **Why Recommended:** Practical implementations of advanced techniques - **Best For:** Production RAG development, learning by example ### RAG Evaluation & Optimization **⭐ explodinggradients/ragas** (9,200+ stars) ⚡ ESSENTIAL (Python) - RAG Assessment framework - evaluation metrics for RAG pipelines - 🔗 https://github.com/explodinggradients/ragas - **Use Case:** Automated RAG evaluation (faithfulness, relevancy, recall) - **Why Essential:** Standard RAG evaluation metrics - **Best For:** RAG quality assessment, A/B testing - **Node.js Alternative:** LangSmith or custom LLM-as-judge patterns **⭐ langchain-ai/langsmith-cookbook** (1,900+ stars) - LangSmith cookbook with RAG evaluation examples - 🔗 https://github.com/langchain-ai/langsmith-cookbook - **Use Case:** Production RAG monitoring with LangSmith - **Best For:** LangChain projects, production observability ### Production RAG Templates **⭐ a16z-infra/ai-getting-started** (4,900+ stars) ⚡ HIGHLY RECOMMENDED - A16z's production-ready RAG stack (Next.js + Pinecone + OpenAI) - 🔗 https://github.com/a16z-infra/ai-getting-started - **Use Case:** Starting point for production RAG applications - **Why Recommended:** Best practices from leading VC - **Best For:** Next.js RAG applications --- ## Multi-Agent Systems ### Multi-Agent Collaboration Frameworks **⭐ langgenius/dify** (116,335+ stars) ⚡ ESSENTIAL - Complete AI agent development platform with visual workflows, RAG, and observability - 🔗 https://github.com/langgenius/dify - **Use Case:** THE production platform for agentic workflows (2025 leader) - **Why Essential:** Most popular LLM platform, combines workflows + RAG + monitoring - **Best For:** Enterprise agent applications, visual development **⭐ crewAIInc/crewAI** (39,166+ stars) ⚡ ESSENTIAL (Python) - Role-based collaborative AI agent framework (1M+ monthly downloads) - 🔗 https://github.com/crewAIInc/crewAI - **Use Case:** Multi-agent teams working as coordinated crew - **Why Essential:** #1 for team-based agents, minimal setup, production-proven - **Node.js Alternative:** `langchain-ai/langgraphjs` for TypeScript multi-agent **⭐ microsoft/autogen** (Python + .NET) ⚡ HIGHLY RECOMMENDED - Microsoft's multi-agent conversation framework - 🔗 https://github.com/microsoft/autogen - **Use Case:** Enterprise multi-agent with structured interactions - **Why Recommended:** Microsoft-backed, enterprise-grade - **Best For:** Complex research workflows, enterprise deployments ### Visual Agent Builders **⭐ FlowiseAI/Flowise** (45,496+ stars) ⚡ HIGHLY RECOMMENDED - Drag-and-drop AI agent builder on LangChain - 🔗 https://github.com/FlowiseAI/Flowise - **Use Case:** No-code/low-code agent development - **Why Recommended:** Best visual builder, rapid prototyping - **Best For:** Non-developers, rapid prototyping, LangChain workflows ### Specialized Agent Applications **⭐ cline/cline** (51,262+ stars) ⚡ HIGHLY RECOMMENDED - Autonomous coding agent for VSCode - 🔗 https://github.com/cline/cline - **Use Case:** AI pair programmer in IDE - **Why Recommended:** Top coding agent, highly autonomous **⭐ assafelovic/gpt-researcher** (23,793+ stars) ⚡ HIGHLY RECOMMENDED (Python) - Autonomous research agent with web search and citations - 🔗 https://github.com/assafelovic/gpt-researcher - **Use Case:** Deep research with proper citations - **Why Recommended:** Industry-standard research automation - **Node.js Alternative:** Build with LangChain + web search tools **⭐ khoj-ai/khoj** (31,313+ stars) ⚡ HIGHLY RECOMMENDED - Self-hosted AI second brain with web and doc search - 🔗 https://github.com/khoj-ai/khoj - **Use Case:** Personal knowledge management with AI - **Why Recommended:** Best self-hosted knowledge assistant **⭐ CopilotKit/CopilotKit** (24,408+ stars) ⚡ HIGHLY RECOMMENDED - React UI + infrastructure for in-app AI copilots - 🔗 https://github.com/CopilotKit/CopilotKit - **Use Case:** Embed AI agents in React applications - **Why Recommended:** Best React integration for agents ### Learning Resources **⭐ NirDiamant/GenAI_Agents** (17,250+ stars) ⚡ HIGHLY RECOMMENDED - Comprehensive tutorials for Generative AI Agent techniques - 🔗 https://github.com/NirDiamant/GenAI_Agents - **Use Case:** Learn agent patterns and implementations - **Why Recommended:** Comprehensive, practical examples - **Best For:** Learning agent development, reference implementations **⭐ e2b-dev/awesome-ai-agents** (23,397+ stars) - Curated list of AI autonomous agents - 🔗 https://github.com/e2b-dev/awesome-ai-agents - **Use Case:** Discover agent projects and tools --- ## LangChain Ecosystem ### Core Libraries **⭐ langchain-ai/langchainjs** ⚡ ESSENTIAL - Official LangChain JavaScript/TypeScript implementation - 🔗 https://github.com/langchain-ai/langchainjs - **Use Case:** Industry-standard LLM framework for TypeScript - **Why Essential:** Massive ecosystem, production-ready, 15K+ stars - (See [TypeScript Agent Frameworks](#typescript-agent-frameworks) for details) **⭐ langchain-ai/langgraph** ⚡ HIGHLY RECOMMENDED - Stateful, graph-based agent workflows for complex multi-agent systems - 🔗 https://github.com/langchain-ai/langgraph - **Use Case:** Advanced agent orchestration with cyclical graphs - **Why Recommended:** Best for complex agent interactions and state management ### LangChain/LangGraph Examples **⭐ wassim249/fastapi-langgraph-agent-production-ready-template** (1,379+ stars) - Production-ready FastAPI + LangGraph template - 🔗 https://github.com/wassim249/fastapi-langgraph-agent-production-ready-template - **Use Case:** Production LangGraph starting point **⭐ von-development/awesome-LangGraph** (1,153+ stars) - Comprehensive LangChain + LangGraph ecosystem index - 🔗 https://github.com/von-development/awesome-LangGraph - **Use Case:** Discover LangGraph resources and tools --- ## Observability & Evaluation **⭐ langfuse/langfuse** (17,000+ stars) ⚡ ESSENTIAL - Open source LLM engineering platform with observability, metrics, and evals - 🔗 https://github.com/langfuse/langfuse - **Use Case:** THE platform for LLM observability and monitoring - **Why Essential:** Industry-standard for production LLM monitoring - **Features:** Tracing, metrics, evals, prompt management, cost tracking - **Best For:** Production LLM apps, debugging, optimization **⭐ Built-in Observability** - **Mastra** - Built-in observability in TypeScript agent framework - **VoltAgent** - TypeScript framework with first-class observability - **LangSmith** - Official LangChain observability (paid/free tiers) **⭐ Evaluation Frameworks** - **RAGAS** - RAG evaluation metrics (Python) - **LangSmith Cookbook** - Evaluation examples for LangChain - **Custom LLM-as-judge** - Build evaluation with GPT-4/Claude --- *Part of octocode-mcp resources collection*