MS-RAGS(ALL-IN-ONE) Docs

Reference

Official Resources

Use these links when you need the official service documentation, pricing, setup instructions, API references, or product pages behind a feature in MS-RAGS(ALL-IN-ONE).

How To Use This Page

These are the official websites for the third-party services and libraries mentioned across the MS-RAGS(ALL-IN-ONE) Docs. Use them to confirm current pricing, model names, API limits, deployment instructions, data-handling policies, and SDK setup before using a provider in production.

Before entering credentials

Open the provider's official site, confirm the account is active, confirm the model or database feature exists in your region, and create a scoped API key.

Before changing embeddings

Check the model dimension and create a fresh collection or index when dimensions change. Mixing dimensions breaks ingestion or weakens retrieval.

Before launch

Review rate limits, logs, tracing support, retention policy, backups, and whether your documents may be sent to an external hosted API.

MS-RAGS(ALL-IN-ONE) Project

LLM And Embedding Providers

Use these services for chat generation, embeddings, reranking, query rewriting, or evaluation depending on the selected pipeline. Hosted providers require credentials and send requests to external APIs; local providers keep inference on the user's machine when configured that way.

  • OpenAI
    General hosted chat and embedding default.
  • Anthropic
    Hosted chat models for careful long-form responses.
  • Cohere
    Embeddings and hosted reranking.
  • Hugging Face
    Hosted inference endpoints and model ecosystem.
  • Google Gemini API
    Google-hosted generation workflows.
  • Mistral AI
    Hosted chat and embedding models.
  • Groq
    Low-latency hosted inference.
  • Together AI
    Hosted open model inference.
  • Replicate
    Hosted model experimentation.
  • Azure OpenAI Service
    Azure enterprise OpenAI deployments.
  • AWS Bedrock
    AWS-managed foundation model access.
  • Ollama
    Local Ollama plus Ollama Cloud for chat; embeddings should use local or self-hosted Ollama.

Frameworks And Runtime Libraries

These libraries power the generated RAG pipeline, agentic workflows, terminal UI, automated tests, and production telemetry.

  • LangChain
    Core chains, retrievers, loaders, vector store integrations, and LCEL composition.
  • LangGraph
    State-machine workflows for Agentic RAG, Self-RAG, CRAG, and Adaptive RAG.
  • Rich terminal UI
    Readable tables, panels, progress, and status output in the CLI.
  • Hypothesis testing
    Property-based tests for validation and generated-code invariants.
  • OpenTelemetry
    Vendor-neutral traces for ingestion, retrieval, query, and generation phases.

Vector Databases

Vector databases store embeddings and metadata. The most important production checks are embedding dimension compatibility, collection naming, filtering support, backup strategy, and a live add/query smoke test with the exact backend users will run.

  • ChromaDB
    Local-first development and simple persisted collections.
  • FAISS
    Fast local similarity search with file-backed persistence in MS-RAGS(ALL-IN-ONE).
  • Pinecone
    Managed vector database for production deployments.
  • Qdrant
    Local or cloud vector search with filtering-friendly production behavior.
  • Weaviate
    Schema-rich vector database with hybrid search options.
  • Milvus
    Large-scale vector database for heavier infrastructure teams.
  • Redis
    Low-latency vector search when Redis is already in the stack.
  • PGVector
    PostgreSQL-native vector search through the pgvector extension.
  • Elasticsearch
    Enterprise search combining lexical and vector retrieval.
  • OpenSearch
    Open search stack commonly used in AWS environments.
  • Azure AI Search
    Azure-hosted search with vector search capabilities.
  • MongoDB Atlas Vector Search
    Vector search for MongoDB Atlas document applications.

Document Parsing And Web Extraction

Extraction quality controls the quality of the final answers. When a simple loader loses layout, tables, OCR text, or web-rendered content, use a more specialized parser and inspect retrieved chunks before launch.

  • Unstructured
    Document parsing for PDFs, Word, PowerPoint, HTML, images, and more.
  • LlamaParse
    Advanced document parsing for complex PDF and layout-heavy content.
  • Firecrawl
    Web crawling and page extraction for RAG ingestion.
  • Apify
    Web scraping and automation platform for dynamic websites.
  • pypdf
    Python PDF text extraction and manipulation.
  • pdfplumber
    PDF extraction with stronger layout and table inspection.

Evaluation And Observability

Production RAG should be measured. Use evaluation tools for answer quality and tracing tools for operational visibility into ingestion, retrieval, reranking, generation, and provider failures.

  • RAGAS
    Faithfulness, context precision, recall, and answer quality evaluation.
  • DeepEval
    LLM evaluation test cases and metrics.
  • TruLens
    Feedback functions and RAG quality analysis.
  • LangSmith
    Tracing, debugging, datasets, and evaluation for LangChain apps.
  • Langfuse
    Open-source LLM observability and prompt monitoring.
  • Arize Phoenix
    LLM observability, tracing, and evaluation workflows.

Security

Before public release, review the security checklist so secrets, allowlists, logging, and generated code stay safe.

Security checklist for MS-RAGS(ALL-IN-ONE)

Beginner Shortcut

If you do not know which service to pick yet, start with one hosted LLM, one hosted embedding model, ChromaDB locally, Recursive Character chunking, Dense Vector retrieval, and no reranker. Once answers work, move one layer at a time: better extraction, better chunking, better retrieval, reranking, evaluation, and tracing.