Workflow automationRAG & knowledgeAgent frameworks

Dify

langgenius/dify

Agentic workflows and RAG pipelines in one collaborative workspace, deployable to cloud, VPC, or your own servers.

Repository
Stars
152k
Forks
24k
Open issues
932
Language
TypeScript
Licence
See repo
Created
2023-04-12
Last push
2026-08-10
Effort to adopt

Some wiring

A few days of integration work — credentials, data mapping, a deploy target.

01

What it actually does

Dify is an LLM application development platform that puts agentic workflows and RAG pipelines in the same workspace, so a team can move from prototype to production without rebuilding on a different stack.

That last point is the actual pitch. The common failure pattern is prototyping in one tool, discovering it will not hold production traffic or meet a compliance requirement, and starting again. Dify's answer is one workspace with cloud, VPC, and self-hosted deployment options.

It is collaborative by design — several people work in the same workspace rather than one engineer owning a notebook. It supports a wide range of models and tools, MCP, and both low-code and no-code building alongside custom Python.

152,000 stars and a fast release cadence. The licence is non-standard, so read it before embedding Dify in a commercial product.

02

Who it's for

  • 01

    Teams building customer-facing LLM features who need a production path

  • 02

    Companies with VPC or on-premise requirements that rule out hosted-only tools

  • 03

    Mixed teams where a product person and an engineer build together

  • 04

    Anyone combining retrieval over their own documents with agentic steps

EngineeringOperationsSupportMarketingSaaSFintechHealthtechAny industry
03

Where it earns its keep

  • Building a support assistant grounded in your own documentation
  • Standing up a RAG pipeline without assembling five separate components
  • Prototyping with a product manager then hardening the same artifact for production
  • Deploying inside a VPC where data cannot reach a third-party service
  • Swapping model providers without rewriting the application
04

Use it, or skip it

Reach for it when

  • You need both retrieval and agentic orchestration in one place
  • Deployment flexibility is a hard requirement
  • More than one person needs to work on the same LLM application
  • You want to avoid the prototype-to-production rewrite

Skip it when

  • Your need is pure workflow integration without retrieval — n8n is simpler
  • You are building deep custom agent logic; a framework gives you more control
  • The licence terms conflict with how you intend to commercialise
  • You have one developer and one use case; the platform is overhead
05

10 automations

  1. 01Support

    Docs-grounded support assistant

    Answer customer questions from your live documentation, with citations, and escalate anything below a confidence threshold.

  2. 02Sales

    Sales knowledge assistant

    Let reps query product docs, pricing rules, and past deal notes mid-call and get a sourced answer.

  3. 03Operations

    Internal policy lookup

    RAG over HR and finance policies so routine questions stop landing in someone's inbox.

  4. 04Support

    Onboarding assistant per customer

    Build an assistant grounded in a specific customer's configuration to guide their team through setup.

  5. 05Sales

    VPC-deployed assistant

    Run the same application inside a customer's VPC for enterprise deals that will not allow external processing.

  6. 06Marketing

    Content research pipeline

    Chain retrieval over your research library into a drafting step, with a human approving before publication.

  7. 07Finance

    Model cost comparison

    Run the same workflow against two providers and compare quality and cost before committing.

  8. 08Operations

    Structured data extraction

    Extract consistent fields from inbound documents and write them into your systems automatically.

  9. 09Engineering

    Prompt version control

    Manage prompt iterations in the shared workspace so changes are reviewable rather than pasted around.

  10. 10Operations

    Human-approval publishing

    Insert an approval step before any AI-generated content reaches a customer-facing surface.

Want one of these running by Friday?

LimeDock builds these as real workflows inside your stack — deployed to your cloud, wired into your Slack and CRM, with the code in your repo. You pay a build fee and your own API keys, nothing else.

Book a workflow call
07

Source

Repository stats were read from the GitHub API and reflect the last time we refreshed this entry. The editorial breakdown above is LimeDock’s own analysis — we are not affiliated with langgenius.

https://github.com/langgenius/dify