Dify

by LangGenius

System Card

OrganizationLangGenius

Released2023-04

Architectureagentic-workflow / LLMOps platform with workflow + RAG memory

DetailsFull-stack LLM app platform: workflow editor, RAG pipeline, agent capabilities, model management, observability. Conversation memory + variable memory at node level; Postgres + object storage.

Parameters—

Domainagent-memoryrag-retrieval

Open SourceYes

WebsiteVisit

CodeRepository

llmopsworkflowagentk8s

Capability Profile

Benchmark Scores

6 of 14 benchmarks

Long-Context Retrieval

1/5

RULER

no data

NIAH

no data

LooGLE

no data

LongBench

603p

∞Bench

no data

Multi-Turn Recall

1/2

LoCoMo

78.980p

MemoryBank

no data

Cross-Session Memory

1/1

LongMemEval

76.957p

Multi-Hop QA

2/3

BABILong

no data

MultiHop-RAG

7579p

HotpotQA

7891p

Agent Task Memory

1/1

AgentBench-Mem

7226p

Personalization

0/1

PerLTQA

no data

Factuality / Grounding

0/1

RAGAS

no data

Sources:Dify (langgenius/dify); evaluated on HotpotQA: A Dataset for Diverse, Explainable Multi-hop Question Answering (Stanford / CMU, 1809)Dify (langgenius/dify); evaluated on MultiHop-RAG: Benchmarking Retrieval-Augmented Generation for Multi-Hop Queries (HKUST, 2401)Dify (langgenius/dify); evaluated on AgentBench Memory Track (Tsinghua KEG, 2308)Dify (langgenius/dify); evaluated on LoCoMo: Long-Term Conversational Memory Benchmark (Snap Research, 2402)Dify (langgenius/dify); evaluated on LongBench: A Bilingual, Multitask Benchmark for Long Context Understanding (Tsinghua KEG, 2308)Dify (langgenius/dify); evaluated on LongMemEval: Benchmarking Chat Assistants on Long-Term Interactive Memory (Salesforce AI Research, 2410)