Home / Products / ThinKM
Suite-wide KB

ThinKMA unified knowledge base for the whole suite

Store once, ask anywhere. Built on Dify, ThinKM parses documents, augments retrieval, builds graphs and isolates permissions — linked with ThinSpider collection and the ThinkWork bench, pointing models to ThinkMaaS, so the whole suite shares one knowledge.

Get a demo
thinkm.thinkalike.com.cn / KB console
4sources
1.2kchunks
3linked
Capabilities

One knowledge base serving the whole suite

Parse, retrieve, graph and isolate — make org knowledge askable, trustworthy and controlled

Document ingest

Structure-aware parsing, chunking and vectorization across formats to build a high-quality corpus.

Hybrid retrieval

Keyword + vector + rerank hybrid retrieval balances precision and semantics with traceable citations.

Knowledge graph

Entities and relations grow a graph for associative reasoning and structured QA.

Multi-way recall

Multi-strategy recall and fusion ranking improve hits on long-tail, complex questions.

Permission isolation

Single-tenant form and fine-grained permissions keep knowledge boundaries clear and access controlled.

SaaS + on-prem

Both self-operated SaaS subscription and per-instance SN-licensed on-prem for air-gapped use.

Retrieval

Askable, trustworthy, traceable knowledge

A Dify-based RAG pipeline: hybrid retrieval and multi-way recall raise hit rates while every answer cites its source — no fabrication.

  • Hybrid: keyword, vector and rerank combined for precision and semantics.
  • Traceable: each answer points to its source for verification.
  • Graph QA: structured relations power complex associative questions.
Retrieval pipeline
hybrid
recall
Suite link

One knowledge shared across the suite

As the unified knowledge layer, ThinKM links ThinSpider collection, the ThinkWork bench and ThinkDE employees, pointing models to ThinkMaaS — maintain once, effective everywhere.

  • Collect: ThinSpider-crawled content flows straight into the KB.
  • Consume: ThinkWork retrieves this KB via its sidecar.
  • Supply: RAG is fed by unified ThinkMaaS models.
ThinKM · suite link
ThinSpider
ThinkWork
Hybrid
keyword + vector search
Multi
recall with rerank
3
linked products
2
delivery forms
Workflow

From documents to askable knowledge

Store, augment, link, control

1

Parse

Structure-aware parsing, chunking and vectorization.

2

Retrieve

Hybrid retrieval + multi-way recall + rerank with citations.

3

Graph

Distill entities and relations into a reasoning graph.

4

Link

Share one knowledge across collection, bench and employees.

Use cases

Enterprise knowledge hubPolicy & document QAProduct libraryResearch KBOn-prem knowledge layer
Preview

Interface at a glance

ThinKM — console
corpusretrieval
KB consoleDocs, chunks and retrieval hits in one screen with live health.
ThinKM — QA
citationsgraph
QA & citationsConversational QA with sources and a visualized knowledge graph.

Store knowledge once, ask anywhere

Try ThinKM's parsing, retrieval and suite-linking capabilities.

Try ThinKM now
FAQ

Frequently asked about ThinKM

What is ThinKM?

ThinKM is a unified knowledge base by ThinkAlike built on Dify, offering parsing, hybrid retrieval, knowledge graphs and permission isolation for the whole suite.

SaaS or on-prem?

Both: single-tenant SaaS subscription and per-instance SN-licensed on-prem for air-gapped deployment.

How does it link with other products?

It is the knowledge layer linked with ThinSpider, ThinkWork and ThinkDE employees, pointing models to ThinkMaaS.

How do I try it?

Visit https://thinkm.thinkalike.com.cn for a trial, or contact us for on-premises deployment.