A next-generation AI application platform that can be built even on existing systems centered around Postgres.
Many AI applications adopt RAG (Retrieval-Augmented Generation) or Graph RAG,
- Vector DB
- Graph DB
- Dedicated search platform
- Synchronized pipeline
This requires new infrastructure elements, among others.
However , ConceptMiner Engine offers an alternative.
An AI engine that uses a conceptual network model based on GNG + MST at its core,and can be implemented on existing database infrastructures such as traditional PostgreSQL.
In other words, it means that sophisticated semantic applications can be built without necessarily requiring a complex, dedicated database configuration .
The advantage is that a dedicated database is not required.
It can leverage existing technology stacks.
Many companies and development teams already have the following environments:
- PostgreSQL
- MySQL
- Microsoft SQL Server
- Internal authentication infrastructure
- Access control system
- Business application APIs
ConceptMiner Engine can be designed to be easy to implement while leveraging these existing environments.
The system configuration becomes simpler.
Typical AI search systems tend to have the following configuration:
App Server
├─ RDB
├─ Vector DB
├─ Graph DB
├─ Search Engine
└─ Sync Jobs
In ConceptMiner Engine, the core processing of the conceptual network is handled by the engine itself, making it easier to store the data in a regular relational database (RDB).
App Server
├─ PostgreSQL
└─ ConceptMiner Engine
As a result:
- Few components
- Easy to use
- The number of points of failure decreases.
- Easy to review for security reasons
- Fast implementation speed
Examples of data that ConceptMiner can handle in relational databases.
It can be managed as a regular table.
- Document / Text Chunk
- Embedded ID
- Node information
- Edge information
- cluster
- label
- coordinate
- User Permissions
- Audit Log
- Linking with business data
In other words, it’s easy to integrate with existing business systems.
Differences between RAG and Graph RAG
| project | RAG | Graph RAG | ConceptMiner Engine |
|---|---|---|---|
| Proximity search | strong | strong | strong |
| Explicit relationship exploration | weak | strong | Medium to strong |
| Automatic discovery of conceptual structure | weak | middle | strong |
| Dedicated DB dependent | middle | high | Low |
| Integration of existing RDBs | middle | Low | strong |
| Exploratory UX | weak | middle | strong |
Practical benefits for developers
1. Low learning cost
You won’t need to learn a lot of new database products, query languages, and operational methods.
This configuration makes it easy to leverage your existing SQL knowledge.
2. Easy to transition from PoC to production.
There are many cases where the proof-of-concept (PoC) is highly sophisticated, but the actual production environment becomes much more complex.
ConceptMiner is designed with integration into business systems in mind from the outset.
3. Easy to implement on-premises
In closed environments, internal company environments, and regulated industries, adding infrastructure can sometimes be difficult.
ConceptMiner is well-suited for enterprise deployments because it is easy to design around existing databases.
4. Easy to differentiate
Many AI products
Chat + Search + RAG
They are becoming quite similar.
ConceptMiner is
Chat + Structure + Discovery
This allows you to create a different experience.
However, Vector DB / Graph DB can also be used.
ConceptMiner does not negate them.
We can collaborate if necessary.
- Connect to an existing Vector DB
- Export to Graph DB
- Conceptual cluster selection in the pre-RAG stage
- Used as a complementary layer in Graph RAG
What’s important is,
It can be started without requiring a dedicated database.
That’s the point.
Message for Developers
Handling the semantic structure of AI without increasing the number of dedicated databases.
ConceptMiner Engine is a developer-friendly AI engine that does not require Vector DB or Graph DB and is easy to integrate into existing PostgreSQL-centric architectures.
In short
RAG searches for answers.Graph RAG traces relationships.ConceptMiner discovers structures that are not yet visible.
And it’s easy to implement on top of existing systems.
Expected implementation locations
- SaaS product
- Internal Knowledge Search
- VOC analysis
- Consulting support tools
- Research support system
- Market analysis app
- New business exploration tool
Development Examples
As an example of an application developed using ConceptMiner, we are offering the thought support system ThinkNavi for free. Please give it a try.
