CloudPuter CloudPuter
Corelamo
Co-Relational Language Model Database

A native XML document database built for the LLM era. Full-text search, document relation traversal, semantic queries - unified in a single engine. Zero compromise on performance and safety.

Read the Docs → Meet the team
Rust & C++
Natively XML
Compleatly customizable
Open-source

Live demo on english Wikipedia corpus

See it in action.

WikiSearch runs on the full english Wikipedia corpus - full-text search, structure, and semantic meaning search. Type in anything and watch the graph surface what's actually related, not just what matches keywords.

cloudputer.org/wikisearch
Full Wikipedia corpus · 7M+ articles indexed
origin of the internet ARPANET
ARPANET 99.2%
The Advanced Research Projects Agency Network was the first wide-area packet-switched network with distributed control — the direct precursor to the modern internet, commissioned by DARPA in 1969.
History of the Internet 96.7%
The internet originated in the United States as ARPANET in the late 1960s, funded by DARPA to enable resilient communication across military and academic research networks.
Packet switching 91.4%
The core technology that made ARPANET possible — routing data in independent packets rather than dedicated circuits, enabling the fault-tolerant distributed network architecture the internet still relies on.
Try the live demo →

Why Corelamo?

The database for unstructured data.

Corelamo lets you search, correlate, link, filter, treverse in large datasets all at once in a couple milisecods. You can set it up on your hardware in minutes for free.

Graph Database
Document Graph Traversal
Declare any XML field as a link between documents and Corelamo lets you treverse it. Traverse concept relationships, follow references across the corpus, and filter by structure
Configurable Index
Advanced Full-Text Search
Configure the index training weights in a single policy file. Full-text indexing happens at write time, and is compleatly customizable based on the application.
Explainable AI
Built-in LLM Capabilities
Semantic similarity and concept-level retrieval, language like answering capabilities with pointers to data

Trainable index?

The policy weight model

Declare the right fields and your document store becomes a traversable concept graph. A single query can follow links between documents, match full-text, and retrieve by semantic similarity.

-- Find articles about quantum computing, -- semantically close to cryptography, -- then traverse to related concepts in the graph SEARCH /corpus/article WHERE FULLTEXT(*, "quantum computing") AND LLM_MATCH( abstract, "applications in modern cryptography", 0.82 ) AND @year > 2020 TRAVERSE @links -> /corpus/article DEPTH 2 WHERE LLM_MATCH(*, "post-quantum encryption", 0.75) RETURN title, abstract, @year, SCORE() AS relevance ORDER BY relevance DESC LIMIT 20
FULLTEXT() — keyword & phrase search
Searches across all fields of the document simultaneously. Supports proximity, phrase matching, and stemming with no separate configuration.
LLM_MATCH() — semantic similarity
Retrieves documents whose meaning is close to the query, not just their words. The similarity threshold is tunable per clause — tighter for seed documents, looser for traversed neighbours.
TRAVERSE — graph traversal across documents
Declare any field (like @links) as a graph edge and TRAVERSE follows it. Specify depth, direction, and per-hop filters. The graph is built from your data — no separate schema or edge table required.
SCORE() — unified relevance ranking
A single score merges full-text relevance, semantic similarity, and graph proximity signals. Sort, filter, or paginate by it — no post-processing in your application.

Capabilities

Everything we needed ourselves

We speciallize in performant systems operating with sensitive data. Every feature in Corelamo exists because we needed it somewhere.

01
At Speed of Thought
Rust and C++ from the ground up. Complex queries take no more than 0.01 seconds.
02
Air-gap architecture
Made for sensitive data safety, an air-gapped Corelamo database with a cactus TPM is unhackable by design. Best choice for medical and defense intelligence.
03
Scale Without Limits
Hierarchical clustering - sharding, mirroring & querying through a master node. Hyperscale ready.
04
LLM Capabilities, Built In
Semantic meaning search, human language like answer capabilities with pointers back to source data. Explainable AI.
05
Live analytics
By design all metrics and query costs are logged and available through a modern web-interface. All errors and transactions are easily tracked.

Use cases

Built for demanding applications.

Anywhere documents need to be stored, searched, and reasoned over — Corelamo replaces multiple systems with one.

Ready to get started?

Read the documentation, explore the architecture, or just drop us a line if you have questions.

Read the Docs Get in touch →