What you can build today.
One query surface across rows, documents, time-ordered events and vectors, available now. Relationships and objects are the next access paths, and are on the roadmap.
Three steps, no setup.
You can look at the query surface before you talk to anyone. The playground runs the same statements you read on this page.
Pick a data model
Rows, documents, events or vectors. Start with the one that removes the most glue, usually rows.
See the modelsRun a request
Five sample statements run against data that ships with the page, right in your browser. Nothing to install.
Open the playgroundRead the reference
The query surface for what is available today, plus what is deliberately not documented yet.
Read the docsWhat you can build on, and what is ahead.
Solid means available today; dashed means in development; dotted means roadmap — useful as direction, not as something to build against.
SQL
Transactional application dataRelational workloads, tables, joins, transactions.
- Records, keys and joins
- Consistent reads and writes
- Users, orders, inventory
JSON
Application documentsDocuments and nested objects, queried alongside rows.
- Nested objects stored beside rows
- Fields that keep changing shape
- One query across both shapes
Key-Value
Direct accessShort-lived state, addressed by key.
- Sessions and short-lived state
- Lookups beside relational rows
- Cache-style access patterns
- One key, one value, one address
Time series
Telemetry and eventsEvents, measurements and history over time.
- Time-ordered events and measures
- History beside current state
- Windowed reads and downsampling
Geo
Spatial dataWhere it is, asked like any other predicate.
- Nearby and radius queries
- Location beside the other models
- Region-aware filtering
- Distance as a queryable value
Vector
Search and retrievalEmbeddings and nearest-neighbour retrieval over stored data.
- Embeddings stored beside the records they describe
- Top-k similarity with ORDER BY distance LIMIT k
- WHERE filtering beside the vector ordering
Graph
Connected dataNodes, edges and traversal.
- Relationship exploration
- Dependency and impact analysis
- Connected-data workflows
- Network and knowledge relationships
Objects
Large objects and assetsLarge objects and file-like assets.
- Files and media assets
- Artifacts and datasets
- Unstructured binary objects
- Assets with queryable metadata
Roadmap Graph and blobs are being explored. There is nothing to run for them yet, and no dates are published.
Four requests, four data shapes.
These are the sample statements the playground ships with. Run them there to see the plan form and the result return.
Mixed workload
5 rows from the sample dataset · rows and document fields read together
Run in the playground-- structured rows and flexible documents in one query
SELECT c.id,
c.region,
c.profile ->> 'plan' AS plan,
COUNT(o.id) AS orders,
MAX(o.created_at) AS last_order
FROM customers c
JOIN orders o ON o.customer_id = c.id
WHERE o.created_at >= NOW() - INTERVAL '90 day'
GROUP BY c.id, c.region, plan
ORDER BY orders DESC
LIMIT 5; SELECT profile
FROM customers
WHERE id = 'C10284'; Time series
6 windows · sample events bucketed by hour
Run in the playgroundSELECT DATE_TRUNC('hour', ts) AS window,
COUNT(*) AS events,
MAX(latency_ms) AS p_max
FROM events
WHERE ts >= NOW() - INTERVAL '6 hour'
GROUP BY window
ORDER BY window; Similarity
3 rows · exact top-k over the sample embeddings
Run in the playgroundSELECT id, label,
embedding <-> [1.0, 0.0, 0.0] AS dist
FROM docs
ORDER BY embedding <-> [1.0, 0.0, 0.0]
LIMIT 3; Parsing, planning, executing.
The console shows these three stages as a request runs. They are the real sequence a request moves through, a plan is chosen before anything is read.
Parsing
The statement is read and resolved against the schema of the data layer.
Planning
A plan is chosen across the data models the request touches.
Executing
Operators run over the stored layout and produce a result set.
Note Stage timings shown anywhere on this site are illustrative of the interaction, not measurements of the product.
How to get your hands on it.
The playground is open. Everything beyond it goes through a short conversation, because what you get depends on what you are running.
- Available models
- SQL, JSON / documents, time series and vectors.
- Surfaces
- SQL statements, document field access, and time-windowed reads inside the same statement.
- Access
- Community builds for macOS, Windows and Linux are published with checksums on the download page; evaluation with the team is on the contact page.
- The playground
- Runs entirely in your browser against sample data that ships with this site. No connection, no telemetry, no account.
Describe the workload
What you store, what you query, and which part of it is currently spread across more than one system.
Scope one data model
We start with the model that removes the most glue, usually rows, sometimes documents, vectors or events.
Run it side by side
Measure against the existing stack. If it does not hold up, you keep what you already have.
GET PLOMID, community builds for macOS, Windows and Linux are published with checksums on the download page; the playground, the documentation and the repository need no install at all.