Tech Maven Geospatial
DuckDB for ArcGISEnterprise Custom Data Feeds

P1 capabilities delivered · October 2026

Spatial data, queried
where it lives.

A read-only DuckDB provider for ArcGIS Enterprise Custom Data Feeds. Read spatial files, push supported queries into DuckDB, and connect the results to your ArcGIS layers.

Explore capabilities Read the documentation

Node.js 22 / 24 · ArcGIS Enterprise 12.1 Linux target · MIT license

data/sample.csvREAD ONLY
where
population > 4000000
outFields
id,name,population
mode
pass-through
Illustrative file-to-feature map showing the New York sample pointNew York · ID 1 SCHEMATIC VIEW · EPSG:4326
1 matching featureStable ID · mapped fields · GeoJSON
Illustration using the included CSV fixture. This website is not connected to a live query service.
6delivered P1 capabilities
232passing tests on Node 22 and 24
97.76%supported runtime line coverage

Release status: Locally tested on Windows. Production certification still requires ArcGIS 12.1 Linux staging, container execution and real cloud/operational checks. Review the release evidence →

What is available now

More capable queries.
Clearer data contracts.

The P1 release adds correctness and scale foundations to the hardened provider. Pass-through queries and multiple workers are opt-in; full-fetch and one worker remain the defaults.

01 / QUERY

Safe pass-through queries

Schema-checked filters, object IDs, field selection and deterministic paging. Counts, extents and grouped statistics run in DuckDB with bound values.

where · outFields · count · statistics
Explore the supported query subset →
02 / GEOMETRY

CRS and ArcGIS geometry

Explicit EPSG and XY transformations. Accept ArcGIS points, envelopes, paths and rings, with topology checks and WGS84 feature output.

4326 · 3857 · configured source EPSG
Review projection and geometry limits →
03 / SCHEMA

Versioned schema contracts

Approve strict or additive native-field contracts. Detect drift, expose CDF field metadata, and preserve large integer and decimal attributes as strings.

approved versions · fingerprints · UTC dates
Understand field compatibility →
04 / FRESHNESS

Source-aware caching

Use TTL, local file identity, immutable snapshot versions or strong HTTPS ETags. Check observed identity before and after reads and cache hits.

ttl · stat · version · public HTTPS ETag
Choose a freshness policy →
05 / SCALE

Measured worker pooling

Opt into 1–8 exclusive native workers, sharing a total memory and thread budget. Separate scratch paths and interruption keep worker ownership isolated.

single-worker default · benchmark before scaling
Read the fixture measurements →
06 / OPERATIONS

Monitoring assets

Queue, latency, cache, RSS, sampled native spill and source-age metrics. Includes an authenticated ServiceMonitor, configurable alerts and a Grafana dashboard.

Prometheus Operator · Grafana · bearer auth
Set up monitoring →

Files and cloud access

Know what each source supports.

The provider is read-only. Source geometry and globally unique positive 32-bit feature IDs must satisfy the configured contract.

Current source support and qualification status
SourceCurrent scope
CSV / JSON recordsLocal reads with an explicit geometry column
GeoParquetLocal native geometry or explicit WKB
GeoJSONLocal GDAL reader tested; qualify other formats
HTTPS / S3 GeoParquetReaders wired; production endpoints need qualification
Iceberg / Delta / streaming / graphRoadmap and research scope
OPT-IN · ACTIVATION GATED

CloudFS integration

The read adapter supports provider credential references and CSV, JSON records and GeoParquet readers. Native loopback agent fixtures pass using an isolated compatible test engine.

SharePoint · spfs://OneDrive · odfs://Google Drive · gdfs://Dropbox · dbxfs://VPS agent · vfs://
Read compatibility and setup requirements →

Production needs a reviewed, signed build for the pinned DuckDB 1.5.6 engine and real-provider tests. SFTP and remote TLS-agent routing are blocked for the reviewed revision. Delta, Iceberg and COPY TO are not exposed by this app.

From source to layer

One provider. Explicit boundaries.

  1. 01 / QUALIFY THE SOURCE

    Files and snapshots

    Declare geometry, CRS, IDs, schema and freshness. Preflight checks the data before it enters a service.

  2. 02 / QUERY WITH DUCKDB

    Bounded native reads

    Full-fetch or the supported pass-through subset, with query deadlines, resource budgets and bounded result caches.

  3. 03 / CONNECT TO ARCGIS

    Custom Data Feed

    Use the matching Enterprise SDK scaffold to package and register the provider, then certify the service in staging.

The standalone diagnostic API returns GeoJSON and CDF metadata. It is separate from ArcGIS FeatureServer routes and is not an ArcGIS REST replacement.

Try the included sample

Start locally.
Qualify your deployment.

  1. Install and checkUse Node.js 22.14+ or 24. The build runs lint, coverage tests and configuration validation.
  2. Enable the sample contractThe pass-through example uses three CSV point features, an approved schema and local file freshness.
  3. Follow the ArcGIS procedureBuild a genuine CDF package with the matching SDK. Complete Linux staging and client checks before release.
ArcGIS 12.1 Linux certification guide →
LOCAL QUICK START / bash
# From a cloned repository
npm ci
npm run build

# Check the approved sample source
node scripts/cdf-cli.js preflight \
  -f config/examples/pass-through.json -d default

# Start the diagnostic API on loopback
CDF_CONFIG=config/examples/pass-through.json npm start

# In another terminal
curl --get http://127.0.0.1:3000/api/sources/default/query \
  --data-urlencode 'where=population > 4000000' \
  --data-urlencode 'outFields=id,name,population'

First use needs access to the official DuckDB spatial extension repository. Configure a bearer token for production or non-loopback access; terminate TLS at a controlled proxy.

Delivery and next steps

Delivered capabilities.
Visible release gates.

Research modules are excluded from the supported distribution. The roadmap distinguishes implemented foundations from certification work and future features.

P1 DELIVERED

Correctness and scale

  • Safe queries and deterministic pages
  • CRS, geometry and schema contracts
  • Source identities, worker pooling and monitoring
P0 QUALIFICATION

Production release

  • ArcGIS 12.1 Linux package and staging
  • Container, cloud permission and recovery checks
  • Representative load, freshness and live monitoring
P2 / P3 ROADMAP

Connectors and workflows

  • Qualify cloud accounts and fix TLS/SFTP blockers
  • Lakehouse, broader formats and write workflows
  • Catalogs, advanced services and reproducible releases

Read the full roadmap and acceptance criteria →

Build with the current guidance

Documentation that matches the code.

Documentation hub

Configuration, query contracts, source setup and migration guidance in one place.

Browse the guides →

Operations and evidence

Preflight, capacity tests, freshness assumptions, monitoring and recovery.

Review the runbook →

Source and contribution

Repository access requires GitLab sign-in. Public guides describe the checks and enhancement roadmap.

Open GitLab →