Custom Functions

NOTE

Verification Status: Tested and Verified against rekuiper engine with live telemetry stream load on 2026-10-01 16:47:58 UTC.
Scorecard: Plugin-Dependent Extensions:
Custom functions (echo, countPlusOne, accumulateWordCount, image, geohash, and TensorFlow plugins) require compiling and registering external plugin dynamic libraries (/plugins/functions) or JavaScript UDF functions (/udf/javascript). Without the plugin loaded, the engine validates functions strictly and returns function not found.

rekuiper supports custom user-defined functions. For details on how to configure and install native functions and extensions, refer to Native Function Extensions and WebAssembly Extensions.

echo Plugin

Function	Example	Description
echo	echo(avg)	Outputs the input parameter value without modification.

Example:

Assuming column avg has data type int with value 30, the output is [{"r1":30}]:

SELECT echo(avg) as r1 FROM test;

countPlusOne Plugin

Function	Example	Description
countPlusOne	countPlusOne(avg)	Outputs the array length plus one.

Example:

Assuming column avg has data type []int with value [1, 2, 3], the output is [{"r1":4}]:

SELECT countPlusOne(avg) as r1 FROM test;

accumulateWordCount Plugin

Function	Example	Description
accumulateWordCount	accumulateWordCount(avg, sep)	Counts the number of words separated by sep.

Example:

Assuming avg is a string with value "My name is Bob" and sep is " ", the output is [{"r1":4}]:

SELECT accumulateWordCount(avg, sep) as r1 FROM test;

Image Processing Plugin

Image processing functions support png and jpeg formats.

Function	Example	Description
resize	resize(avg, width, height, [isRaw])	Scales an image to target dimensions. If width or height is 0, the function maintains the aspect ratio. The optional boolean isRaw specifies whether to return raw image bytes rather than encoded formats.
thumbnail	thumbnail(avg, maxWidth, maxHeight)	Scales down an image to maximum dimensions while preserving the aspect ratio.

Example resize(avg, width, height):

Input avg has data type []byte:

SELECT resize(avg, width, height) as r1 FROM test;

Example thumbnail(avg, maxWidth, maxHeight):

Input avg has data type []byte:

SELECT thumbnail(avg, maxWidth, maxHeight) as r1 FROM test;

Geohash Plugin

The Geohash plugin provides geospatial encoding and decoding functions:

Function	Signature	Description
geohashEncode	geohashEncode(la, lo float64)(string)	Encodes latitude and longitude into a geohash string.
geohashEncodeInt	geohashEncodeInt(la, lo float64)(uint64)	Encodes latitude and longitude into an unsigned 64-bit integer geohash.
geohashDecode	geohashDecode(hash string)(la, lo float64)	Decodes a geohash string into latitude and longitude coordinates.
geohashDecodeInt	geohashDecodeInt(hash uint64)(la, lo float64)	Decodes an integer geohash into latitude and longitude coordinates.
geohashBoundingBox	geohashBoundingBox(hash string)(string)	Returns the boundary box coordinates encoded by a geohash string.
geohashBoundingBoxInt	geohashBoundingBoxInt(hash uint64)(string)	Returns the boundary box coordinates encoded by an integer geohash.
geohashNeighbor	geohashNeighbor(hash string, direction string)(string)	Returns the neighbor in the specified cardinal direction (North, NorthEast, East, SouthEast, South, SouthWest, West, NorthWest).
geohashNeighborInt	geohashNeighborInt(hash uint64, direction string)(uint64)	Returns the integer neighbor in the specified cardinal direction.
geohashNeighbors	geohashNeighbors(hash string)([]string)	Returns an array containing all eight neighboring geohash strings.
geohashNeighborsInt	geohashNeighborsInt(hash uint64)([]uint64)	Returns an array containing all eight neighboring integer geohashes.

Geohash Examples

Example geohashEncode:

• Input: {"lo": 131.036192, "la": -25.345457}
• Output: {"geohashEncode": "qgmpvf18h86e"}

SELECT geohashEncode(la, lo) FROM test

Example geohashEncodeInt:

• Input: {"lo": 131.036192, "la": -25.345457}
• Output: {"geohashEncodeInt": 12963433097944239317}

SELECT geohashEncodeInt(la, lo) FROM test

Example geohashDecode:

• Input: {"hash": "qgmpvf18h86e"}
• Output: {"geohashDecode": {"Longitude": 131.036192, "Latitude": -25.345457099999997}}

SELECT geohashDecode(hash) FROM test

Example geohashDecodeInt:

• Input: {"hash": 12963433097944239317}
• Output: {"geohashDecodeInt": {"Longitude": 131.03618861, "Latitude": -25.345456300000002}}

SELECT geohashDecodeInt(hash) FROM test

Example geohashBoundingBox:

• Input: {"hash": "qgmpvf18h86e"}
• Output: {"geohashBoundingBox": {"MinLat": -25.345457140356302, "MaxLat": -25.34545697271824, "MinLng": 131.03619195520878, "MaxLng": 131.0361922904849}}

SELECT geohashBoundingBox(hash) FROM test

Example geohashBoundingBoxInt:

• Input: {"hash": 12963433097944239317}
• Output: {"geohashBoundingBoxInt": {"MinLat": -25.345456302165985, "MaxLat": -25.34545626025647, "MinLng": 131.0361886024475, "MaxLng": 131.03618868626654}}

SELECT geohashBoundingBoxInt(hash) FROM test

Example geohashNeighbor:

• Input: {"hash": "qgmpvf18h86e", "direction": "North"}
• Output: {"geohashNeighbor": "qgmpvf18h86s"}

SELECT geohashNeighbor(hash, direction) FROM test

Example geohashNeighborInt:

• Input: {"hash": 12963433097944239317, "direction": "North"}
• Output: {"geohashNeighborInt": 12963433097944240129}

SELECT geohashNeighborInt(hash, direction) FROM test

Example geohashNeighbors:

• Input: {"hash": "qgmpvf18h86e"}
• Output: {"geohashNeighbors": ["qgmpvf18h86s", "qgmpvf18h86u", "qgmpvf18h86g", "qgmpvf18h86f", "qgmpvf18h86d", "qgmpvf18h866", "qgmpvf18h867", "qgmpvf18h86k"]}

SELECT geohashNeighbors(hash) FROM test

Example geohashNeighborsInt:

• Input: {"hash": 12963433097944239317}
• Output: {"geohashNeighborsInt": [12963433097944240129, 12963433097944240131, 12963433097944240130, 12963433097944237399, 12963433097944237397, 12963433097944150015, 12963433097944152746, 12963433097944152747]}

SELECT geohashNeighborsInt(hash) FROM test

LabelImage Plugin

The LabelImage sample plugin runs TensorFlow Lite image classification models. This plugin is included in Docker images with the -slim suffix.

The function receives a bytea image payload and returns the predicted classification label.

Assuming input payload self contains binary data for peacock.jpg, the function outputs "peacock":

SELECT labelImage(self) FROM tfdemo

tfLite Plugin

The tfLite plugin executes TensorFlow Lite model inference. This plugin is included in Docker images with the -slim suffix.

To execute inference:

1. Upload the .tflite model using the uploads REST API.
2. Specify the model name without the .tflite extension in model_name.
3. Pass input data as a one-dimensional array in input_data.

SELECT tfLite(model_name, input_data) FROM tfdemo

WebAssembly (Wasm) Plugin: wasm_run

wasm_run(module_name, func_name, arg1, arg2, ...)
wasm(module_name, func_name, arg1, arg2, ...)

Executes an exported function from a compiled WebAssembly (.wasm) module registered in the embedded rekuiper Wasm runtime.

Modules are registered via the REST API (POST /plugins/wasm) or the MCP register_wasm_plugin tool. Exported functions can either be invoked directly as UDFs or dispatched dynamically via wasm_run.

Parameters

• module_name: A string identifier matching the registered Wasm plugin module name (e.g. 'math', 'crypto', 'detector').
• func_name: A string identifying the exported function inside the Wasm module to execute.
• arg1, arg2, ...: Zero or more arguments passed directly to the Wasm function.

Return Value

The calculated return value from the WebAssembly function execution. Returns null if the module or function does not exist, argument count/types mismatch, or execution encounters an internal trap.

Examples

Execute an arithmetic addition function inside a Wasm module named math:

SELECT wasm_run('math', 'add', 10, 20) AS total FROM demo;

Result:

{"total": 30}

Invoke an edge anomaly detection algorithm compiled from Rust or C to Wasm:

SELECT wasm('detector', 'evaluate_vibration', vibe, rpm) AS anomaly_score FROM telemetry;