Home > Reference > ModelOp Center CLI > The ModelOp Center CLI User Guide

ModelOp Center CLI

ModelOp Center exposes a CLI that is built using Go, allowing end users to interface with ModelOp Center micro services via CLI. The Go CLI is the primary supported ModelOp Center CLI.

For 1.x versions of ModelOp Center, there is also a Python CLI (Python CLI reference). However, the Python CLI will be deprecated, starting in ModelOp Center version 2.x.

Installation Instructions

  1. Download the CLI here:

  2. For Mac (Darwin) and Linux make it executable:

     chmod +x fastscore
  3. Add to $Path

    For Macs: /usr/local/bin/fastscore

  4. Confirm proper installation with fastscore

ModelOp Center CLI User Guide

This guide describes the complete set of commands and options supported by the ModelOp Center CLI.

Command index

   fastscore - Control fastscore from the command line.

   fastscore [global options] command [command options] [arguments...]


     connect     Establish a fastscore Connection
     use         Select the target engine instance
     run         A shortcut for running a particular model and stream configuration
     config      Configure the fastscore fleet
     fleet       Examine status of the fastscore fleet
     model       Manage analytic models
     schema      Manage schemata
     stream      Manage streams
     sensor      Manage sensors
     engine      Manage engines
     attachment  Manage attachments
     pneumo      Print pneumo messages
     create      Create new fastscore services
     destroy     Destroy existing fastscore services
     conductor   Interact with the orchestrator
     composer    Create and manage workflows
     workflow    Create and manage workflows
     monitor     Monitor an engine
     login       Login to fastscore
     debug       Watch debugging messages
     stats       Get various statistics
     help, h     Shows a list of commands or help for one command

   --help, -h     show help
   --version, -v  print the version

WEBSITE: http://www.opendatagroup.com
SUPPORT: support@opendatagroup.com

Connecting to ModelOp Center

fastscore connect [command options] <prefix>
   --verbose, -V  Be verbose
   --nowait       Do not retry connecting.

The command establishes a connection between the command line interface and the ModelOp Center deployment. The <prefix> is the URL of the ModelOp Center proxy, e.g.

By default, the command waits for the provided prefix to become available. Use --nowait option to disable the check and return immediately.


$ fastscore connect https://localhost:8000 -V
Connecting... Done.

Choosing an instance

fastscore use <instance-name>

The command selects <instance-name> as a target of subsequent commands. In addition, it makes <instance-name> the preferred instance of a specific service. For example, fastscore use engine-1 makes engine-1 the target of sensor management commands. Any command that require an instance of an engine service will now use the engine-1 instance.


$ fastscore use engine-1

Running models

fastscore run <model> <input stream> <output stream>

The run is a convience command mostly equivalent to the following sequence of commands:

fastscore model load <model-name>
fastscore stream attach <stream-0> 0
fastscore stream attach <stream-1> 1

Models with a single input and single output stream are widespread. The command allows building such pipelines quickly.


$ fastscore model add cube
action <- function(x) emit(x*x*x)
$ fastscore run cube rest: rest:
$ fastscore model interact
> 2
> 10

Model input/output using REST

fastscore model input [ <slot> ]
fastscore model output [ <slot> ] [ -nowait ] [ -c ]
fastscore model interact

The model input command reads its standard input and sends the data to the REST stream attached to <slot> (or slot 0 if <slot> is omitted). Each line is interpreted as a separate data record.

The model output reads data from the REST stream attached to <slot> (or slot 1 if <slot> is omitted). By default, the command waits for the record to become available and retrieves exactly one record. The -nowait option limits the waiting time to 0.5s. The -c option asks the command to continuously read data from the stream.


$ fastscore run sqrt rest: rest:
$ fastscore model input
$ fastscore model output -c

The model interact command combines the effects of model input and model output commands. It allows the user to both send inputs to the REST stream and observe the outputs. The model interact also shows records rejected either by encoding or the schema.


$ fastscore run sqrt rest: rest:
$ fastscore model interact
> 7
> foobar
> 11
> 100
> ^D

Note that these commands do not close input/output streams. They may be run multiple times.

Configuring ModelOp Center

fastscore config set <config-file>
fastscore config show

The first command sets the configuration of the ModelOp Center deployment. The second - prints the current configuration. The <config-file> is the name of the configuration file. The configuration file uses YAML format. Its syntax is described here.


$ fastscore config set config.yaml
$ fastscore config show
    host: database
    password: root
    port: 3306
    type: mysql
    username: root
  - api: engine
    host: engine-2
    name: engine-2
    port: 8003
  - api: engine
    host: engine-1
    name: engine-1
    port: 8003
  - api: model-manage
    host: model-manage
    name: model-manage-1
    port: 8002
    type: REST

Listing services

fastscore fleet [command options] 
   --verbose, -V  be verbose
   --json         as json
   --wait         Wait for all sewrvices to report healthy

The command outputs the information about the currently running ModelOp Center services. The verbose output includes information about CLI and Connect.

Use -wait option to make the command wait for all services to report a healthy status.


$ fastscore fleet -wait -V
       NAME      |     API      |     HOST     | PORT | RELEASE |           BUILTON            | HEALTH
  engine-2       | engine       | engine-2     | 8003 |     1.9 | Thu Jan 10 18:48:14 UTC 2019 | ok
  engine-1       | engine       | engine-1     | 8003 |     1.9 | Thu Jan 10 18:48:14 UTC 2019 | ok
  model-manage-1 | model-manage | model-manage | 8002 |     1.9 | Thu Jan 10 17:14:49 UTC 2019 | ok

Managing models

fastscore model add <model-name> <source-file> [ --type <model-type> ]
fastscore model show <model-name> [ -e ]
fastscore model list
fastscore model remove <model-name>

The model add command introduces a model to fastscore. The type of the model can be set using --type <model-type> option:

Option Model type
-type python2 Python 2.x
-type python3 Python 3.x
-type R or -type:r R
-type java Java
-type c C
-type pfa-json PFA (JSON)
-type pfa-yaml PFA (YAML)
-type pfa-pretty PFA (PrettyPFA)
-type h2o-java h20.ai (Java)
-type octave Matlab using Octave
-type sas SAS (experimental)

If the --type option is omitted the model type is determined by the extension of the <source-file>:

Extension Model type
.py Python 2.x
.py3 Python 3.x
.R R
.c C
.pfa or .json PFA (JSON)
.yaml PFA (YAML)
.ppfa PFA (PrettyPFA)
.m Matlab using Octave

If the commands reads the model source from the standard input it still tries to gets the model type. For example, if the source code starts with def action( the command assumes that the model type is Python.

The model show prints the model source code to the standard output. The -e (edit) option provides for a convenient modification of the model. The edit option spawns an editor and updates the model when the editor closes. The editor name is taken from the EDITOR environment variable. By default, the editor name is ‘vi’.

The model list command shows the list of models known to ModelOp Center. The model remove command removes the corresponding model.


$ ~/fastscore model add cube ./cube.py3 --type python3 
$ fastscore model list
  cube    | Python3 |
$ fastscore model show cube
def action(x):
  return x*x*x
$ fastscore model remove cube
$ fastscore model list

Loading/unloading models

fastscore model load <model name> [override attachments]
fastscore model unload [engine name]
fastscore model verify <model-name>
fastscore mode inspect

The model load command performs all preparatory steps and gets ready to start data processing. The target engine can be selected using the use command. The preparatory steps include verification of the model syntax, checking if the model follows the ModelOp Center model conventions. If an unexpected slot has a stream attached the model load command will fail. If all requied streams are attached, the loaded model will start the data processing. See also engine pause command.

Note that the model load command can be called repeatedly, even if the data processing is underway. The command will attempt to replace the model without stopping the flow of data. The replacement model may be implemented in a different language.

The model unload command removes the active model from the engine. The operation is only allowed if the engine is in the INIT state. When data is being processed it is not possible to unload the model. See also engine reset command.

The model verify command performs all preparatory steps without actually loading the model. The verbose command output shows how the model smart comments were understood.


$ fastscore model verify echo -V
Model Name     Model Type     SLOC
----------     ----------     ----
echo           Python3        5

Slot     Schema     Action     Recordsets
----     ------     ------     ----------
0        string     action     false
1        string                false

The model contains no errors

The model inspect prints information about the currently loaded model. Its output is similar to the output of the model verify command.


$ fastscore model inspect
| echo | python3 | none      | none        |

|    0 | action | false      |
|    1 |        | false      |

| JET # | PID |  SANDBOX  |
|     1 |  45 | 100796087 |

Managing schemas

fastscore schema add <schema-name> <schema-file>
fastscore schema show <schema-name> [ -e ]
fastscore schema list
fastscore schema remove <schema-name>

The schema add command adds an Avro schema to ModelOp Center.

The schema show prints the schema to the standard output. The -e (edit) option allows editing the schema in-place. The edit option spawns an editor and updates the schema after the editor closes. The editor name is taken from the EDITOR environment variable. By default, the editor name is ‘vi’.

The schema list command shows the list of schemas known to ModelOp Center. The schema remove command removes the corresponding schema.


$ fastscore schema add sch-1 ./sch-1.avsc
$ fastscore schema list
$ fastcore schema remove sch-1 -v

Schema Verify

fastscore schema verify <schema name> <data source>
   --verbose, -V  be verbose

The schema verify command check that the named schema is a well-formed. If <data source> is present the command in addition checks data in the file for conformance against the schema. The data file contains JSON-encoded records separated by newlines.


$ fastscore schema add s1
$ fastscore schema verify s1
Error: Type "foobar" is undefined
$ fastscore schema add s2
$ fastscore schema verify s2 -v
Schema OK
$ cat <<EOF>>data1
$ fastscore schema verify s2 data1
OK   2

FAIL "fastscore"
     Int expected, not "fastscore"

     Int expected, not []

OK   3

Managing streams

fastscore stream add <stream-name> <descriptor-file>
fastscore stream show <stream-name> [ -e ]
fastscore stream list
fastscore stream remove <stream-name>

The stream add command adds a stream descriptor to fastscore. The syntax of the stream descriptor is explained here.

The stream show prints the stream descriptor to the standard output. The -e (edit) option allows editing the descriptor in-place. The edit option spawns an editor and updates the stream descritor when the editor closes. The editor name is taken from the EDITOR environment variable. By default, the editor name is ‘vi’.

The stream list command shows the list of streams known to fastscore. The stream remove command removes the corresponding stream.


$ fastscore stream add input-1 ./input-stream.json
$ fastscore stream list
$ fastcore stream remove input-1

Attaching/detaching streams

fastscore stream attach <stream-name> <slot>
fastscore stream detach <slot>
fastscore stream verify <stream-name> <slot>
fatsscore stream inspect

The stream attach opens the stream and attaches it to the <slot>. Typically, the command retrieves the stream descriptor from Model Manage using <stream-name>.

The stream detach command closes and detaches the stream. It is not possible to detach the stream when the model is running. However, it is possible to attach a different stream to the same slot using the stream attach command while data is being processed.

The stream verify command performs all the same steps as stream attach without actually attaching the stream to the <slot>. It prints the internal representation of the stream descriptor with all default values substituted.


$ fastscore stream verify s1 0
  "batching": {
    "nagle_time": 500,
    "watermark": 1000
  "linger_time": 3000,
  "loop": false,
  "transport": {
    "type": "rest",
    "mode": "simple"
  "encoding": null
The stream descriptor contains no errors

The stream inspect command shows the list of streams attached to slots at the current engine.


$ fastscore stream inspect
  SLOT |       NAME   | TRANSPORT |  EOF
     0  inline-526507    REST       False
     1  inline-809896    REST       False

URL-like stream descriptors

An URL-like or literal stream descriptor is a shortened representation of a stream descriptor. It can be used instead of the stream name for commands, such as stream attach. The commands look for a ‘:’ character to distinguish a stream name from a literal stream descriptor. The table below contains examples of literal stream descriptors.

Example Description
inline:1,2,3 An inline stream 3 records
inline-chunked:%06%02%04%06 An inline stream with Avro-encoded [1,2,3] array
rest: A REST stream (simple mode)
rest-chunked: A REST stream (chunked mode)
kafka:kafka-1.company.com:9092/mytopic.json A Kafka stream
file:/vol/data/data1.dat A (local) file stream
tcp: A TCP stream
udp::9099 A UDP stream (input only)
https://dev.company.com:1234/data1.json An HTTP stream
exec:/root/myscript.sh An exec stream
discard: A discard stream

Use stream verify command to see how the literal stream descriptor is understood by the engine.


$ fastscore stream verify inline:1,2,3 0
  "encoding": "json",
  "linger_time": 3000,
  "transport": {
    "data": [
    "type": "inline"
  "batching": {
    "nagle_time": 500,
    "watermark": 1000
  "loop": false,
  "schema": "undefined"
The stream descriptor contains no errors

Sampling streams

fastscore stream sample <stream-name> [ sample count ]

The stream sample reads a few data records from the stream. This happens in the context of the engine. Otherwise, stream sampling is unrelated to the data pipeline. If sample count option is omitted, the commands attempts to read 10 records max.

$ fastscore stream sample inline:2,3,5,7,11 3
   1: 2
   2: 3
   3: 5

Managing sensors

fastscore sensor add <sensor-name> <sensor-file>
fastscore sensor show <sensor-name> [ -e ]
fastscore sensor list
fastscore sensor remove <sensor-name>

The sensor add command adds a sensor descriptor to ModelOp Center. The syntax of a sensor descriptor is described here.

The sensor show prints the sensor descriptor to the standard output. The -e (edit) option allows editing the sensor descriptor. The edit option spawns an editor and updates the descriptor after the editor closes. The editor name is taken from the EDITOR environment variable. By default, the editor name is ‘vi’.

The sensor list command shows the list of sensor descriptors known to ModelOp Center. The sensor remove command removes the corresponding sensor descriptor.

Note that the above commands operate on sensor descriptors, not active sensors. See sensor install/uninstall/inspect commands for more.


$ fastscore sensor add watchdog-1 ./watchdog.json
$ fastscore sensor list
$ fastcore sensor remove watchdog-1

Active sensor operations

fastscore sensor install <sensor-name>
fastscore sensor uninstall <tap-id>
fastscore sensor inspect [ <tap-id> ]
fastscore sensor points

The sensor install command adds the sensor to the current service instance. See use command. Upon success, the command prints <tap-id> of installed sensor. Other active sensor operations take <tap-id> as an argument.

The sensor inspect prints information about all active sensors for the current service instance or a particular active sensor if <tap-id> is present.

The sensor uninstall command removes the active sensor identified by <tap-id>.

The sensor points command outputs the list of tapping points provided by the current instance. A sensor descriptor includes a tapping point as a value of the “Tap” element.


$ fastscore sensor install sensor1
$ fastscore sensor inspect 15 -v
Sensor id 15 is attached to manifold.debug at engine-1.
$ fastscore sensor uninstall 15
$ fastscore sensor points

Managing engine state

fastscore engine reset [ <engine-name> ] [ --all ]
fastscore engine inspect [ <engine-name> ]
fastscore engine pause [ <engine-name> ]
fastscore engine unpause [ <engine-name> ]

A ModelOp Center engine has one of the following seven states:

State Description
INIT Initialization - data pipleline is incomplete
RUNNING Data is being processed
PIGGING Pigging is in effect
FINISHING All inputs at EOF - processing continues
FINISHED Both inputs and outputs at EOF
PAUSED Data processing is paused
ERROR Unrecoverable error encountered - see logs

The engine reset command puts the engine into the INIT state regardless of its current state. The commands unload the model and detaches all streams. Use engine reset to start using an engine again after it reached the FIHISHED state.

The engine inspect prints the current state of the engine.


$ fastscore engine inspect
engine engine-1 is init
$ fastscore engine reset

The engine pause puts the engine into the PAUSED state. It is not possible to pause an engine which is currently in FINISHING, FINISHED, or ERROR state. If the pause is requested during initialization, the engine becomes paused when it is about to enter the RUNNING state. Use engine unpause to continue data processing.


$ fastscore engine inspect
engine engine-1 is init
$ fastscore engine pause
$ fastscore engine inspect
engine engine-1 is init
$ fastscore run cube rest: rest:
$ fastscore engine inspect
engine engine-1 is paused
$ fastscore engine unpause
$ fastscore engine inspect
engine engine-1 is running

Model attachments

fastscore attachment upload <model-name> <file-to-upload>
fastscore attachment download <model-name> <attachment-name>
fastscore attachment list <model-name>
fastscore attachment remove <model-name> <attachment-name>

ModelOp Center models may have attachments - archives containing arbitrary data made available to the model when it runs. The attachment upload command adds the attachment to the model. The <file-to-upload> must have one of the following extensions:

Extension Archive format
.zip zip
.tgz gzipped tar
.tar.gz gzipped tar

Upon upload the <file-to-upload> becomes the name of the model attachment. The attachment can be downloaded using the attachment download command. Use attachment remove command to remove the attachment.

Note that there is no hard limit on the size of the attachment.


$ fastscore upload cube data1.zip
$ fastscore attachment list cube
     NAME   | TYPE | SIZE
  data1.zip | zip  |  1048904
$ fastscore attachment remove cube data1.zip

Pneumo access

fastscore pneumo [ history ]

The pneumo commands provide an access to Pneumo – the internal message bus of fastscore. pneumo history prints a few most recent Pneumo messages (60 max). The pneumo command continuously prints Pneumo messages as they appear on the bus.

Monitoring engine operations

fastscore monitor

The monitor command shows a live text-based UI that shows information about the state of the engine, loaded model, attached streams, and running jets.

Example (assuming the engine is running):

$ fastscore monitor
Engine: engine-1 [RUNNING]
Model:  cube (r)

input          I:0          4,649 rps     0.0 mbps
output         O:1          5,496 rps     0.0 mbps

Jet                             Input       Output
cube                        1,833 rps    1,885 rps
cube                        1,825 rps    1,935 rps
cube                        1,845 rps    1,944 rps

User authentication

fastscore login <authentication mode> [ <username> ] [ <password> ]
   Use any of the login method: basic, oauth2, ldap, session cookie. (Case insensitive)

   --verbose, -V  Be verbose

Troubleshooting data pipelines

fastscore debug manifold
fastscore debug stream [ <slot> ]

The debug commands enable generation of a debug messages that describe internal operations of the engine. These commands are not related to debugging of the model source code. The commands are for advanced users only.


# In a separate window run: fastscore run cube inline:1,2,3 discard
$ fastscore debug manifold
14:40:22.491: model loaded
14:40:22.528: schema match: ok
14:40:22.536: stream open: ok
14:40:22.568: schema match: ok
14:40:22.575: stream open: ok
14:40:22.576: requesting data from slot 0
14:40:22.576: starting initial jet 1
14:40:22.882: state changes to RUNNING
14:40:22.885: jet <0.1474.0> is ready
14:40:22.885: 3 record(s) received from slot 0
14:40:22.885: dispatched to jet <0.1474.0>
14:40:22.885: more data requested from slot 0
14:40:22.885: stream at slot 0 reaches EOF
14:40:22.886: pending state change: FINISHING
14:40:22.892: writing output 1 record(s) to slot 1
14:40:22.893: write confirmed for slot 1
14:40:22.893: writing output 1 record(s) to slot 1
14:40:22.893: write confirmed for slot 1
14:40:22.893: writing output 1 record(s) to slot 1
14:40:22.893: write confirmed for slot 1
14:40:22.893: jet <0.1474.0> is ready
14:40:22.893: state changes to FINISHING
14:40:22.893: sending EOF to jet <0.1474.0>
14:40:22.894: jet <0.1474.0> finished
14:40:22.894: state changes to FINISHED

Collecting statistics

fastscore stats memory
fastscore stats cpu-utilization
fastscore stats jets
fastscore stats streams

The stats commands continuously report various statistics. The stats memory and stats cpu-utilization commands show the current memory footprint and CPU utilization of the engine. stats jets shows the records per second throughput of all runnings jets. stats streams reports the throughput at each stream slot.

Example (assuming the engine is running):

$ fastscore use engine-1
$ fastscore stats memory
engine-1: 2163.5mb
engine-1: 2163.5mb
engine-1: 2161.5mb
$ fastscore stats cpu-utilization
engine-1: kernel: 152.1 user: 69.0
engine-1: kernel: 152.7 user: 69.2
engine-1: kernel: 153.0 user: 69.5
$ fastscore stats jets
engine-1: 1000.0 rps/--- 
engine-1: 1745.2 rps/1696.0 rps 
engine-1: 1766.8 rps/1739.2 rps 
engine-1: ---/---
engine-1: 1823.2 rps/1832.0 rps 
engine-1: 1556.4 rps/1626.4 rps 
$ fastscore stats streams
engine-1: 0:25.0 rps | 1:1.0 rps
engine-1: 0:6000.0 rps | 1:5675.0 rps
engine-1: 0:6000.0 rps | 1:5711.0 rps
engine-1: 0:6000.0 rps | 1:5574.0 rps

The stats commands rely on sensor to collect statistics. --- indicate that data is not available at the moment.