hops — Hopsworks CLI

A Go CLI for the Hopsworks Feature Store. Works as a standalone tool and as an AI tool interface for LLMs.

Philosophy

Build a great CLI, and both humans and LLMs benefit. hops is the simplest way to give an AI agent access to Hopsworks: just allow Bash(hops *) and it works. No protocol, no server, no overhead.

Install

# From source
go build -o hops . && sudo mv hops /usr/local/bin/

# Or via go install
go install github.com/MagicLex/hopsworks-cli@latest && mv ~/go/bin/hopsworks-cli ~/go/bin/hops

Quick Start

# Login (prompted for host + API key)
hops login

# Or if inside Hopsworks terminal, it auto-configures
hops project list
hops project use myproject

# Explore feature groups
hops fg list
hops fg info customer_transactions
hops fg preview customer_transactions --n 5
hops fg features customer_transactions

# Keywords (visual tags for feature groups)
hops fg keywords customer_transactions
hops fg add-keyword customer_transactions ml production
hops fg remove-keyword customer_transactions production

# Feature views (single FG or multi-FG joins + transforms)
hops fv list
hops fv create my_view --feature-group transactions
hops fv create enriched --feature-group transactions \
  --join "products LEFT product_id=id p_" \
  --transform "standard_scaler:amount"

# Online feature vector lookup
hops fv get my_view --entry "id=42"

# Batch read from feature view
hops fv read my_view --n 100
hops fv read my_view --output data.parquet

# Insert data
hops fg insert customer_transactions --file data.csv
hops fg insert customer_transactions --generate 100

# Derive new FG from joins (with provenance tracking)
hops fg derive enriched --base transactions \
  --join "products LEFT id" --primary-key id

# Embeddings + similarity search
hops fg create documents --primary-key doc_id \
  --features "doc_id:bigint,title:string" \
  --embedding "text_embedding:384:cosine"
hops fg search documents --vector "0.1,0.2,..." --k 5

# Training datasets (materialize + retrieve)
hops td compute my_view 1
hops td compute my_view 1 --split "train:0.8,test:0.2"
hops td compute my_view 1 --filter "price > 100"
hops td compute my_view 1 --start-time "2026-01-01" --end-time "2026-02-01"
hops td read my_view 1 --td-version 1 --output train.parquet

# Transformations
hops transformation list
hops transformation create --file my_scaler.py

# Browse files
hops dataset list

# Model Registry
hops model list
hops model info fraud_detector --version 1
hops model register fraud_detector ./model_dir --framework sklearn --metrics "accuracy=0.95"
hops model register fraud_detector ./model_dir \
  --feature-view my_view --td-version 1 \
  --input-example sample.json \
  --schema "in:age:int,salary:float out:prediction:float" \
  --program train.py
hops model download fraud_detector --output ./local_dir

# Deployments (serving)
hops deployment list
hops deployment create fraud_detector --version 1 --script predictor.py
hops deployment start testmodel
hops deployment predict testmodel --data '{"instances": [[1, 2, 3]]}'
hops deployment logs testmodel --tail 100
hops deployment stop testmodel
hops deployment delete testmodel

# Storage connectors (external data sources)
hops connector list
hops connector create snowflake my_sf \
  --url "https://xyz.snowflakecomputing.com" \
  --user admin --password secret --database MY_DB --schema PUBLIC --warehouse MY_WH
hops connector test my_sf
hops connector databases my_sf
hops connector tables my_sf --database MY_DB
hops connector preview my_sf --database MY_DB --table sales
hops connector delete my_sf

# External feature groups (backed by a connector)
hops fg create-external sales_fg \
  --connector my_sf \
  --query "SELECT id, name, amount FROM sales" \
  --primary-key id

# Jobs (full lifecycle)
hops job list
hops job create my_etl --type python --app-path Resources/jobs/etl.py
hops job create spark_job --type pyspark --app-path "hdfs:///Projects/myproject/Resources/jobs/etl.py"
hops job run my_etl --wait
hops job logs my_etl
hops job history my_etl
hops job schedule my_etl "0 0 * * * ?"    # every hour (Quartz cron)
hops job schedule-info my_etl
hops job unschedule my_etl
hops job stop my_etl
hops job delete my_etl

# Charts & Dashboards
hops chart generate --fg orders --x status --type pie
hops chart generate --fg orders --x category --y revenue --type bar --dashboard 1
hops chart list
hops dashboard list
hops dashboard create "My Dashboard"
hops dashboard add-chart 1 --chart 5

# Context dump (for LLMs)
hops context

Commands

Command	Description
hops login	Authenticate with Hopsworks
hops project list|use|info	Manage projects
hops fs list	List feature stores
hops fg list|info|preview|features|stats|keywords|add-keyword|remove-keyword|create|create-external|delete|insert|derive|search	Feature groups (with embeddings + KNN + keywords)
hops connector list|info|test|databases|tables|preview|create|delete	Storage connectors (Snowflake, JDBC, S3)
hops fv list|info|create|get|read|delete	Feature views (joins + transforms + online/batch read)
hops transformation list|create	Transformation functions
hops td list|create|compute|read|delete	Training datasets (materialize + retrieve with splits)
hops model list|info|register|download|delete	Model registry
hops deployment list|info|create|start|stop|predict|logs|delete	Model deployments (serving)
hops job list|info|create|run|stop|status|logs|history|delete	Full job lifecycle (Python, PySpark, Spark, Ray)
hops job schedule|schedule-info|unschedule	Cron scheduling (Quartz v2)
hops chart list|info|create|update|delete|generate	Charts (Plotly HTML from FG/FV data)
hops dashboard list|info|create|delete|add-chart|remove-chart	Dashboards (chart grid layout)
hops dataset list|mkdir	Browse project files
hops init	Set up Claude Code integration
hops context	Dump project state for LLMs

Claude Code Integration

hops init

This creates .claude/skills/hops/SKILL.md with:

• Frontmatter for auto-invocation when discussing feature stores
• Dynamic context (feature groups, feature views, jobs load automatically)
• Full CLI reference

It also adds Bash(hops *) to .claude/settings.local.json.

Open Claude Code in the project directory — type /hops or just ask about your feature store.

Output Modes

# Human-readable tables (default)
hops fg list

# JSON for programmatic use / LLMs
hops fg list --json

Authentication

• Inside Hopsworks terminal: Auto-detects REST_ENDPOINT, PROJECT_NAME, and JWT token. Zero config.
• Outside: Run hops login or set HOPSWORKS_API_KEY and REST_ENDPOINT env vars.

Config saved to ~/.hops/config.

Global Flags

--host <url>       Override Hopsworks host
--api-key <key>    Override API key
--project <name>   Override active project
--json             JSON output

Architecture

hopsworks-cli/
├── main.go              # Entry point
├── cmd/                 # Cobra commands (one file per resource)
│   └── templates/       # Embedded SKILL.md template
├── pkg/
│   ├── client/          # HTTP client for Hopsworks REST API
│   ├── config/          # Config management (~/.hops/config)
│   └── output/          # Table vs JSON formatting
└── docs/                # Plans & known issues