CLI

daeda is the Typer-based command group that exposes every Daedalus operator surface: catalog/service inspection, the Version Protocol, materialization, the embedding store, the training pipeline, controlled serving, agent config, and the control-plane servers.

Every command and option below is transcribed verbatim from daeda ... --help. Run a command with --help at any level to see the same text inline.

bash

uv run daeda --help

Global options

bash

uv run daeda [OPTIONS] COMMAND [ARGS]...

🛸 Daedalus CLI — Feature store utilities for smoother engineering experience.

Option	Description
`--version`	Show CLI version and exit.
`--install-completion`	Install completion for the current shell.
`--show-completion`	Show completion for the current shell, to copy it or customize the installation.
`--help`	Show this message and exit.

Commands

Command	Description
`lineage`	Show upstream/downstream lineage for one feature view or service.
`materialize-day`	Materialize one training-day partition so an agent can inspect sample output.
`mcp`	Run the Daedalus MCP server over stdio.
`serve-api`	Run the Daedalus JSON-RPC control plane server.
`catalog`	Inspect the feature catalog: list views, show schema, dump lineage.
`config`	Persist AI-agent CLI configuration to `~/.daedalus/config.toml`.
`dev`	Local dev utilities.
`pipeline`	Load and run YAML-defined feature pipelines.
`service`	Inspect feature services: list contracts, show schema, dump lineage.
`materialize`	Query and manage materialization metadata: list, show, gc.
`registry`	Version Protocol: lock, check, and diff the resolved feature schemas.
`serving`	Inspect and query materialized serving relations.
`store`	Build and inspect the declared embedding stores.

catalog

Inspect the feature catalog: list views, show schema, dump lineage.

bash

uv run daeda catalog [OPTIONS] COMMAND [ARGS]...

catalog list

List feature views in the catalog with a one-line summary each.

bash

uv run daeda catalog list [OPTIONS]

Option	Default	Description
`-d, --definitions-dir DIRECTORY`	`feature_views`	Directory containing YAML feature view definitions.
`-f, --format TEXT`	`table`	Output format: 'table' (default) or 'json'.
`--help`		Show this message and exit.

catalog show

Print a feature view's metadata + schema as JSON.

bash

uv run daeda catalog show [OPTIONS] NAME

Argument: name (required) — Feature view name.

Option	Default	Description
`-d, --definitions-dir DIRECTORY`	`feature_views`	Directory containing YAML feature view definitions.
`--help`		Show this message and exit.

catalog lineage

Dump the catalog's lineage graph as JSON or Graphviz DOT.

bash

uv run daeda catalog lineage [OPTIONS]

Option	Default	Description
`-d, --definitions-dir DIRECTORY`	`feature_views`	Directory containing YAML feature view definitions.
`-f, --format TEXT`	`json`	Output format: 'json' (default) or 'dot' (Graphviz).
`--help`		Show this message and exit.

service

Inspect feature services: list contracts, show schema, dump lineage.

bash

uv run daeda service [OPTIONS] COMMAND [ARGS]...

service materialize

Materialize a feature service's offline serving snapshot to Parquet.

bash

uv run daeda service materialize [OPTIONS] NAME

Argument: name (required) — Feature service name.

Option	Default	Description
`--feature-views-dir DIRECTORY`	`feature_views`	Directory containing YAML feature view definitions.
`--feature-services-dir DIRECTORY`	`feature_services`	Directory containing YAML feature service definitions.
`--output-root DIRECTORY`	`output`	Directory where materialized Parquet snapshots and metadata.db are stored.
`--repo-root DIRECTORY`		Root used to resolve repo-relative feature-view source paths. Omit to preserve the catalog paths unchanged.
`--run-id TEXT`		Explicit run identifier. Auto-generated if omitted.
`--attempt INTEGER RANGE [x>=1]`	`1`	Attempt number (increment when retrying the same run_id).
`-f, --format TEXT`	`table`	Output format: 'table' (default) or 'json'.
`--online`		Publish the materialized snapshot to the online store.
`--online-backend TEXT`	`redis`	Online store backend used with --online: 'memory' or 'redis'.
`--redis-url TEXT`	`redis://localhost:6379`	Redis/Valkey URL used when --online is set.
`--online-prefix TEXT`	`daedalus:`	Redis key prefix used when --online is set.
`--help`		Show this message and exit.

service list

List feature services with a one-line summary each.

bash

uv run daeda service list [OPTIONS]

Option	Default	Description
`--feature-views-dir DIRECTORY`	`feature_views`	Directory containing YAML feature view definitions.
`--feature-services-dir DIRECTORY`	`feature_services`	Directory containing YAML feature service definitions.
`-f, --format TEXT`	`table`	Output format: 'table' (default) or 'json'.
`--help`		Show this message and exit.

service show

Print a feature service's full resolved contract as JSON (redacted).

bash

uv run daeda service show [OPTIONS] NAME

Argument: name (required) — Feature service name.

Option	Default	Description
`--feature-views-dir DIRECTORY`	`feature_views`	Directory containing YAML feature view definitions.
`--feature-services-dir DIRECTORY`	`feature_services`	Directory containing YAML feature service definitions.
`--no-redact`		Show unredacted sensitive metadata (trusted local use only).
`-f, --format TEXT`	`json`	Output format: 'json' (default).
`--help`		Show this message and exit.

service lineage

Dump the platform lineage graph (service → view → source).

bash

uv run daeda service lineage [OPTIONS]

Option	Default	Description
`--feature-views-dir DIRECTORY`	`feature_views`	Directory containing YAML feature view definitions.
`--feature-services-dir DIRECTORY`	`feature_services`	Directory containing YAML feature service definitions.
`-f, --format TEXT`	`json`	Output format: 'json' (default) or 'dot' (Graphviz).
`--help`		Show this message and exit.

registry

Version Protocol: lock, check, and diff the resolved feature schemas.

bash

uv run daeda registry [OPTIONS] COMMAND [ARGS]...

registry lock

Write registry.lock.json from the current view + service definitions.

bash

uv run daeda registry lock [OPTIONS]

Option	Default	Description
`-d, --feature-views-dir DIRECTORY`	`feature_views`	Directory containing YAML feature view definitions.
`-s, --feature-services-dir DIRECTORY`	`feature_services`	Directory containing YAML feature service definitions.
`-l, --lock-file PATH`	`registry.lock.json`	Path to registry.lock.json.
`--help`		Show this message and exit.

registry check

Fail (exit non-zero) on a breaking change or a missing version bump.

bash

uv run daeda registry check [OPTIONS]

Option	Default	Description
`-d, --feature-views-dir DIRECTORY`	`feature_views`	Directory containing YAML feature view definitions.
`-s, --feature-services-dir DIRECTORY`	`feature_services`	Directory containing YAML feature service definitions.
`-l, --lock-file PATH`	`registry.lock.json`	Path to registry.lock.json.
`--help`		Show this message and exit.

registry diff

Show the structural diff against the lock (informational; never fails).

bash

uv run daeda registry diff [OPTIONS]

Option	Default	Description
`-d, --feature-views-dir DIRECTORY`	`feature_views`	Directory containing YAML feature view definitions.
`-s, --feature-services-dir DIRECTORY`	`feature_services`	Directory containing YAML feature service definitions.
`-l, --lock-file PATH`	`registry.lock.json`	Path to registry.lock.json.
`-f, --format TEXT`	`table`	Output format: 'table' (default) or 'json'.
`--help`		Show this message and exit.

materialize

Query and manage materialization metadata: list, show, gc.

bash

uv run daeda materialize [OPTIONS] COMMAND [ARGS]...

materialize list

List materialization records, newest first.

bash

uv run daeda materialize list [OPTIONS]

Option	Default	Description
`--output-root DIRECTORY`	`output`	Directory containing metadata.db and output artifacts.
`-s, --service TEXT`		Filter records by service name.
`--status TEXT`		Filter by status: running, succeeded, or failed.
`-n, --limit INTEGER RANGE [x>=1]`	`20`	Maximum records to return.
`-f, --format TEXT`	`table`	Output format: 'table' (default) or 'json'.
`--help`		Show this message and exit.

materialize show

Show the full materialization record(s) for a run_id.

bash

uv run daeda materialize show [OPTIONS] RUN_ID

Argument: run_id (required) — The run_id of the record to show.

Option	Default	Description
`--output-root DIRECTORY`	`output`	Directory containing metadata.db and output artifacts.
`-f, --format TEXT`	`json`	Output format: 'json' (default).
`--help`		Show this message and exit.

materialize gc

Garbage-collect old materialization records and optional output directories.

bash

uv run daeda materialize gc [OPTIONS]

Option	Default	Description
`--output-root DIRECTORY`	`output`	Directory containing metadata.db and output artifacts.
`-s, --service TEXT`		Limit garbage collection to this service.
`--older-than TEXT`		Prune records older than an ISO datetime or day count (e.g. '2024-01-01T00:00:00Z' or '7d').
`-k, --keep-last INTEGER RANGE [x>=0]`		Keep the most recent N records per service.
`--delete-outputs`		Also delete the output directories for pruned records.
`--execute`		Execute the prune (default is dry-run).
`--help`		Show this message and exit.

pipeline

Load and run YAML-defined feature pipelines.

bash

uv run daeda pipeline [OPTIONS] COMMAND [ARGS]...

pipeline train

Canonical training entry point. Compile a feature service and run it through the operator executor as one operator DAG. Output is byte-identical to the now-deprecated daeda skeleton + daeda enrich (the executor wraps the same proven engine entry points).

--only selects a DAG sub-range, not a stage

There are no separate "skeleton" and "enrich" stages — just one operator DAG. --only source runs a contiguous sub-range of that DAG (the source-side operators that read sources in place and build the spine); --only enrich runs the complementary sub-range (the enrichment operators that append the Ray/Lance embeddings). --only sql / --only sql-arrow-udf / --only pythonic narrow to a single engine. See the Architecture Overview.

bash

uv run daeda pipeline train [OPTIONS] SERVICE

Argument: service (required) — Feature service name.

Option	Default	Description
`--feature-views-dir PATH`	`feature_views`	Directory of feature view YAML.
`--feature-services-dir PATH`	`feature_services`	Directory of feature service YAML.
`--runtime-config-path PATH`	`feature_services/pixai_feed_relevance/settings.yaml`	Training settings.yaml.
`--aggregation-config-path PATH`	`config/training/aggregations.yaml`	Training aggregations.yaml.
`--output-root PATH`		Override the skeleton output root.
`--target-date TEXT`		A single YYYY-MM-DD target day.
`--day-workers INTEGER`		Concurrent target days (default config).
`--only TEXT`	(full DAG)	Run only selected engine units: `source` \| `enrich` \| `sql` \| `sql-arrow-udf` \| `pythonic`.
`--ray-num-cpus INTEGER`	`16`	Ray logical CPUs (enrich).
`--help`		Show this message and exit.

pipeline compile

Compile a feature service into an editable operator-pipeline YAML. The emitted DAG records each operator's declared engine (sql / sql_arrow_udf / pythonic); hand-edit <output-dir>/<service>.yaml to override. <service>.generated.yaml is the regenerated artifact.

bash

uv run daeda pipeline compile [OPTIONS] SERVICE

Argument: service (required) — Feature service name.

Option	Default	Description
`--feature-views-dir PATH`	`feature_views`	Directory of feature view YAML.
`--feature-services-dir PATH`	`feature_services`	Directory of feature service YAML.
`--runtime-config-path PATH`	`feature_services/pixai_feed_relevance/settings.yaml`	Training settings.yaml.
`--aggregation-config-path PATH`	`config/training/aggregations.yaml`	Training aggregations.yaml.
`--output-dir PATH`	`config/pipelines`	Where to write the generated pipeline YAML.
`--write` / `--no-write`	`write`	Write the YAML (else stdout only).
`--help`		Show this message and exit.

pipeline list

List YAML-defined pipelines in pipelines-dir.

bash

uv run daeda pipeline list [OPTIONS]

Option	Default	Description
`-d, --pipelines-dir DIRECTORY`	`config/pipelines`	Directory of YAML pipeline definitions.
`-f, --format TEXT`	`table`	Output format: 'table' (default) or 'json'.
`--help`		Show this message and exit.

pipeline show

Print a pipeline's definition as JSON.

bash

uv run daeda pipeline show [OPTIONS] NAME

Argument: name (required) — Pipeline name.

Option	Default	Description
`-d, --pipelines-dir DIRECTORY`	`config/pipelines`	Directory of YAML pipeline definitions.
`--help`		Show this message and exit.

pipeline run

Execute a pipeline end-to-end (Source → Ops → Sink).

bash

uv run daeda pipeline run [OPTIONS] NAME

Argument: name (required) — Pipeline name.

Option	Default	Description
`-d, --pipelines-dir DIRECTORY`	`config/pipelines`	Directory of YAML pipeline definitions.
`--no-sink`		Skip the sink write step. Useful for dry-runs and previewing the pipeline output without touching the destination.
`--help`		Show this message and exit.

serving

Inspect and query materialized serving relations.

bash

uv run daeda serving [OPTIONS] COMMAND [ARGS]...

serving health

Print the backend health status as JSON.

bash

uv run daeda serving health [OPTIONS]

Option	Default	Description
`--output-root PATH`	`output`	Root directory containing materialized snapshots.
`--backend TEXT`	`local`	Serving backend to use (local or quack).
`--help`		Show this message and exit.

serving relations

List the approved relation names for a service.

bash

uv run daeda serving relations [OPTIONS] SERVICE

Argument: service (required) — Service name.

Option	Default	Description
`--output-root PATH`	`output`	Root directory containing materialized snapshots.
`--feature-views-dir DIRECTORY`	`feature_views`	Directory containing YAML feature view definitions.
`--feature-services-dir DIRECTORY`	`feature_services`	Directory containing YAML feature service definitions.
`--help`		Show this message and exit.

serving schema

Print the Arrow schema (field names + types) for a relation as JSON.

bash

uv run daeda serving schema [OPTIONS] SERVICE

Argument: service (required) — Service name.

Option	Default	Description
`-r, --relation TEXT` (required)		Relation name to inspect.
`--output-root PATH`	`output`	Root directory containing materialized snapshots.
`--feature-views-dir DIRECTORY`	`feature_views`	Directory containing YAML feature view definitions.
`--feature-services-dir DIRECTORY`	`feature_services`	Directory containing YAML feature service definitions.
`--backend TEXT`	`local`	Serving backend to use (local or quack).
`--help`		Show this message and exit.

serving scan

Scan a relation and print rows as JSON.

bash

uv run daeda serving scan [OPTIONS] SERVICE

Argument: service (required) — Service name.

Option	Default	Description
`-r, --relation TEXT` (required)		Relation name to scan.
`--output-root PATH`	`output`	Root directory containing materialized snapshots.
`--feature-views-dir DIRECTORY`	`feature_views`	Directory containing YAML feature view definitions.
`--feature-services-dir DIRECTORY`	`feature_services`	Directory containing YAML feature service definitions.
`--backend TEXT`	`local`	Serving backend to use (local or quack).
`-n, --limit INTEGER`		Maximum rows to return.
`-c, --columns TEXT`		Comma-separated column names to include.
`--help`		Show this message and exit.

serving serve

Start a long-running serving daemon for a service. Blocks until SIGINT or SIGTERM, then shuts down cleanly.

bash

uv run daeda serving serve [OPTIONS] SERVICE_NAME

Argument: service_name (required) — Service name.

Option	Default	Description
`--backend TEXT`	`quack`	Serving backend to use (local or quack).
`--host TEXT`	`127.0.0.1`	Host to bind the Quack server to.
`--port INTEGER RANGE [0<=x<=65535]`	`0`	Port to bind the Quack server to (0 = auto-pick).
`--token TEXT`		Bearer token for Quack auth. If omitted, Quack generates one.
`--output-root PATH`	`output`	Root directory containing materialized snapshots.
`--feature-views-dir DIRECTORY`	`feature_views`	Directory containing YAML feature view definitions.
`--feature-services-dir DIRECTORY`	`feature_services`	Directory containing YAML feature service definitions.
`--help`		Show this message and exit.

config

Persist AI-agent CLI configuration to ~/.daedalus/config.toml (override path via $DAEDALUS_CONFIG).

bash

uv run daeda config [OPTIONS] COMMAND [ARGS]...

config init

Write the default agent config to disk.

bash

uv run daeda config init [OPTIONS]

Option	Default	Description
`--force`		Overwrite an existing config file with defaults.
`--help`		Show this message and exit.

config show

Print the effective agent CLI config.

bash

uv run daeda config show [OPTIONS]

Option	Default	Description
`-f, --format TEXT`	`toml`	Output format: 'toml' (default) or 'json'.
`--help`		Show this message and exit.

config set

Persist one config key/value pair.

bash

uv run daeda config set [OPTIONS] KEY VALUE

Arguments: key (required) — Config key to update; value (required) — String value to persist.

Option	Default	Description
`--help`		Show this message and exit.

config path

Print the config file path.

bash

uv run daeda config path [OPTIONS]

Option	Default	Description
`--help`		Show this message and exit.

dev

Local dev utilities.

bash

uv run daeda dev [OPTIONS] COMMAND [ARGS]...

dev seed

Seed synthetic parquet for the catalog's declared source paths.

bash

uv run daeda dev seed [OPTIONS]

Option	Default	Description
`--data-root DIRECTORY` (required)		Root directory where catalog source paths will be seeded.
`--rows INTEGER RANGE [x>=0]`	`1000`	Rows to write per unique source.
`--seed INTEGER`	`42`	Base RNG seed.
`--id-pool INTEGER RANGE [x>=1]`	`200`	Shared ID pool size per entity.
`--feature-views-dir DIRECTORY`	`feature_views`	Directory containing YAML feature view definitions.
`--help`		Show this message and exit.

store

Build and inspect the declared embedding stores.

bash

uv run daeda store [OPTIONS] COMMAND [ARGS]...

The Lance embedding pool is the one artifact Daedalus cannot read in place — it is local-path only and needs a filesystem with atomic rename. See Local Lance embedding store for the full setup, including the download step that produces the source parquet.

store list

List the declared embedding stores and whether each is built.

bash

uv run daeda store list [OPTIONS]

Option	Default	Description
`--aggregation-config FILE`	`config/training/aggregations.yaml`	Config declaring `embedding_stores`.
`--help`		Show this message and exit.

store build

Build a declared Lance store from pre-downloaded hive-partitioned parquet.

bash

uv run daeda store build [OPTIONS] NAME

Option	Default	Description
`NAME` (required)		Name of an `embedding_stores` entry to build.
`--source DIRECTORY`	declared `source_parquet`	Root of the pre-downloaded hive-partitioned vectors.
`--store-path PATH`	declared `path`	Build somewhere other than the declared path.
`--source-filter TEXT`	declared `source_filter`	SQL predicate scoping the pool; `''` builds every partition.
`--rebuild`	off	Delete an existing store first. Without it, an existing store is an error.
`--duckdb-threads INTEGER RANGE [x>=1]`	`16`	DuckDB threads for the streaming read.
`--duckdb-memory-limit TEXT`	`32GiB`	DuckDB `memory_limit` for the streaming read.
`--allow-network-filesystem`	off	Skip the filesystem check. Lance needs atomic rename; a store on GPFS/NFS corrupts under concurrent actors.
`--aggregation-config FILE`	`config/training/aggregations.yaml`	Config declaring `embedding_stores`.
`--help`		Show this message and exit.

lineage

Show upstream/downstream lineage for one feature view or service.

bash

uv run daeda lineage [OPTIONS] NAME

Argument: name (required) — Feature view or service name.

Option	Default	Description
`-f, --format TEXT`	`table`	Output format: 'table' (default) or 'json'.
`--help`		Show this message and exit.

materialize-day

Materialize one training-day partition so an agent can inspect sample output.

bash

uv run daeda materialize-day [OPTIONS] SERVICE DATE

Arguments: service (required) — Training service name; date (required) — Partition date (YYYY-MM-DD).

Option	Default	Description
`--help`		Show this message and exit.

serve-api

Run the Daedalus JSON-RPC control plane server.

bash

uv run daeda serve-api [OPTIONS]

Option	Default	Description
`--host TEXT`	`127.0.0.1`	Host interface to bind the FastAPI server to.
`--port INTEGER`	`8000`	TCP port to bind the FastAPI server to.
`--help`		Show this message and exit.

mcp

Run the Daedalus MCP server over stdio.

bash

uv run daeda mcp [OPTIONS]

Option	Default	Description
`--help`		Show this message and exit.

CLI ​

Global options ​

Commands ​

catalog ​

catalog list ​

catalog show ​

catalog lineage ​

service ​

service materialize ​

service list ​

service show ​

service lineage ​

registry ​

registry lock ​

registry check ​

registry diff ​

materialize ​

materialize list ​

materialize show ​

materialize gc ​

pipeline ​

pipeline train ​

pipeline compile ​

pipeline list ​

pipeline show ​

pipeline run ​

serving ​

serving health ​

serving relations ​

serving schema ​

serving scan ​

serving serve ​

config ​

config init ​

config show ​

config set ​

config path ​

dev ​

dev seed ​

store ​

store list ​

store build ​

lineage ​

materialize-day ​

serve-api ​

mcp ​

CLI

Global options

Commands

catalog

catalog list

catalog show

catalog lineage

service

service materialize

service list

service show

service lineage

registry

registry lock

registry check

registry diff

materialize

materialize list

materialize show

materialize gc

pipeline

pipeline train

pipeline compile

pipeline list

pipeline show

pipeline run

serving

serving health

serving relations

serving schema

serving scan

serving serve

config

config init

config show

config set

config path

dev

dev seed

store

store list

store build

lineage

materialize-day

serve-api

mcp