dss-ai-tools

Tools for exploring Data Shape Server (DSS) knowledge graph schemas — classes, properties, their relations, and namespaces — over the DSS HTTP API. There are three ways to use it, all sharing one small client:

Component	What it is	Where
CLI	A single-file Python client (dss), standard library only. Requires Python >= 3.10.	dss.py
Claude skill	A self-contained Claude Code skill that drives the CLI.	skill/
MCP server	An MCP server exposing the same endpoints as tools.	mcp-server/

The CLI implements the core functionality; the skill and MCP server bundle a copy of dss.py so each is self-contained. Run scripts/sync.sh after editing the canonical dss.py to refresh the copies.

Configure

All three approaches use the same environment variables:

export DSS_BASE_URL=https://dss.semtech.lv   # default
export DSS_TIMEOUT=30                        # optional, seconds

The CLI and MCP tools also accept a per-call override (--base-url / base_url).

CLI

python3 dss.py --help

Install from PyPI to get the dss command on your PATH:

pipx install dss-ai-tools     # provides the `dss` command
# or run without installing:
uvx --from dss-ai-tools dss --help

Or install from a checkout of this repo:

pipx install .          # provides the `dss` command (see pyproject.toml)

Commands

DSS refers to data schemas as ontologies and uses that term throughout its API; the CLI follows the same naming.

Command	Description
ontologies [--variant ...] [--tag TAG]	List the ontologies (schemas) loaded in DSS; --variant selects an alternate info view, --tag filters (with variant 3).
schema-tags	List the tags used to group ontologies.
public-ns	List the globally known namespace prefixes shared across ontologies.
namespaces <ont>	List the namespace prefixes declared in one ontology.
classes <ont> [--limit N] [--filter STR]	List classes in an ontology; narrow with a name filter.
properties <ont> [--limit N] [--filter REGEX] [--kind ...]	List properties in an ontology, optionally by kind; narrow with a name filter.
resolve-class <ont> <name>	Look up a class by its prefixed name; returns its metadata, including its id.
resolve-property <ont> <name>	Look up a property by its prefixed name; returns its metadata, including its id.
class-out-properties <ont> <class_id> [--limit N]	List the properties that originate from a class (outgoing).
class-in-properties <ont> <class_id> [--limit N]	List the properties that point to a class (incoming).
class-pairs <ont> <p_list>	List the (source, target) class pairs that a property connects.
call <ont> <fn> [--body JSON|-] [--param k=v ...]	Generic API call: invoke any DSS function not covered by the commands above.

The class-* commands use numeric ids: get a class id from resolve-class and a property id from resolve-property.

Output is indented human-readable JSON by default; pass --compact for single-line (pipe-friendly).

dss ontologies | jq '.[].db_schema_name'
dss classes dbpedia --filter Person --limit 20
dss resolve-property dbpedia dbo:birthPlace
dss class-pairs war_sampo 116

Claude skill

skill/ is a self-contained Claude Code skill. Install it user-level:

mkdir -p ~/.claude/skills
cp -R skill ~/.claude/skills/dss

See docs/USAGE.md for setup, pointing it at a server, and example prompts.

MCP server

See mcp-server/README.md.

You can register it from the published PyPI package (no checkout needed).

Exit codes (CLI)

• 0 — success, JSON on stdout
• 2 — HTTP / transport / argument error, message on stderr
• 130 — interrupted (Ctrl-C)

License

MIT