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:

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.

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