Ceres — Claude Code Skill

A Claude Code Skill for Ceres, the harvest-first toolkit for open data portals.

What it does

This skill gives Claude deep knowledge of Ceres as it exists today: a harvesting-first system with optional embedding, search, export, and API layers. It helps with:

Using Ceres CLI (harvest, embed, search, export, stats)
Configuring portals in portals.toml
Running and querying the REST API
Understanding the harvest pipeline, incremental sync, and delta detection
Working with Ollama-first local embeddings and optional hosted providers
Understanding CKAN and DCAT udata portal support
Implementing custom EmbeddingProvider, PortalClient, or DatasetStore traits
Contributing to the Ceres codebase

Install

Fastest way

Ask Claude to kindly install the skill for you!

Claude Code (CLI)

git clone https://github.com/AndreaBozzo/Ceres-Claude-Skill.git
mkdir -p ~/.claude/skills
ln -s "$(pwd)/Ceres-Claude-Skill/ceres" ~/.claude/skills/ceres

The skill is now available in all Claude Code sessions. Since it's a symlink, git pull is all you need to update.

Claude (Web/Desktop)

Build the ZIP:

cd ceres && zip -r ../ceres-skill.zip .

Upload to Claude → Settings > Capabilities > Custom Skills

Structure

ceres/
├── SKILL.md                        # Entry point — overview, pipeline, traits, quick start
└── references/
  ├── architecture.md             # Crate graph, services, job system, DB schema
  ├── cli-and-server.md           # CLI commands, REST API, env vars, deployment
  ├── harvesting.md               # Harvest-first sync model, delta detection, metadata-only flow
  ├── extending.md                # Implementing providers, portal clients, stores, queues
  └── contributing.md             # Dev setup, testing, CI, release workflow

License

Apache-2.0