ai-kit

Host an MCP server as a private, single-tenant Keboola data app — a dedicated remote MCP endpoint on your own Keboola compute (Keboola MCP is the built-in example, but it wraps any streamable-HTTP / FastMCP server) with client auth. Use when a user wants a private or self-hosted MCP endpoint, or would benefit from one but hasn't asked (see "When to use this"). Keywords: host/self-host MCP server as Keboola data app, private/remote MCP endpoint, deploy MCP to Keboola.

Use right after a Keboola semantic-context tool returns, when building an app, query, report, or transformation from a semantic model. Ground the SQL in the model's own physical mapping (dataset `tableId`/`fqn`, metric SQL) and confirm the real columns before writing any query — do not build SQL from the business-facing display names.

Use when building, modifying, deploying, or debugging Keboola Apps (Streamlit or Python/JS). Covers the full lifecycle — choosing app type, configuring keboola-config/, storage access (RO workspace, RW Query Service, input mapping), authentication, DuckDB caching for performance, default Keboola styling, dashboard patterns, optional Kai chat integration, and the three client paths (MCP-only, Claude Code with filesystem, kbagent CLI).

Use when accessing a Keboola-managed Git (Forgejo) repo for a python-js data app via the kbagent CLI — provisioning the repo, finding its repo_url, minting a git_clone_url / push credential with create_python_js_data_app_git_credential, pushing source to Keboola git, copying source between GitHub and Keboola git, working around the ~15MB / HTTP 413 push cap with build-at-deploy, or deploying a data app from its managed git repo. Triggers: kbagent, Keboola managed git, Forgejo, data app repo, repo_url, git_clone_url, create git credential, push to Keboola git, build-at-deploy, HTTP 413, deploy from git.

Use this skill to migrate an existing Microsoft Power BI semantic model into a Keboola semantic layer model. Accepts two input shapes: (1) TMDL folder extracted with the Microsoft Power BI Modeling MCP server in `--readonly` mode (canonical, captures DAX + Power Query M expressions verbatim); (2) per-table JSON produced by older `get_semantic_model` calls (missing M steps). Maps Power BI tables → Keboola semantic-dataset, measures → semantic-metric (DAX preserved verbatim, flagged for SQL translation), relationships → semantic-relationship. Produces JSON files ready to push to the Keboola metastore via `sl-toolkit` or the REST API. Use for: migrate PowerBI semantic model, convert Power BI to Keboola SL, translate TMDL, import PowerBI tables, powerbi-to-sl, pbi-migration, brownfield semantic layer. Trigger when the user has PowerBI artifacts (TMDL folder, `*_semantic_layer.json`, `.bim`, `.pbip`) and wants them as a Keboola semantic layer.

Semantic layer reference for Keboola. Auto-loads when the user mentions: semantic layer, semantic model, SL, metastore, semantic-metric, semantic-dataset, semantic-relationship, semantic-glossary, semantic-constraint, inspect model, validate model, threshold constraints, metric sql, model entities, add dataset, create metric, update metric, delete metric, remove constraint. Provides API reference, payload shapes, CRUD procedures, validation rules, and operational gotchas so Claude can work with the metastore directly.

Use this skill for managing and reviewing Keboola projects. Activates when syncing project configs (init, pull, push, diff), running project reviews, analyzing SQL transformations, auditing security, assessing performance, validating financial logic, or evaluating template readiness. Covers the full Keboola CLI workflow and 10-agent review team.

Expert knowledge for writing, optimizing, and migrating DuckDB transformations in Keboola. Covers SQL dialect, block orchestration, dynamic backends, Parquet format, case sensitivity, Snowflake-to-DuckDB migration, type casting patterns, and best practices. Use when the user is writing DuckDB SQL, migrating from Snowflake, debugging DuckDB transformations, or needs guidance on DuckDB-specific syntax and configuration in Keboola.

Canonical template files for Keboola Python components — Dockerfile, pyproject.toml, push.yml, build_n_test.sh, docker-compose.yml, pre-commit-config.yaml, and config-schema.md. Load this whenever creating or modifying any of these files in a component, or when checking whether non-source files are aligned with the official cookiecutter template. Any deviation from these templates must have an explicit reason. Internal utility — usually invoked via Task from develop-component or migrate-to-uv, but should be consulted any time one of these files is being touched.

Migrate Keboola Python components to modern uv build system with deterministic dependencies and ruff linting.

Expert skill for debugging Keboola Python components. Use when a component is failing, a job returned an error, or behavior is unexpected. Uses whatever tools are available — Keboola MCP for job/config inspection, Datadog for logs, Linear/Jira for issue context, Slack for incident history, and local Bash for reproducing issues. Invoke for "failing job", "exit code 2", "component throwing an error", "why is my component not working".

Expert Keboola Python component developer for ongoing development work — implementing features, extending extractors/writers/applications, adding incremental loads, designing configuration schemas, separating API clients, applying self-documenting workflow patterns, and maintaining code quality with Ruff. Use this skill whenever the user is working on component logic, implementation, or architecture — whether building something new or extending existing code. Do NOT use for initial project scaffolding (that's get-started) or UI-only schema work (that's build-component-ui).

Keboola platform context — automatically load this whenever working on any Keboola Python component task (developing, testing, debugging, reviewing, schema design). Contains platform-level knowledge about how the Keboola Connection platform executes components, manages configuration, handles state, and runs jobs. This is not about code patterns — it is about platform behaviour that affects how code should be written and tested. Always pull this before making decisions about config structure, row handling, state, parallelism, or test data layout.

Full code review for Keboola Python components — covers both code quality (architecture, config/client patterns, typing, Pythonic best practices) and backward compatibility (configSchema changes, sync actions, output tables, state files, telemetry impact). Use for PRs, pre-merge checks, and component audits. Invoke whenever someone asks to review code, check a PR, audit before release, or verify that changes are safe for existing users.

Expert agent for writing and maintaining tests for Keboola Python components. Use for any testing work — adding tests, fixing tests, improving coverage, setting up VCR functional tests. Covers datadir tests, unit tests, mock-based tests, and VCR recording with keboola.datadirtest. Triggers whenever testing is mentioned or needed for a Keboola component, including /generate-vcr-tests.

Expert in Keboola configuration schemas, conditional fields (options.dependencies), UI elements, sync actions, and schema testing. Can launch schema-tester and run Playwright tests. Specialized for configSchema.json and configRowSchema.json development.

Mandatory initialization workflow for new Keboola Python components — use this skill before writing any code. Contains the exact cookiecutter command, Keboola-specific post-template cleanup steps, KBC_DATADIR directory structure, data/config.json format, and first commit format that are easy to get wrong without this reference. Invoke whenever the user is starting from a fresh or empty repository, has never built a Keboola component before, wants to run the cookiecutter template, or says things like "new extractor for X", "build ex-something from scratch", "fresh repo", "empty git repo", "brand new component", "never done this in keboola", "initialize a component", "scaffolding a new component", "just initialized the git repo", "nothing set up yet". Not for existing components that already have src/component.py.

Use this skill when working with Keboola project configurations, understanding JSON config files, editing transformations, or analyzing Keboola project structure. Triggers on questions about Keboola configs, transformations, orchestrations, extractors, writers, or .keboola directories.