feat(AI-3112): support keboola.duckdb-transformation via generic config tools

[Matovidlo]

Description

Linear: AI-3112

Change Type

• Major (breaking changes, significant new features)
• Minor (new features, enhancements, backward compatible)
• Patch (bug fixes, small improvements, no new features)

Summary

Adds keboola.duckdb-transformation as a supported transformation type, routed through the
existing generic create_config/update_config tools — the same path Python/R transformations
already use. create_sql_transformation/update_sql_transformation stay Snowflake/BigQuery-only,
since their component ID is derived from the project's workspace SQL dialect and DuckDB has no
cloud workspace backend.

Changes:

• utils.py: new DUCKDB_TRANSFORMATION_ID constant, added to FOLDER_SUPPORTING_COMPONENT_IDS
  so folder metadata management works the same as for Python/R.
• tools.py: docstrings for create_config, create_sql_transformation, and
  update_sql_transformation, plus the update_sql_transformation 404 error message, updated to
  mention DuckDB alongside Python/R so the model routes DuckDB requests correctly.
• No new tools, no workspace changes. DuckDB-specific sync actions (syntax_check,
  lineage_visualization, execution_plan_visualization, expected_input_tables) and config
  parameters (backend_size, timeout, duckdb_version, etc.) require no new code — they're
  already handled generically by run_sync_action and the opaque-JSON parameters field.

RFC: feature_spec/duckdb_transformation_support/RFC.md

Follow-up (separate Linear issue, own RFC, not in this PR): AI-3662 —
consolidating the per-backend transformation tools and moving shared guidance into
project_system_prompt.md.

Testing

• Tested with Cursor AI desktop (Streamable-HTTP transports)

Optional testing

• Tested with Cursor AI desktop (all transports)
• Tested with claude.ai web and canary-orion MCP (Streamable-HTTP)
• Tested with In Platform Agent on canary-orion
• Tested with RO chat on canary-orion

Checklist

• Self-review completed
• Unit tests added/updated (if applicable)
• Integration tests added/updated (if applicable)
• Project version bumped according to the change type
• Documentation updated (if applicable)

[linear-code]

AI-3112

[Matovidlo]

@claude review

[Copilot]

Pull request overview

Adds DuckDB transformation support (keboola.duckdb-transformation) by routing it through the existing generic configuration tools, aligning it with the existing Python/R transformation flow (and keeping the SQL-transformation tools Snowflake/BigQuery-only).

Changes:

• Added DUCKDB_TRANSFORMATION_ID and enabled folder metadata handling for DuckDB transformations via update_config.
• Updated tool docstrings and the update_sql_transformation 404 guidance to explicitly point DuckDB users to create_config/update_config.
• Extended unit tests to cover DuckDB folder behavior and updated error-message expectations; bumped project version to 1.75.0 (plus regenerated docs/lock updates).

Reviewed changes

Copilot reviewed 6 out of 7 changed files in this pull request and generated no comments.

Show a summary per file

File	Description
src/keboola_mcp_server/tools/components/utils.py	Adds DuckDB transformation component ID constant and includes it in folder-support set.
src/keboola_mcp_server/tools/components/tools.py	Updates docstrings and improves 404 guidance to steer DuckDB to generic config tools.
tests/tools/components/test_tools.py	Extends existing parametrized tests to cover DuckDB folder handling and error message content.
TOOLS.md	Regenerated tool documentation reflecting updated docstrings/guidance.
feature_spec/duckdb_transformation_support/RFC.md	Adds an RFC documenting the motivation, scope, and verification plan for DuckDB support.
pyproject.toml	Version bump to 1.75.0.
uv.lock	Lockfile updated to reflect the version bump.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Matovidlo]

Rebased onto the AI-3664 fix branch (martinvasko-ai-3664-migrate-flow-integration-tests-off-legacy-orchestrator) as an interim step to unblock Integration Tests, which were failing due to the now-disabled legacy Orchestrator flow API (see #653). This PR's diff will temporarily include #653's commits until that PR merges to main, at which point I'll rebase this branch onto main to drop the duplication.

[Copilot]

Pull request overview

Copilot reviewed 9 out of 10 changed files in this pull request and generated no new comments.

Suppressed comments (2)

integtests/tools/flow/test_tools.py:833

• The cleanup block swallows deletion exceptions but logs without traceback, which makes integration-test failures much harder to debug. Use structured logging with exc_info=True (or LOG.exception) when catching and continuing.

            except Exception as e:
                LOG.warning(f'Failed to delete {flow_type} flow {flow_id}: {e}')

integtests/tools/flow/test_tools.py:424

• get_flows() lists all flows in the project (it does not filter to only those created by these fixtures), so asserting exact equality of IDs will fail in projects that already have flows. This test should assert that the two fixture-created flow IDs are present (subset), not that they are the only results.

    assert {f.configuration_id for f in flows.flows} == {
        initial_cf.configuration_id,
        initial_cf2.configuration_id,
    }

[keboola-pr-reviewer-bot]

Verdict: auto_approve (risk 2/5) · profile keboola-mcp-server

Additive DuckDB component-ID support via existing generic config tools, with linked RFC, version bump, and extended tests — safe to auto-approve.

Impact flags: scope creep — see Check Run summary.

Concerns:

• integtests/tools/flow/test_tools.py: Unrelated flow-test refactor bundled into a DuckDB PR; drops legacy Orchestrator coverage.

Dismissing prior approval — a new commit was pushed and this review was for an earlier SHA. Run @keboola-pr-reviewer-bot review to get a fresh verdict.

[keboola-pr-reviewer-bot]

New commit on 4fab217 — dismissed 1 stale bot approval. Comment @keboola-pr-reviewer-bot review when you want a fresh review.

[tomasfejfar]

👀

[tomasfejfar]

Doesn't do the same for update.

Also, while it works, it's not great. DuckDB transformation is essentially a SQL transformation and should be treated as such. Limitations of our tool shouldn't dictate that it goes to code-like transformations. But I wouldn't block on that alone. The missing update is the blocker.

[tomasfejfar]

Also:

Friction summary: get_components and get_config_examples — the two tools the system prompt explicitly directs you to consult before calling create_config — return nothing usable for keboola.duckdb-transformation. The only way to actually learn the config shape was docs_query, and even that needed a second, more targeted question after the first answer resolved to the wrong (but similarly-named) component. A user following the documented discovery steps literally would hit two dead ends before finding a working path.