fix: shared fixes so the dbt Power User tool PRs merge in any order

HK
2m ago
+185-14

Issue for this PR

None

Type of change

  • Bug fix
  • New feature
  • Refactor / code improvement
  • Documentation

What does this PR do?

This is one change to the files that the nine dbt Power User tool PRs share (#1729, #1730, #1731, #1732, #1733, #1734, #1735, #1737, #1738). With it, those PRs can merge in any order with green tests, and none of them needs to touch a shared file.

  • Placeholder tests no longer assume every tool is a stub (test/altimate/dbt-adapter-access.test.ts). A tool counts as a placeholder only while its source file still contains the stub's sentence "is not implemented in this build of altimate-code yet." The not-implemented checks run over those tools only, and zero placeholders is allowed. New tests check that:

    • every feature tool has its source file, so a renamed file can't silently drop out of the check;
    • every tool, built or not, is registered under its own id;
    • the marker agrees with what each tool actually answers.

    The argument-check test still covers all nine tools, because real tools also reject a non-object.

  • dbt_sql_graph answers as MCP structured content (mcp-server/server.ts): model, nodes, edges and error, the same pattern as dbt_lineage. Before this, an MCP client got the graph as text only. The test sits next to the other structuredResult tests in column-reach.test.ts.

  • Run-results readers can take a target folder (validators/validator-utils.ts):

    • resolveDbtTargetPath(dbtRoot, targetPath?): a folder the caller names wins over the project's own.
    • readRunResults(dbtRoot, targetPathOverride?) also returns invocationId, elapsedSec and each node's executionTimeSec. It refuses a file that is rewritten while it is being read, checked by its bigint stat identity.

    Existing callers are unchanged. Tests are in dbt-build-green.test.ts and dbt-nothing-built.test.ts.

Not included: exporting sessionProject from dbt-lineage.ts, because nothing outside the file uses it.

Merge this PR before the nine tool PRs.

How did you verify your code works?

  • git merge-tree --write-tree against each of the nine tool branches: no conflicts.
  • tsgo --noEmit and bun test test/altimate test/tool-cli on four setups: main plus this branch; that plus each tool branch alone; and that plus all nine together.
    • Typecheck passed in every setup.
    • The only test failures were 5-second timeouts in suites this PR doesn't touch, mostly carry-forward/agent-safety.test.ts. They pass when rerun alone (52/52 in the all-nine checkout).
  • With all nine merged, the touched suites (dbt-adapter-access, column-reach, validators) pass 1056/1056.

Cost side effects

None. These are test changes, one more MCP structured-content entry, and a local file read.

Screenshots / recordings

Not a UI change.

Checklist

  • I have tested my changes locally
  • I have not included unrelated changes in this PR
  • I checked for cost side effects (model calls, queries or cloud resources that grow with user data) and described them, or wrote "none"

🤖 Generated with Claude Code

Summary by CodeRabbit

  • New Features
    • SQL graph results now include structured model and graph details. Successful results show a brief report, while failed results retain the full report and error details.
    • Build results now include invocation, elapsed-time, and per-model execution-time details. A custom target location can be used to find build results.
  • Bug Fixes
    • Build freshness checks correctly handle matching timestamps.
    • Results that change while being read are treated as unavailable.