fix(datasets): improve Dataset.generate polling robustness and timeout handling

[thiagobomfin-galileo]

User description

Shortcut:
NewSDK: extend() and generate() do NOT work
Description:

Summary

• Fix premature loop exit when steps_total=0 on the first poll of a synthetic dataset
  extension job, which caused Dataset.generate to return empty rows
• Replace fragile "Unexpected error" string matching on progress_message with stall
  detection: raises DatasetAPIException if steps_completed doesn't advance for 30 consecutive
  seconds
• Add timeout_seconds parameter (default 300s) to Datasets.extend, Dataset.generate, and
  the extend_dataset convenience function, replacing a hardcoded constant and preventing
  infinite polling loops
• Add a TODO to track replacing stall detection with a proper status/error field once the
  API exposes one in JobProgress
• Update docstrings across datasets.py and dataset.py to reflect new parameters and
  exception behavior

Test plan

• Call Dataset.generate(...) against a live environment and verify rows are returned
  correctly
• Verify a job that returns steps_total=0 on the first poll does not prematurely return
  empty rows
• Verify a stalled job raises DatasetAPIException with a meaningful message after 30s of no
  progress
• Verify a timed-out job raises DatasetAPIException after timeout_seconds
• Verify passing a custom timeout_seconds is respected
• Run poetry run pytest tests/test_datasets.py — all extend-related tests pass

Tests:

• Unit Tests Added
• E2E Test Added (if it's a user-facing feature, or fixing a bug)

---

Generated description

Below is a concise technical summary of the changes proposed in this PR:
Harden the Datasets.extend and extend_dataset polling loop with timeout, stall detection, and updated docstrings so synthetic extension jobs no longer return empty rows and surface DatasetAPIException when stalled. Teach Dataset.generate to forward the same timeout_seconds contract so callers receive generated rows instead of empty lists, with regression tests covering zero-step starts, stalls, and timeouts.

Topic	Details
Generation flow	Teach Dataset.generate to pass through timeout_seconds so that generated rows respect the new polling contract. Modified files (1) src/galileo/__future__/dataset.py Latest Contributors(1) User Commit Date vamaq@users.noreply.gi... fix-dataset-204-respon... February 07, 2026
Synthetic extension	Harden the extend polling loop by logging progress, adding timeout_seconds, detecting 30s stalls before raising DatasetAPIException, and documenting the new behavior, with regression tests covering zero-step starts, timeouts, and stalls. Modified files (2) src/galileo/datasets.py tests/test_datasets.py Latest Contributors(2) User Commit Date bipin@galileo.ai fix-normalize-ground_t... February 18, 2026 vamaq@users.noreply.gi... fix-dataset-204-respon... February 07, 2026

This pull request is reviewed by Baz. Review like a pro on (Baz).

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 82.03%. Comparing base (5a7c747) to head (753441b).

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #494      +/-   ##
==========================================
+ Coverage   82.00%   82.03%   +0.02%     
==========================================
  Files          96       96              
  Lines        9260     9273      +13     
==========================================
+ Hits         7594     7607      +13     
  Misses       1666     1666              

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.