Persist commitment costs on scheduling job meta

[Flix6x]

The bug

make_schedule records the scheduler's commitment cost breakdown on the job:

if rq_job and result.get("name") == "commitment_costs":
    rq_job.meta["scheduler_info"]["commitment_costs"] = result["data"]

…but this runs after the function's last rq_job.save_meta() call, and RQ persists a finishing job with include_meta=False. So the costs were computed and then silently dropped: no scheduling job has ever carried commitment_costs in the meta actually stored in Redis. The ordering has been wrong since the feature was introduced in #1946, so this never worked. Any consumer reading job meta (e.g. the jobs API) saw a scheduler_info without them.

Observed live: after a successful StorageScheduler job, Job.fetch(...) returns scheduler_info containing only {"scheduler": ...}, while the in-memory job object in the worker (e.g. seen from an RQ success callback) still holds the full cost dict.

The fix

Call rq_job.save_meta() right where the costs are recorded, with a comment explaining why the explicit save is load-bearing.

The test

The assertions live in test_scheduling_a_battery, which already runs a real battery scheduling job with these fixtures. After the job finishes it fetches a fresh Job instance from Redis and asserts the persisted meta carries a non-empty, finite cost breakdown — deliberately not inspecting the worker's in-memory object, which held the data even when the bug was live.

Mutation-tested (each mutant applied to scheduling.py, test rerun):

Mutant	Outcome
Fix reverted (no save_meta)	killed
save_meta() before the assignment	killed
Costs stored under a wrong key	killed
Costs replaced by {}	killed
Branch condition inverted	killed

Pristine fix: passes. Full test_scheduling_jobs.py module: 14 passed.

No changelog entry: the feature this repairs is itself part of unreleased v1.0.0 (#1946).

🤖 Generated with Claude Code

https://claude.ai/code/session_01UKx19DmqHiTBS7iEeFTijs

[Copilot]

Pull request overview

Fixes a bug where make_schedule computed and attached commitment_costs to the RQ job meta but never persisted it to Redis, meaning downstream consumers (e.g. jobs API) could not read commitment cost breakdowns from stored job meta.

Changes:

• Persist commitment_costs immediately via rq_job.save_meta() at the point they are written to rq_job.meta["scheduler_info"]
• Add a regression test that runs a real scheduling job and then re-fetches the Job from Redis to assert the persisted meta contains finite, non-empty commitment costs

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 1 comment.

File	Description
flexmeasures/data/services/scheduling.py	Persist commitment_costs into Redis by saving job meta immediately after writing the field
flexmeasures/data/tests/test_scheduling_jobs.py	Add regression test verifying commitment_costs are present on a freshly fetched finished job

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[Copilot]

Pull request overview

Copilot reviewed 2 out of 2 changed files in this pull request and generated no new comments.

[read-the-docs-community]

Documentation build overview

📚 flexmeasures | 🛠️ Build #33972054 | 📁 Comparing 811b236 against latest (8c4e7d3)

  🔍 Preview build  

1 file changed

± api/v3_0.html