feat(b2b_learner_records): serve /courses from mv_b2b_contract_courserun

[blarghmatey]

What are the relevant tickets?

N/A. Serves listCourses from the spec in #55. Stacked on #56, which left /courses out for lack of a backing view. Reads the MV added in mitodl/ol-data-platform#2684.

Description (What does it do?)

Adds GET /organizations/{organization_id}/courses to the learner-records tenant. It takes contract_id and limit/offset, is ordered by (contract_id, courserun_id), and reads mv_b2b_contract_courserun. The grant dependency is the router's, same as /learners and /enrollments.

queries.courses() builds its SQL without _assemble. The rows carry no personal data, so there's no consent projection, and outcomes_withheld_count is a literal 0, which is what the envelope documents for endpoints with no consent gate. CourseRun follows the spec schema: contract dates as date, run start/end as UTC date-time, and every field required, nulls included.

How can this be tested?

• uv run pytest: 237 passed. The new cases cover the view and ordering, the bound contract_id, the envelope's zero withheld count, date serialization, and CourseRun in the required-field schema test. The existing client-credentials security test iterates every path, so it covers this one too.
• uv run ruff check and ruff format --check on the tenant and its test file, plus uv run mypy src/ol_analytics_api/tenants/b2b_learner_records: clean.
• The SQL hasn't been run on StarRocks; the MV doesn't exist until the data-platform PR merges.

Additional Context

Measured on production data (see the data-platform PR): contract_start_date is null for every contract, null in the raw MITx Online extract as well, and contract_end_date is null for all but one. Clients will see nulls there until that's populated upstream.

Checklist:

• Merge feat(b2b_learner_records): add the learner-records tenant #56 first
• Confirm b2b_learner_records.mv_b2b_contract_courserun exists before deploying

🤖 Generated with Claude Code

https://claude.ai/code/session_01EDJRuj6mDpGRqNbQSsXZ9o

[Copilot]

🟡 Changes recommended

The generated CourseRun schema omits the contract’s int64 format and non-negative seat-limit constraint.

Get a fresh assessment by requesting another Copilot review.

Pull request overview

Adds the learner-records /courses endpoint backed by the contract/course-run materialized view.

Changes:

• Adds contract and course-run response modeling and queries.
• Supports contract filtering and deterministic pagination.
• Adds endpoint and schema coverage.

File summaries

File	Description
models.py	Defines CourseRun.
queries.py	Queries and counts course runs.
routers/organizations.py	Exposes listCourses.
tests/test_learner_records.py	Tests filtering, ordering, serialization, and schema.

Review details

Suppressed comments (1)

src/ol_analytics_api/tenants/b2b_learner_records/models.py:104

• The contract declares seat_limit with minimum: 0, but this field currently accepts negative values and omits that constraint from the generated schema. Add the non-negative bound so runtime validation and generated clients match docs/openapi/b2b-learner-records-v1.yaml:744-747.

    seat_limit: int | None = Field(description="Null means uncapped, not zero.")

• Files reviewed: 4/4 changed files
• Comments generated: 1
• Review effort level: Balanced

---

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

[blarghmatey]

Addressed Copilot's suppressed finding on models.py:104 (not posted as an inline thread): fixed in 60336cd, CourseRun.seat_limit now enforces ge=0 to match the contract's minimum: 0. The int64-format thread (PRRT_kwDOTPlRi86iq6F2) was already resolved and stays that way.