Adds Learn AI tutor as a feedback source

[rachellougee]

What are the relevant tickets?

#2534

Description (What does it do?)

Adds Learn AI tutor as a second source of the feedback system, so it flows through the rest of the feedback pipeline

• A new dbt model int__feedback__learn_ai_tutor
• Adds the chatbot data to int__feedback__unioned

How can this be tested?

Run dbt build --target dev_production --select int__feedback__learn_ai_tutor int__feedback__unioned

Additional Context

[Copilot]

🟡 Changes recommended

Downstream facts currently drop or incompletely resolve Learn AI records, and the selected business key is not migration-stable.

Get a fresh assessment by requesting another Copilot review.

Pull request overview

Adds Learn AI tutor messages as a second feedback source in the dbt pipeline.

Changes:

• Adds and documents the Learn AI feedback adapter.
• Unions tutor messages with Zendesk feedback.
• Adds cross-database support for nullable varchar arrays.

File summaries

File	Description
src/ol_dbt/seeds/feedback_sources.csv	Registers Learn AI Tutor.
src/ol_dbt/models/intermediate/feedback/int__feedback__unioned.sql	Adds tutor feedback to the common stream.
src/ol_dbt/models/intermediate/feedback/int__feedback__learn_ai_tutor.sql	Maps chatbot messages to the feedback contract.
src/ol_dbt/models/intermediate/feedback/_feedback_models.yml	Documents and tests the new model.
src/ol_dbt/macros/cross_db_functions.sql	Adds portable nullable varchar-array casts.

Review details

Suppressed comments (4)

src/ol_dbt/models/intermediate/feedback/int__feedback__learn_ai_tutor.sql:33

• This uses the Postgres row PK as the feedback business key, but the feedback contract explicitly specifies Learn AI's stable checkpoint_id. Because feedback_pk is derived from this value, a later source swap/backfill will mint different keys for the same turns. Select and use checkpoint_id here (including any namespace/thread qualification needed by its uniqueness contract), and retain djangocheckpoint_id only as an ordering tie-breaker.

    , cast(human_turns.djangocheckpoint_id as varchar) as source_record_ref

src/ol_dbt/models/intermediate/feedback/int__feedback__learn_ai_tutor.sql:48

• CanvasSyllabusBot is a documented value emitted by int__learn_ai__chatbot (_learn_ai__models.yml.yml:64-66), but it falls through this case to null. Since tfact_feedback.subject_type has a not_null test, any such conversation will fail the downstream build. Map it to the same course subject type as SyllabusBot.

    , case human_turns.chatsession_agent
        when 'TutorBot' then 'courseware_block'
        when 'VideoGPTBot' then 'courseware_block'
        when 'SyllabusBot' then 'course'
        when 'ResourceRecommendationBot' then 'resource'

src/ol_dbt/models/intermediate/feedback/int__feedback__learn_ai_tutor.sql:39

• This correctly emits a global user ID, but tfact_feedback currently resolves every subject_user_ref only against lower(dim_user.email) (tfact_feedback.sql:74-75). Authenticated Learn AI turns will therefore get a null user_fk unless the UUID happens to equal an email. Add source-aware identity resolution against dim_user.user_global_id, including the incremental re-key comparison.

    , human_turns.user_global_id as subject_user_ref

src/ol_dbt/models/intermediate/feedback/int__feedback__learn_ai_tutor.sql:43

• The seed marks this source as course-scoped and this adapter emits a course-run ID, but platform is always null and tfact_feedback still hardcodes both courserun_fk and platform_fk to null (tfact_feedback.sql:43-46). Consequently these rows cannot support the instructor/course filtering this source is meant to add. Populate the platform and resolve the conformed course/platform FKs downstream.

    , human_turns.courserun_readable_id
    , cast(null as varchar) as platform

• Files reviewed: 5/5 changed files
• Comments generated: 3
• Review effort level: Balanced

---

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

[github-actions]

🔎 ol-dbt impact — column-level blast radius

0 breaking, 0 surrogate-key regeneration, 1 warning, 1 info across 2 changed model(s).

Details

• ⚠️ macro: cross_db_functions.sql → 79 downstream model(s)
• ℹ️ int__feedback__learn_ai_tutor — channel_slug, conversation_ref, courserun_readable_id, created_at, explicit_rating, is_conversation_opening, occurred_at, platform, source_metadata, source_record_ref, source_slug, source_url, subject_ref, subject_type, subject_url, subject_user_ref, text, title, turn_index, updated_at → 0 downstream model(s)

_{Posted by ol-dbt impact (annotate-only — does not block merge).}