fix: allows nullable source_id in taxonomy

[pandeymangg]

Problem

feedback_records.source_id is nullable in the base Hub schema (001_initial_schema.sql), so feedback can be ingested with no attributed source. But taxonomy treated source_id as required everywhere, so those records could never get a taxonomy — they were silently invisible to discovery and rejected at run creation.

Approach

Taxonomy now supports a nullable source scope instead of forcing feedback_records.source_id to NOT NULL (which would be a broader ingestion contract change).

Key decision: rather than making source_id nullable inside the composite PK/FK/UNIQUE constraints — where NULL != NULL breaks key matching and forces a source_key/surrogate-PK workaround — taxonomy canonicalizes "no source" to the empty string '' at the Go boundary. The Go layer always sends a string, never NULL, so
source_id stays NOT NULL in taxonomy tables and '' is a valid, comparable key value. NULL stays confined to feedback_records, and only the queries that read it use null-safe matching.

Changes

• migrations/010_taxonomy_schema.sql — dropped the three *_source_id_required blank-check constraints (taxonomy_runs, taxonomy_active_runs, taxonomy_node_events).
  source_id stays NOT NULL (satisfied by ''); composite keys unchanged. Added a header note documenting the empty-string convention. (Edited in place — taxonomy is unreleased,
  no new migration needed.)
• internal/models/taxonomy.go — TaxonomyScope.SourceID validation relaxed from required,min=1 to omitempty,no_null_bytes,max=255.
• internal/service/id_validation.go — added normalizeOptionalIdentifier (empty → ""); shared checks extracted into normalizeIdentifierValue.
• internal/service/taxonomy_service.go — normalizeTaxonomyScope now treats source_id as optional.
• internal/repository/taxonomy_repository.go — ListFieldOptions removed the source_id IS NOT NULL AND btrim(...) <> '' filter and now groups on and emits
  COALESCE(NULLIF(btrim(source_id), ''), '') so NULL and blank collapse into a single "no source" bucket; CountScopeInput and GetRunInput match feedback_records with
  NULLIF(btrim(fr.source_id), '') IS NOT DISTINCT FROM NULLIF(btrim($3), '').

Tests

• tests/taxonomy_no_source_test.go (new) — NULL + blank feedback collapse into one bucket; empty-source scope null-safe matches both rows; run creation and the in-progress guard work with the '' scope.
• tests/taxonomy_schema_test.go (existing) — still passes, confirming the constraint drop didn't break schema relationships/cascades.

Verification

• go build ./..., go vet ./..., golangci-lint — clean.
• Migrations roll back and re-apply cleanly.
• Taxonomy integration tests pass against a real Postgres.

Out of scope / follow-ups

• feedback_records.source_id stays nullable — ingestion contract unchanged.
• openapi.yaml — no taxonomy schemas there yet (unreleased), nothing to update.
• ListRuns filter still treats empty source_id as "no filter" (history query). Listing only no-source runs would need an explicit flag — not added here.
• Frontend renders the empty-source_id bucket as "No source" / "Unattributed"; that label lives in the frontend repo. The API now emits an empty source_id for the bucket.

[coderabbitai]

Walkthrough

This PR introduces a canonical "no source" bucket for taxonomy scopes where feedback_records.source_id is NULL or blank. The empty string "" is adopted as the internal representation. Three CHECK constraints requiring btrim(source_id) <> '' are removed from the migration schema. TaxonomyScope.SourceID validation is changed from required to omitempty. In the service layer, a new normalizeOptionalIdentifier function accepts blank values as "" without error, and normalizeTaxonomyScope switches to using it for SourceID. Repository SQL is updated to normalize source_id via COALESCE(NULLIF(btrim(...), ''), '') in projections and grouping, and uses IS NOT DISTINCT FROM with NULLIF for NULL-safe predicate matching in CountScopeInput and GetRunInput. A new integration test validates the end-to-end behavior across ListFieldOptions, CountScopeInput, and CreateRunIfAvailable.

🚥 Pre-merge checks | ✅ 4 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 33.33% which is insufficient. The required threshold is 80.00%.	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (4 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title accurately summarizes the main change: making source_id optional/nullable in taxonomy validation and schema constraints.
Description check	✅ Passed	The PR description is comprehensive, covering problem statement, approach, detailed changes by file, tests, verification steps, and out-of-scope items. All required template sections are well-addressed.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[coderabbitai]

Actionable comments posted: 1

🤖 Prompt for all review comments with AI agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Inline comments:
In `@internal/models/taxonomy.go`:
- Line 64: The change to make TaxonomyScope.SourceID optional with the omitempty
tag creates a filtering problem in the list-runs path where empty source_id
values are dropped, making it impossible to distinguish between "no filter
applied" and "filter by empty source_id". Update the filter contract across the
handler, service, and repository layers for the list-runs operation to use a
tri-state filter mechanism (such as a pointer to string (*string) or an explicit
presence flag) instead of a plain string filter. This ensures that empty-string
filtering is preserved and passed through the entire stack, allowing callers to
properly scope run history to the no-source bucket while maintaining consistent
tenant access rules across all resource access paths.

🪄 Autofix (Beta)

Fix all unresolved CodeRabbit comments on this PR:

• Push a commit to this branch (recommended)
• Create a new PR with the fixes

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Organization UI

Review profile: ASSERTIVE

Plan: Pro

Run ID: fc25f77f-26bb-4355-98dc-2bfe074cc837

📥 Commits

Reviewing files that changed from the base of the PR and between 9d175ee and d396a51.

📒 Files selected for processing (6)

• internal/models/taxonomy.go
• internal/repository/taxonomy_repository.go
• internal/service/id_validation.go
• internal/service/taxonomy_service.go
• migrations/010_taxonomy_schema.sql
• tests/taxonomy_no_source_test.go