dbt-labs / dbt-labs/dbt-context-engineering
Fusion + DuckDB: package-owned seed fails to load (doubled dbt_packages path)
- Dominant language
- Python
- Stars
- 5
- Forks
- 0
- Avg merge
- 15h 54m
- Merged PRs (30d)
- 9
Description
## Summary
On dbt Fusion (`dbt-fusion`), the DuckDB adapter fails to load a seed owned by an installed package. The seed's file path gets doubled, so the loader can never find the CSV.
## Where this shows up in this repo
`integration_tests/duckdb` builds `dbt_context_engineering`'s own `embedding_canary_baseline` seed when `ai_functions_enabled: true` is set (the canary is disabled by default). Running `dbtf build --vars '{ai_functions_enabled: true}'` against that project fails:
```
[error] [DbDriverFailed (dbt1308)]: Database Error in seed embedding_canary_baseline
IO Error: No files found that match the pattern
".../integration_tests/duckdb/dbt_packages/dbt_context_engineering/dbt_packages/dbt_context_engineering/seeds/embedding_canary_baseline.csv"
```
Note the doubled `dbt_packages/dbt_context_engineering/` segment.
A plain `dbtf build` (canary disabled by default) is unaffected.
## Confirmed scope
- Fusion-engine-only: `dbt-core` 1.12 loads the same seed correctly on every adapter.
- DuckDB-adapter-only under Fusion: Fusion loads the same seed fine on Snowflake, BigQuery, and Databricks.
- Not specific to this repo's local self-referencing package setup (`local: ../..`): reproduced identically in a minimal scratch project that installs `dbt_context_engineering` as a normal git package, matching how a real downstream consumer would install it.
## Tracking
Upstream issue: dbt-core#16128. Not fixable from this package's side; resolves automatically once dbt-fusion ships a fix.
## Suggested follow-up here
If a Fusion CI job is ever added to `.github/workflows/ci.yml`, it will need to exclude the canary seed/model on the DuckDB target until the upstream fix lands, or that job will be permanently red for a reason unrelated to this package's own code.
Contributor guide
Research direction
Start by running `dbtf build --vars '{ai_functions_enabled: true}'` in `integration_tests/duckdb` and compare the failing seed path with the dbt-core#16128 report. Read `.github/workflows/ci.yml` if evaluating the suggested CI follow-up; done means Fusion ships the upstream fix or the CI job explicitly excludes the DuckDB canary seed/model until then.
Written by the indexing model from the issue text.
Assessment
- Domain
- databases
- Issue type
- Bug
- Difficulty
- 4/5
- Estimated time
- 3-5 days
- Activity status
- Active
- Clarity
- Needs clarification
- Newbie friendliness
- 25/100