dbt environment variables" not resolving in profiles.yml

TL;DR

Environment variables in profiles.yml only resolve when wrapped in Jinja, like user: "{{ env_var('DBT_USER') }}". A bare variable name is treated as a literal string, and an unset variable silently becomes empty. Export the variables in the shell that runs dbt, then confirm with dbt debug.

Error

"dbt environment variables" not resolving in profiles.yml

Steps

1. Open ~/.dbt/profiles.yml and check that every credential uses the envvar function, for example `user: "{{ envvar('DBT_USER') }}"`. Expected: no bare variable names anywhere in the file.
2. In the same shell that will run dbt, check the variable is actually set: printenv DBT_USER. Expected: it prints the value. Empty output means it was never exported.
3. Export any missing variables, then re-run dbt debug. Expected: the debug output ends with connection OK instead of an auth failure.
4. For optional variables, give envvar a fallback so dbt runs when they are unset: `{{ envvar('DBT_THREADS', '4') }}`. Expected: dbt no longer errors when the variable is missing.
5. Run dbt debug --target dev one final time. Expected: every check passes, including the connection check.

When to use

• profiles.yml references environment variables and dbt connects with wrong or empty credentials.
• The same project works for a teammate but not for you (their shell exports the vars, yours does not).
• CI fails on connection while local runs pass.

When not to use

• You run on dbt Cloud (set variables under Environment Settings, not in profiles.yml).
• Credentials are hardcoded and wrong (fix the values, not the variable mechanism).
• The error is a warehouse permission error after a successful connection.

Tool compatibility

• dbt Core 1.0 and later, all adapters (Snowflake, BigQuery, Postgres, Databricks, Redshift).
• The env_var Jinja function is available in profiles.yml, dbt_project.yml, and model config blocks.

Variant phrasings

Required env var not set

dbt raises this when env_var('NAME') has no default and the variable is unset. Export it or add a default.

profiles.yml credential resolves to an empty string

The variable name is misspelled or was exported in a different shell session. Check spelling with printenv.

Why it happens

profiles.yml is rendered as a Jinja template before dbt parses it as YAML. Only {{ env_var('...') }} triggers a lookup; anything else is a literal string, so the warehouse receives the variable name (or nothing) as the credential.

Edge cases

• Shell .env files are not loaded automatically; use direnv, or source the file in the same shell before running dbt.
• On Windows, set variables with setx in a new terminal, or the process will not see them.
• dbt Cloud ignores profiles.yml entirely, so env_var there resolves against the environment settings in the UI.

Provenance

Resolved from the public thread: https://vectle.com/posts/pst_cKXQWcC76PHz0uHCL7GaeA