dbt-labs/dbt-core · error

get_csv_data: argument must be an AgateTable

Error message

get_csv_data: argument must be an AgateTable

What it means

When `adapter.get_csv_data(...)` receives an argument but it cannot be downcast to the internal `AgateTable` object type, the dispatcher raises this `InvalidOperation` error. The method only understands real agate-table wrapper objects produced by the seed-loading pipeline, not plain strings, dicts, or other Value types.

Source

Thrown at crates/dbt-adapter/src/adapter/mod.rs:4371

                // model: dict, settings: str -> str
                Ok(args.get(1).cloned().unwrap_or_else(|| Value::from("")))
            }
            "get_ch_database" => {
                // schema: str -> str (CH database = schema in 2-part naming)
                Ok(args.first().cloned().unwrap_or_else(|| Value::from("")))
            }
            "get_csv_data" => {
                let table = args
                    .first()
                    .ok_or_else(|| {
                        minijinja::Error::new(
                            minijinja::ErrorKind::MissingArgument,
                            "get_csv_data requires agate_table argument",
                        )
                    })?
                    .downcast_object::<AgateTable>()
                    .ok_or_else(|| {
                        minijinja::Error::new(
                            minijinja::ErrorKind::InvalidOperation,
                            "get_csv_data: argument must be an AgateTable",
                        )
                    })?;
                self.get_csv_data(table)
            }
            "get_credentials" => self.get_credentials(args),
            "render_equals" => {
                let iter = ArgsIter::new(name, &["expr1", "expr2"], args);
                let expr1 = iter.next_arg::<&str>()?;
                let expr2 = iter.next_arg::<&str>()?;
                iter.finish()?;
                self.render_equals(state, expr1, expr2)
            }
            _ => Err(minijinja::Error::new(
                minijinja::ErrorKind::UnknownMethod,
                format!("Unknown method on adapter object: '{name}'"),
            )),

View on GitHub (pinned to 0267ce9170)

Solutions

1. Pass the actual agate table object (typically the `agate_table` variable in seed macros), not a path string or dict.
2. If you only have a CSV file, load it through the engine's seed-loading path to obtain an AgateTable before calling get_csv_data.
3. Log/type-check the argument with `is mapping` / `is string` tests in Jinja to catch wrong inputs before dispatching.

Example fix

// before (Jinja)
{% set csv = adapter.get_csv_data(seed_file_path) %}
// after
{% set csv = adapter.get_csv_data(agate_table) %}

{# Jinja: string path or mapping means you don't have a table object #}
{% if agate_table is string or agate_table is mapping %}
  {{ exceptions.raise_compiler_error("get_csv_data needs an AgateTable, not a path/dict") }}
{% endif %}

{# Jinja: AgateTable is an opaque object; reject obvious non-tables #}
{% set looks_like_table = agate_table is defined and agate_table is not string and agate_table is not mapping and agate_table is not sequence %}

When it happens

Trigger: Calling `adapter.get_csv_data(x)` where `x` is a string path to a CSV, a dict, a Jinja list, or any Value that is not an `AgateTable` object — for example passing `config.get('seed_path')` or the text output of another call instead of the loaded table.

Common situations: Custom macros that load CSV contents themselves and pass the raw string; confusion between a seed's file path and its parsed table; macros ported from dbt-core where the table was a Python agate object but here must be the engine's AgateTable wrapper.

Understand the failure class

Background: "Must be a positive integer", "Invalid value", "Unsupported": the invalid-argument-value error family, when a library rejects the value you pass — this error's family across 35 libraries.

Related errors

• remote_state must be Object
• In adapter.dispatch, got a macro name of "{macro_name}", but
• role must be a string
• get_csv_data requires agate_table argument
• get_csv_data: failed to format CSV: {e}

AI-assisted analysis of dbt-labs/dbt-core@0267ce9170 (2026-09-07). Data as JSON: /api/errors/f845a684951d3c07. Report an issue: GitHub.