name: hasura-metadata-apply description: >- Applies Hasura metadata changes (track tables, relationships, permissions) by editing hasura/metadata.json and running scripts/hasura_apply_metadata.sh against local dev compose. Use when tracking a new Postgres table in Hasura, adding relationships or permissions, or when the user asks to update Hasura metadata.

Hasura metadata apply

Overview

This project manages Hasura metadata as a single JSON file (hasura/metadata.json). Changes are applied via replace_metadata through scripts/hasura_apply_metadata.sh.

When to apply

• Tracking a new Postgres table in Hasura.
• Adding an object or array relationship on an existing tracked table.
• Changing select/insert/update/delete permissions for a role.
• User says "update Hasura", "track table", "add relationship", or "apply metadata".

Workflow

1. Edit metadata

Edit hasura/metadata.json (Hasura v2 metadata format, version 3).

Track a new table — add an entry to metadata.sources[0].tables:

{
  "table": { "name": "my_table", "schema": "public" },
  "object_relationships": [],
  "select_permissions": [
    {
      "role": "user",
      "permission": {
        "columns": ["col_a", "col_b"],
        "filter": {}
      }
    }
  ]
}

Add object relationship (e.g. 1:1 via FK on child table):

{
  "name": "my_table",
  "using": {
    "foreign_key_constraint_on": {
      "column": "parent_id",
      "table": { "name": "my_table", "schema": "public" }
    }
  }
}

Add array relationship (parent → children via FK on child):

{
  "name": "children",
  "using": {
    "foreign_key_constraint_on": {
      "column": "parent_id",
      "table": { "name": "child_table", "schema": "public" }
    }
  }
}

2. Validate JSON

python3 -m json.tool hasura/metadata.json > /dev/null

3. Apply to local Hasura

./scripts/hasura_apply_metadata.sh

Default endpoint: http://127.0.0.1:8080 (compose.dev.yaml). Default admin secret: password.

Override with env vars:

HASURA_URL=http://127.0.0.1:8080 HASURA_GRAPHQL_ADMIN_SECRET=mysecret ./scripts/hasura_apply_metadata.sh

The script uses jq (required dependency). It:

1. Extracts .metadata from the JSON file.
2. Sends replace_metadata to /v1/metadata.
3. Checks HTTP 200, no .error, and is_consistent == true.

4. Update client schema

After Hasura accepts metadata, re-introspect if the client's schema.graphql needs new types. The compose dev stack includes an introspection service, or run manually:

get-graphql-schema -h 'x-hasura-admin-secret=password' -h 'x-hasura-role=user' \
  http://localhost:8080/v1/graphql > packages/client/lib/data/gql/schema.graphql

Then run Ferry codegen:

cd packages/client && dart run build_runner build -d

5. Bump resource_version

Increment resource_version at the top of metadata.json after each change (convention only; Hasura ignores it on replace_metadata but it helps track edits).

Key conventions

• Permissions: most tables use "filter": {} (public read) for role: user. Tables with user-scoped data use "filter": { "user_id": { "_eq": "X-Hasura-User-Id" } }.
• Relationship naming: matches Postgres table name in snake_case (e.g. beacon_review_window).
• Computed fields: require session_argument and often table_argument; see existing examples in metadata.json.
• replace_metadata replaces everything: always edit the full metadata.json, never partial API calls.

Troubleshooting