Files
windmill/system_prompts/base/resources.md
GuilhemandClaude Opus 5 e39dd7eb12 docs: teach agents to pass a resource as $res:<path> in run arguments (#10927)
* docs: teach agents to pass a resource as $res:<path> in run arguments

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01XjRARL7JA7xm772iJP4mJk

* docs: extend run-argument rule to in-editor chats, fix run-as wording

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01XjRARL7JA7xm772iJP4mJk

* docs: tighten resource run-argument rule after review

- Drop the false rationale that "$var:" only works inside a resource value
  from the write_variable description and its runtime rejection message; keep
  the rule (a variable cannot reference itself).
- MCP resource-argument description: the title fallback renders "No title",
  so say the title is only a label rather than that it can be empty. Guard the
  real-newline fix with asserts in the existing enrichment test.
- Eval: assert the full "$res:f/evals/global/github_main" value as one prefix
  so a wrong path with a right prefix fails.
- resources.md: narrow "a trigger's payload" to its configured static args.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01XjRARL7JA7xm772iJP4mJk

* docs: scope the run-argument rule to global chat, add an exact eval matcher

The ai_evals A/B on the two in-editor modes showed no effect: script mode
sonnet 5/5 both with and without the description, flow mode sonnet 5/5 and
haiku 5/5 on the baseline alone. A flow's input schema already carries
`format: resource-<type>`, so those modes have a signal global mode does not
give. Revert both files to keep the tool schemas free of a description that
buys nothing per iteration; global mode keeps it, where haiku goes 0/5 -> 5/5.

Add `stringEqualsAnyOf` to toolCallArgs and use it for the resource reference:
nothing in the eval resolves the value, so a prefix match accepted a near-miss
path like `$res:f/evals/global/github_main_backup`.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01XjRARL7JA7xm772iJP4mJk

* docs: address cubic review — CLI wording, mock resource getter

- `-d --data` help on all four run/preview commands: give $res: and $var:
  their own clauses instead of a parenthetical that read as if a resource
  were a kind of variable.
- Mock backend: `getBenchmarkResource` now resolves AI-provider seeds as well
  as plain ones, so it agrees with `existsResource` and `listResource` — both
  report either kind, and a case that listed a resource and then read it by
  path got a row it could not fetch.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01XjRARL7JA7xm772iJP4mJk

---------

Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
2026-09-03 11:29:54 +02:00

5.7 KiB

Windmill Resources

Resources store credentials and configuration for external services.

File Format

Resource files use the pattern: {path}.resource.json

Example: f/databases/postgres_prod.resource.json

Resource Structure

{
  "value": {
    "host": "db.example.com",
    "port": 5432,
    "user": "admin",
    "password": "$var:g/all/db_password",
    "dbname": "production"
  },
  "description": "Production PostgreSQL database",
  "resource_type": "postgresql"
}

Required Fields

  • value - Object containing the resource configuration
  • resource_type - Name of the resource type (e.g., "postgresql", "slack")

Variable References

Reference variables in resource values:

{
  "value": {
    "api_key": "$var:g/all/api_key",
    "secret": "$var:u/admin/secret"
  }
}

Reference formats:

  • $var:g/all/name - Global variable
  • $var:u/username/name - User variable
  • $var:f/folder/name - Folder variable

Resource References

Reference other resources:

{
  "value": {
    "database": "$res:f/databases/postgres"
  }
}

Passing a Resource or Variable as a Run Argument

A script or flow argument typed as a resource (schema format: resource-<type>) is passed as the bare string $res:<path> — the whole argument value. Same for a variable, with $var:<path>. This applies everywhere job arguments are supplied: wmill script run/preview, wmill flow run/preview, the runScriptByPath / runFlowByPath API, a schedule's args, a trigger's configured static args.

{
  "db": "$res:f/databases/postgres_prod",
  "api_token": "$var:g/all/api_token"
}

The reference is resolved when the job runs, under the job's run-as identity — the caller for an ordinary run, but the configured principal for a schedule, a trigger, or a runnable set to run on behalf of someone else. The run fails if that identity cannot read the referenced resource or variable.

Never wrap it in an object. The resolver only rewrites a JSON value that is a string starting with $res: / $var:; keys are never inspected. These are all wrong and are passed through to the script unchanged:

{ "db": { "$res": "f/databases/postgres_prod" } }
{ "db": { "resource": "f/databases/postgres_prod" } }
{ "db": "f/databases/postgres_prod" }

The string may sit anywhere a string can — a top-level argument, a nested object field ({ "gh_auth": { "token": "$var:g/all/gh_token" } }), or an array element (array elements are walked only while nested at most two levels deep, and only for arrays of at most 1000 items). The prefix must be on the string itself.

Common Resource Types

PostgreSQL

{
  "resource_type": "postgresql",
  "value": {
    "host": "localhost",
    "port": 5432,
    "user": "postgres",
    "password": "$var:g/all/pg_password",
    "dbname": "windmill",
    "sslmode": "prefer"
  }
}

MySQL

{
  "resource_type": "mysql",
  "value": {
    "host": "localhost",
    "port": 3306,
    "user": "root",
    "password": "$var:g/all/mysql_password",
    "database": "myapp"
  }
}

Slack

{
  "resource_type": "slack",
  "value": {
    "token": "$var:g/all/slack_token"
  }
}

AWS S3

{
  "resource_type": "s3",
  "value": {
    "bucket": "my-bucket",
    "region": "us-east-1",
    "accessKeyId": "$var:g/all/aws_access_key",
    "secretAccessKey": "$var:g/all/aws_secret_key"
  }
}

HTTP/API

{
  "resource_type": "http",
  "value": {
    "baseUrl": "https://api.example.com",
    "headers": {
      "Authorization": "Bearer $var:g/all/api_token"
    }
  }
}

Kafka

{
  "resource_type": "kafka",
  "value": {
    "brokers": "broker1:9092,broker2:9092",
    "sasl_mechanism": "PLAIN",
    "security_protocol": "SASL_SSL",
    "username": "$var:g/all/kafka_user",
    "password": "$var:g/all/kafka_password"
  }
}

NATS

{
  "resource_type": "nats",
  "value": {
    "servers": ["nats://localhost:4222"],
    "user": "$var:g/all/nats_user",
    "password": "$var:g/all/nats_password"
  }
}

MQTT

{
  "resource_type": "mqtt",
  "value": {
    "host": "mqtt.example.com",
    "port": 8883,
    "username": "$var:g/all/mqtt_user",
    "password": "$var:g/all/mqtt_password",
    "tls": true
  }
}

Custom Resource Types

Create custom resource types with JSON Schema:

{
  "name": "custom_api",
  "schema": {
    "type": "object",
    "properties": {
      "base_url": {"type": "string", "format": "uri"},
      "api_key": {"type": "string"},
      "timeout": {"type": "integer", "default": 30}
    },
    "required": ["base_url", "api_key"]
  },
  "description": "Custom API connection"
}

Save as: custom_api.resource-type.json

OAuth Resources

OAuth resources are managed through the Windmill UI and marked:

{
  "is_oauth": true,
  "account": 123
}

OAuth tokens are automatically refreshed by Windmill.

Using Resources in Scripts

TypeScript (Bun/Deno)

export async function main(db: RT.Postgresql) {
  // db contains the resource values
  const { host, port, user, password, dbname } = db;
}

Python

class postgresql(TypedDict):
    host: str
    port: int
    user: str
    password: str
    dbname: str

def main(db: postgresql):
    # db contains the resource values
    pass

CLI Commands

# List resources
wmill resource list

# List resource types with schemas
wmill resource-type list --schema

# Get specific resource type schema
wmill resource-type get postgresql

# Deploy resources to the workspace — destructive to remote state, so only run when
# the user explicitly asks to deploy/publish/push. Depending on how the repo is wired,
# deploy via `git push` or `wmill sync push` (see the Deploying section in AGENTS.wmill.md).
wmill sync push