DSL Grammar Overview

import { Aside } from ‘@astrojs/starlight/components’;

The Cyntex DSL is YAML format with file extension .cyn.yml, declared with apiVersion: cyntex/v1.

The JSON Schema for the DSL is available at `docs.cyntex.io/schema/v1/`, supporting IDE completion and real-time validation. VS Code users: associate `"*.cyn.yml"` → `cyntex/v1` schema in your workspace `settings.json`.

Resource Types

Cyntex uses a multi-resource declaration model. Each .cyn.yml file can contain one resource, distinguished by kind:

Kind	Responsibility
source	Defines a data source connection (connector type + config + mining mode)
pipeline	Defines a data pipeline (references source + transforms + output)

Future extensions: workspace (batch multi-resource), serve (API publishing, GA phase)

---

source

Describes a data source: connection configuration + read mode.

mysql-prod.cyn.yml

apiVersion: cyntex/v1
kind: source
id: mysql-prod

connector: mysql
mode: cdc        # batch | cdc | api | file

config:
  host: db.internal
  port: 3306
  database: production
  username: ${MYSQL_USER}
  password: ${MYSQL_PASS}

options:
  start_from: latest    # earliest | latest | <ISO timestamp>

tables:
  - name: users         # literal = fixed table name
  - name: /orders_.*/   # regex = dynamic matching (requires discovery capability)

mode — Four Values

mode	Semantics	Typical connectors
batch	Full snapshot (bounded)	MySQL, PostgreSQL, files
cdc	Continuous change capture (unbounded)	MySQL binlog, PG WAL, MongoDB oplog
api	API / SaaS pull (polling or webhook)	Salesforce, HubSpot
file	File scanning	S3, SFTP, local directory

Valid combinations of connector and mode are verified by the capability matrix (bundled catalog) at validate time — invalid combinations produce an error, not silent failures.

tables Omission Semantics

• tables omitted = all tables (equivalent to /.*/)
• Literal string = fixed table name; does not track new tables
• /regex/ = dynamic: new tables matching the regex automatically join the pipeline (requires discovery capability)

---

pipeline

Describes a data pipeline: reference source → optional transforms → output (sync / push).

user-sync.cyn.yml

apiVersion: cyntex/v1
kind: pipeline
id: user-profile-sync

source: mysql-prod    # references the id of a kind:source

tables:
  - name: users
  - name: /orders_.*/

transforms:
  - name: filter-active
    filter: "record.status == 'active'"   # CEL expression

  - name: enrich-type
    js: |
      record.tier = record.spend > 1000 ? 'premium' : 'standard';
      return record;

sync:
  - source: users
    target:
      collection: user_profiles
    options:
      write_mode: upsert    # upsert | append
      ddl: apply            # apply | ignore | fail (default: fail)

settings:
  error_policy: fail        # fail | skip | dlq
  batch_size: 1000
  parallelism: 1

transforms (Transform Nodes)

Each transform is a processing step in the pipeline, executed in declaration order.

Node type	Field	Semantics
Filter	filter: <CEL>	Discard rows that do not satisfy the condition
JS transform	js: <script>	GraalVM JS row-by-row processing
rename	rename: {old: new}	Field renaming
typeFilter	type_filter: <type>	Filter by data type
unwind	unwind: <array field>	Array flattening
nest	nest:	Master-detail merge (Beta; MERGE family)

`nest` (master-detail merge) is not available in Alpha; it lands in Beta. See [Roadmap](/overview/roadmap/).

sync Output

Full snapshot / CDC write to target storage:

sync:
  - source: <table name or regex>        # references the logical stream from tables
    target:
      collection: <target collection name>
    options:
      write_mode: upsert | append
      ddl: apply | ignore | fail
      auto_create_table: true   # default: true

push Output (Event Stream)

Publish data as an event stream (Kafka, etc.):

push:
  - source: orders
    topic: order-events          # omitted = uses table name
    format: cyntex               # default: cyntex envelope; custom = CEL projection
    options:
      start_from: latest

---

Expressions: CEL

Fields such as filter and format use CEL (Common Expression Language):

filter: "record.amount > 100 && record.currency == 'USD'"
format: "{'id': record.user_id, 'ts': record.created_at}"

• Compiled + type-checked at validate time (not evaluated)
• Evaluated by the engine at runtime
• JS nodes (js: field) are the escape hatch beyond CEL, supporting complex logic

---

id Rules

• Top-level id: globally unique within the workspace; must not contain .
• Inline private resources: <pipeline_id>.<local_id> (auto-naming for inline transforms/views)
• Reference: a string value = reference to the resource with the corresponding id

---

Version Evolution

apiVersion: cyntex/v1

• Strict rejection of unknown fields (additionalProperties: false): any typo causes an immediate error
• Experimental zone: experimental fields go under the experimental: block with no compatibility guarantee

---

validate — Three Layers

cyntex validate <file.cyn.yml> runs three layers of validation:

1. Structural layer: JSON Schema validation (unknown fields, types, required fields)
2. Reference layer: id reference closure within this batch of resources (offline; no server connection needed)
3. Capability matrix layer: connector × mode validity + required config fields (bundled catalog)

Error format: file path / line number / field path + plain-language explanation.