Gtm api

Google Tag Manager API

Execute GTM API operations with validation, error handling, and proper workflow patterns.

Quick Start

1. Understand the Hierarchy

Account
└── Container
    ├── Environments (deployment targets)
    ├── Versions (immutable snapshots)
    └── Workspaces (mutable, where changes happen)
        ├── Tags
        ├── Triggers
        └── Variables

Critical rule: All changes happen in workspaces; publishing creates versions.

2. Four Steps for Every Operation

1. Read algorithm → instructions.md
2. Copy template → request-templates.md
3. Validate → validation-rules.md
4. Execute with error handling → workflows.md

---

Core Execution Workflow

Step 1: Validate Inputs

BEFORE any API call:
  1. Validate required fields present
  2. Validate field types (strings, arrays, etc.)
  3. Verify entity references exist (trigger IDs, variable names)
  4. Check OAuth scopes sufficient
  5. Confirm container type supports operation

Step 2: Get or Create Workspace

workspaces = GET /accounts/{account_id}/containers/{container_id}/workspaces

IF workspaces is empty:
  workspace = POST /workspaces with {"name": "Default Workspace"}
ELSE:
  workspace = workspaces[0]

Store: workspace_id, workspace_path

Step 3: Execute Operation

Use the appropriate reference file:

• Creating entities → request-templates.md
• Step-by-step algorithms → instructions.md
• Complex flows → workflows.md

Step 4: Handle Response

IF status 200/201: Extract IDs, return success
IF status 400: Check JSON syntax, validate fields
IF status 401: Refresh OAuth token
IF status 403: Check scopes OR rate limit (backoff)
IF status 404: Verify resource path/ID
IF status 409: Get fresh fingerprint, retry (max 3)
IF status 429: Exponential backoff (1s → 2s → 4s → 8s → 16s → 32s)
IF status 500: Retry with backoff

---

Critical Rules

Fingerprints

CREATE:  Don't include fingerprint
UPDATE:  MUST include current fingerprint (GET first)
DELETE:  Not needed

IDs vs Paths

API endpoints:     Use full path
                   "accounts/123/containers/456/workspaces/10/tags/5"

Entity references: Use ID only
                   firingTriggerId: ["5"]

Boolean Parameters

WRONG:  {"type": "boolean", "value": true}
RIGHT:  {"type": "boolean", "value": "true"}  // String!

Rate Limits

10,000 requests/day
0.25 QPS (1 request per 4 seconds)
On limit: Exponential backoff

---

Common Operations

Create a Tag

1. GET workspace
2. Verify triggers exist (GET /triggers/{id})
3. Build tag body with:
   - name (required)
   - type (required)
   - firingTriggerId (required, array of IDs)
   - parameter (array of typed params)
4. POST {workspace_path}/tags
5. Return tag_id

Template: request-templates.md → "Tag Templates"

Update a Tag

1. GET {tag_path} → extract fingerprint
2. Merge updates with current state
3. Include fingerprint in body
4. PUT {tag_path}
5. On 409: Get fresh fingerprint, retry

Publish Changes

1. GET {workspace_path}/status
   - Check for conflicts (resolve first)
   - Check for changes (nothing to publish if empty)
2. POST {workspace_path}:create_version with name/notes
3. POST {version_path}:publish
4. Return version_id

Algorithm: instructions.md → "Instruction: Publish Changes"

List Resources with Pagination

all_items = []
page_token = None

LOOP:
  response = GET {path}/{resource_type}?pageToken={page_token}
  all_items.extend(response.items)
  page_token = response.nextPageToken
  IF page_token is None: BREAK

RETURN all_items

---

OAuth Scopes

Action	Scope
Read anything	tagmanager.readonly
Create/edit tags	tagmanager.edit.containers
Publish	tagmanager.publish
Delete containers	tagmanager.delete.containers
Manage users	tagmanager.manage.users

Scopes are additive. Publishing requires: readonly + edit.containers + publish

---

Path Patterns

Resource	Path
Container	accounts/{id}/containers/{id}
Workspace	.../containers/{id}/workspaces/{id}
Tag	.../workspaces/{id}/tags/{id}
Trigger	.../workspaces/{id}/triggers/{id}
Variable	.../workspaces/{id}/variables/{id}
Version	.../containers/{id}/versions/{id}

Special paths:

• /versions/live - Currently published version
• /workspaces/{id}:create_version - Create version
• /versions/{id}:publish - Publish version

---

Tag Parameter Structure

Parameters use typed objects:

{
  "parameter": [
    {"type": "template", "key": "measurementId", "value": "G-XXXXX"},
    {"type": "boolean", "key": "sendPageView", "value": "true"},
    {"type": "list", "key": "eventParameters", "list": [...]}
  ]
}

Types: template (string), boolean (string "true"/"false"), integer, list, map

---

Reference Documentation

Execution (When Doing)

File	Use When
instructions.md	Step-by-step algorithms for operations
request-templates.md	Copy-paste JSON templates
validation-rules.md	Validating before sending
workflows.md	Complex flows, decision trees

Understanding (When Learning)

File	Use When
context.md	Mental model, architecture
schemas.md	Complete entity structures
api-reference.md	All endpoints, methods
examples.md	Real-world request/response pairs
quick-reference.md	Lookup tables, type codes

---

Common Mistakes to Avoid

Including auto-generated fields in CREATE

WRONG: {"tagId": "5", "path": "...", "name": "My Tag"}
RIGHT: {"name": "My Tag", "type": "html", "firingTriggerId": ["5"]}

Missing fingerprint in UPDATE

WRONG: PUT /tags/5 with {"name": "Updated"}
RIGHT: GET /tags/5 first, then PUT with current fingerprint

Using paths in entity references

WRONG: {"firingTriggerId": ["accounts/123/.../triggers/5"]}
RIGHT: {"firingTriggerId": ["5"]}

---

Execution Template

When given a GTM API task:

1. IDENTIFY operation type (create/update/delete/list/publish)
2. FIND algorithm in instructions.md
3. GET template from request-templates.md (if creating/updating)
4. VALIDATE using validation-rules.md
5. EXECUTE with proper auth headers
6. HANDLE errors using workflow from workflows.md
7. RETURN result with IDs or error with fix suggestion