# n8n Workflows

Exported n8n workflow definitions stored as version-controlled JSON files.

## Workflows

### 001 · Zendesk — Evaluation — Change Requests

Runs twice daily at 08:00 and 14:00 (Europe/Helsinki). Scans six Zendesk support groups for unsolved tickets created in the past 24 hours, uses an AI agent to identify ITIL 4 change requests, notifies via Telegram, and logs confirmed findings to a BookStack wiki page.

#### Flow

```
Schedule Trigger (cron 0 8,14 * * *)
  │
  ├─ Get IT-Tukipalvelut          ─┐
  ├─ Get IT-Change Control         │
  ├─ Get Järjestelmähankintaryhmä  ├─ Merge results (6-input)
  ├─ Get Turvatekniikka            │
  ├─ Get Sovelluspaketointi        │
  └─ Get IT-Hankinnat             ─┘
                │
        Aggregate the results  (formats tickets as JSON)
                │
           AI Agent  (Claude Haiku 4.5 via Anthropic)
            │       │
            │   Report potential change requests  →  Telegram (immediate notification)
            │
      Code - Parse Tickets  (regex-parses structured AI output)
            │
      Code - Aggregate Entries  (builds HTML blocks)
            │
      Get current content  (BookStack page #46)
            │
      Code - Merge Results  (prepends new entries to existing HTML)
            │
      Send a text message  (Telegram approval request — double confirmation)
            │
      Update Approved?  (if approved == true)
            │
      Update page contents  (BookStack page #46)
```

#### Required credentials

| Credential name | Type | Used by |
|---|---|---|
| `Zendesk account` | Zendesk API | All six Zendesk fetch nodes |
| `Anthropic account` | Anthropic API | AI Agent (Claude Haiku 4.5) |
| `Telegram account` | Telegram Bot API | Immediate report + approval request |
| `BookStack account` | BookStack API | Get and update page #46 |

> The `Azure OpenAI` node exists in the canvas but is not connected to the AI Agent and is not part of the active execution path.

#### AI output contract

The AI agent must begin every response with `CR report -` and use exactly these field labels for each finding:

```
Ticket ID: <id>
Ticket URL: <url>
Subject: <text>
Short reason: <text>
```

The `Code - Parse Tickets` node uses a regex that matches this exact format. Changes to the system prompt must preserve these labels and order.

## Repository conventions

- Naming: `NNN - <Integration> - <Category> - <Description>.json`
- Keep the `"name"` field inside the JSON in sync with the filename.
- `"connections"` references node names, not IDs — update both when renaming a node.
- Do not modify `"credentials"` blocks; credential IDs and names must match the target n8n instance.

## Importing a workflow

1. Open your n8n instance.
2. Go to **Workflows → Import from file**.
3. Select the JSON file.
4. Re-link any credentials that differ from the source instance before activating.