create-ticket

Create Ticket

Create a ticket — persistent conversation/work thread with state and assignee.

When to use

Use tickets when:

• The topic will come up again (customer retention, recurring bug, ongoing partnership)
• The work needs an assignee and status workflow
• A heartbeat should pull it from an inbox (tickets are the inbox)
• Multiple people / agents will discuss over time (threaded comments)

Don't use tickets for:

• Ephemeral questions (use a chat session)
• Deterministic scheduled work (use a routine)
• State-checking protocols (use a heartbeat + decision prompt)

Step 1: Collect fields

Ask the user:

1. Title — short, specific ("Cliente X reclama de latência")
2. Description — context (optional but recommended)
3. Priority — urgent, high, medium (default), low
4. Assignee agent — which agent handles this? Common picks:
5. zara-cs — support / CS
6. flux-finance — billing / payments
7. atlas-project — project blockers
8. aria-hr — HR / hiring
9. lex-legal — contracts / compliance
10. Goal link? Optional goal_id if this work moves a specific goal. Skip if not.
11. Project link? Optional project_id for grouping without a specific goal.

Step 2: Call the API

Use from dashboard.backend.sdk_client import evo — auto-handles URL + auth.

The runtime injects the current agent slug and session id into your system prompt — pass them through so the ticket records provenance.

import json
from dashboard.backend.sdk_client import evo
ticket = evo.post("/api/tickets", {
    "title": "Cliente X reclama de latência",
    "description": "Print anexado no Intercom, 4s para carregar /dashboard",
    "priority": "high",
    "assignee_agent": "zara-cs",
    "goal_id": 3,
    "source_agent": "<agent-slug>",         # injected via system prompt at runtime
    "source_session_id": "<session-uuid>",  # injected via system prompt at runtime
})
print(json.dumps(ticket))   # use json.dumps, NOT print(ticket) — the UI auto-binds the session to the ticket when it sees a valid JSON ticket in stdout

Response includes the created ticket with id, status=open, created_at.

Step 3: Show what happens next

Explain to the user:

1. Ticket appears in /issues?assignee=zara-cs inbox
2. Zara's next heartbeat (step 3 — query inbox) will see it
3. If priority is urgent, it jumps to front of queue
4. When Zara acts, the ticket gets:
5. locked_at / locked_by = zara-cs (atomic checkout)
6. Comments added by Zara showing what was done
7. Status eventually changes (in_progress → review → resolved → closed)
8. Auto-release after lock_timeout_seconds (default 1800s) if Zara crashes mid-work

Step 4: Mentions (optional)

If the user says "and tell Flux to check billing", add a comment with:

from dashboard.backend.sdk_client import evo
evo.post(f"/api/tickets/{ticket['id']}/comments", {
    "body": "@flux-finance please confirm this customer's billing status",
})

The @flux-finance is parsed and fires a wake trigger — Flux's heartbeat wakes (within 30s debounce window) and picks up the ticket.

Step 5: Direct user to UI

• /issues — global list with filters and search
• /tickets/<id> — detail view with full timeline (comments + activity + status changes)
• Bulk actions in /issues: close, reopen, delete, reassign, relink_goal

Notes

• Tickets vs sessions: tickets persist (days/weeks), sessions are ephemeral. A session can bind to a ticket.
• Closing a ticket doesn't delete comments or activity — the timeline remains.
• Priority order: urgent > high > medium > low (enum with internal rank).

Related: .claude/rules/tickets.md, .claude/rules/heartbeats.md, .claude/rules/goals.md.