Agent: dashboard-manager

Purpose

Create and manage Corezoid dashboards that visualize task counters from running processes. Dashboards show how many tasks are in specific nodes — useful for monitoring process health and business metrics.

When to invoke

• User asks to "create a dashboard" or "visualize" process data
• User wants to monitor how many tasks succeed/fail in a process
• User wants to track business metrics from Corezoid processes
• After deploying a new process, to add observability

Required inputs

Parameter	Description
API_LOGIN, SECRET, BASE_URL, COMPANY_ID	Auth credentials
Process ID(s)	Which process(es) to measure
Metric goals	What the user wants to track (completion rate, error count, etc.)

MCP tools used

get_process_scheme    — find node IDs to measure
create_dashboard      — create dashboard container
add_chart             — add visualization charts
get_dashboard         — verify chart series after creation

Workflow

Step 1 — Identify nodes to measure

result = client.get_process_scheme(process_id)
nodes = result['ops'][0]['scheme'][0]['scheme']['nodes']

# End nodes (obj_type: 2) accumulate tasks — best for measurement
end_nodes = [n for n in nodes if n['obj_type'] == 2]
# Note: intermediate nodes pass tasks through instantly — measuring them shows 0 in real-time

Step 2 — Create dashboard

result = client.make_request({'ops': [{
    'obj': 'dashboard',
    'obj_type': 'create',
    'title': 'Weather Connector Monitoring',
    'description': 'Task success and error rates'
}]})
dashboard_id = result['ops'][0]['obj_id']

Step 3 — Add charts

result = client.make_request({'ops': [{
    'obj': 'chart',
    'obj_type': 'create',
    'dashboard_id': dashboard_id,
    'title': 'Success vs Error Tasks',
    'chart_type': 'column',   # ALWAYS 'column', NEVER 'bar'
    'metrics': [
        {'conv_id': process_id, 'node_id': success_node_id, 'title': 'Completed'},
        {'conv_id': process_id, 'node_id': error_node_id, 'title': 'Errors'}
    ]
}]})
chart_id = result['ops'][0]['obj_id']

Step 4 — Verify series

After creating, ALWAYS verify that series is populated:

verify = client.make_request({'ops': [{'obj': 'chart', 'obj_type': 'show', 'id': chart_id}]})
series = verify['ops'][0].get('series', [])
if not series:
    # Re-create chart — metrics didn't attach correctly
    pass

Chart types

Type	When to use
column	Compare values — success vs error, node A vs node B
pie	Show proportions — what % of tasks succeed
funnel	Show drop-off through sequential steps
table	Tabular view of multiple metrics

bar does NOT exist — will error silently.

Modifying charts — full payload required

When modifying a chart, ALWAYS include obj_type and the complete series array. Partial modify returns validation error.

client.make_request({'ops': [{
    'obj': 'chart',
    'obj_type': 'modify',
    'id': chart_id,
    'title': 'Updated Title',
    'chart_type': 'column',
    'series': [
        {'conv_id': process_id, 'node_id': node_id, 'title': 'Metric Label'}
    ]
}]})

Grid layout

Use width/height fields (NOT w/h) — wrong field names cause validation errors.

{'width': 6, 'height': 4}

Real-time monitoring

Real-time mode works only for these node types:

• End nodes (obj_type: 2) — tasks finished the process
• Waiting for Callback — tasks waiting for external callback
• Delay nodes — tasks paused for a time period

Intermediate nodes (Code, API Call, Condition) pass tasks through instantly — they always show 0 in real-time.

Knowledge references

• knowledge/gotchas.md — #15: dashboard chart type is column not bar
• knowledge/auth.md — credentials and batching
• docs/dashboards/README.md — full dashboard documentation