Build a dashboard briefing

Turn a dashboard into a Markdown brief: each forecast's latest number and how much it moved since the previous run. These are the same dashboards you see in the Mantic web app.

Full script: dashboard_brief.py.

1. Fetch the dashboard's forecasts

python dashboard_brief.py --list prints your dashboard IDs. iter_dashboard_questions then reads every page of the dashboard's forecasts, each with its latest run:

return paginate(
    client,
    f"/dashboards/{dashboard_id}/questions",
    "questions",
    {"show_forecasts": "true"},
)

2. Turn the latest run into a headline

latest_forecast is null until the forecast's first run completes, so check that first:

if forecast is None:
    return "First run pending"

The headline function then branches on the run's type. For an event run, the probability becomes a percentage:

return f"{forecast['probability']:.0%}"

For example, 0.64 becomes 64%. The full function handles every type; see Read a forecast's latest run.

3. Find the previous run

The dashboard gives only the latest run. Request the forecast's history with GET /questions/{question_id}/forecasts, newest first:

history = paginate(
    client,
    f"/questions/{question_id}/forecasts",
    "forecasts",
    {"show_forecasts": "true"},
)

A new run could complete between the dashboard request and the history request. Find the run you displayed by its id, then take the next, older one:

for forecast in history:
    if forecast["id"] == latest["id"]:
        return [forecast, *islice(history, 1)]
return []

4. Calculate movement

For two runs of an event forecast, latest_change subtracts the previous run's probability and converts the difference to percentage points (pp):

points = 100 * (latest["probability"] - previous["probability"])
return f"{points:+.1f}pp"

From 0.57 to 0.64, this returns +7.0pp. For other types, see Measure the change.

Connect these functions for each forecast on the dashboard:

latest = question["latest_forecast"]
change = "Change unavailable"
if latest is not None and latest["type"] not in {"redacted", "unknown"}:
    pair = previous_forecast(client, question["id"], latest)
    change = latest_change(pair) if pair else "Displayed run missing from history"

5. Assemble the Markdown entry

render_brief starts with the dashboard title. For each forecast, it adds a linked heading, the two values you just calculated, then the run's completion time and ID:

lines.extend(
    [
        (
            f"## [{escape_link_label(question['title'])}]"
            f"(https://app.mantic.com/question/{question['id']})"
        ),
        "",
        f"**{headline(latest)}** · {change}",
        "",
    ]
)

Run and inspect the brief

List your dashboards, then replace YOUR_DASHBOARD_ID with one of the returned IDs:

python dashboard_brief.py --list
python dashboard_brief.py YOUR_DASHBOARD_ID > dashboard-brief.md

Open dashboard-brief.md. Example output:

# UK macro outlook

## [Will UK CPI inflation be below 2.5% in December 2026?](https://app.mantic.com/question/123)

**64%** · +7.0pp

Run completed: `2026-09-21T08:30:00Z` · Run ID: `987`