Skip to content

[spending-forecast] Daily spending forecast - 2026-11-08 #75882

Closed as not planned

Description

@github-actions

Warning

Threat detection engine error

Threat detection could not complete its analysis due to a tooling error. This is not a security finding.

What happened

The threat detection engine failed to produce results.

Review the workflow run logs to troubleshoot the error.

Overview

The daily spending forecast is unavailable for 2026-11-08 UTC. Both forecast attempts produced zero-byte output, and the authorized run-artifact fallback was rate-limited before it could list runs. Total observed AIC is therefore unavailable, not 0; no reliable budget can be inferred.

  • Weekly forecast: P10 (10th percentile — optimistic scenario: 9 out of 10 months will cost at least this much), P50 (50th percentile — median/expected scenario: equal probability of spending more or less than this), and P90 (90th percentile — conservative scenario: only 1 out of 10 months is expected to exceed this amount) are all unavailable.
  • Monthly forecast: P10, P50, and P90 are all unavailable.
  • Requested history window: 30 days, 2026-10-09 10:05:03 UTC through 2026-11-08 10:05:03 UTC.
  • Evidence recovered: 0 workflow records, 0 run samples, and 0 AIC observations.

Caution

Missing observations are not evidence of zero spending. This report must not be used to set or validate a budget.

Charts

The PNG inputs could not be produced from an empty, invalid forecast, so the required compact ASCII fallbacks are shown instead.

Spending Trend — Last 30 Days

AIC/run
  N/A |............... no observations ................|
      +------------------------------------------------+
      2026-10-09                              2026-11-08
Series: unavailable; 7-day rolling average: unavailable

Weekly Forecast Distribution

In this section, P10 (10th percentile — optimistic scenario: 9 out of 10 months will cost at least this much), P50 (50th percentile — median/expected scenario: equal probability of spending more or less than this), and P90 (90th percentile — conservative scenario: only 1 out of 10 months is expected to exceed this amount) are unavailable.

Workflow scope                  P10        P50        P90
All workflows                   N/A        N/A        N/A
Scale: no projection distribution was produced

Key metrics and issues

For this section, P50 (50th percentile — median/expected scenario: equal probability of spending more or less than this) and P90 (90th percentile — conservative scenario: only 1 out of 10 months is expected to exceed this amount) are unavailable rather than zero.

Workflow Available samples Observed AIC P50/run P90/run Weekly projected AIC Monthly projected AIC Success rate Confidence range
All workflows — discovery failed 0* Unavailable Unavailable Unavailable Unavailable Unavailable Unavailable Unavailable

*Zero records were available to analyze; this does not mean zero runs occurred.

The checkout contains 362 compiled workflow definitions, while the forecast returned no workflow objects. That local count is not proof that all 362 are active remotely, but it makes an empty fleet forecast implausible and leaves workflow completeness unresolved.

Data quality and accuracy

In this section, P10 (10th percentile — optimistic scenario: 9 out of 10 months will cost at least this much), P50 (50th percentile — median/expected scenario: equal probability of spending more or less than this), and P90 (90th percentile — conservative scenario: only 1 out of 10 months is expected to exceed this amount) refer to the absent Monte Carlo bounds.

Check Finding Likely forecast impact Follow-up result
Parseability forecast.json is zero bytes; metadata records exit code 1. No totals or per-workflow fields can be reconciled. Unresolved.
Targeted rerun A 30-day, 100-run-per-workflow monthly JSON rerun failed during workflow discovery because the local GitHub CLI was unauthenticated. No replacement JSON was produced. Alternate credentials were not attempted.
Authorized artifact fallback agentic-workflows.logs requested info and usage evidence for the exact 30-day window, capped at 3,000 runs. Its GitHub installation was rate-limited before run listing. Observed AIC could not be derived independently. Unresolved.
Missing or zero AIC No AIC values exist in the available evidence, so missing values cannot be separated from genuine zero-AIC runs. Observed spending and all projections are unavailable. Unresolved; no zero values were imputed.
Sample size and freshness No run dates or samples were returned. Sparse, stale, and per-workflow reliability checks cannot be performed. Unresolved.
Outliers No run-level distribution exists. High-cost runs could materially change the budget without appearing here. Unresolved.
Date-window consistency The requested window is known, but no sample dates exist to test inclusion or coverage. The effective history window is unknown. Unresolved.
Run frequency and success rate No completed-run counts or conclusions exist. Schedule-driven and event-driven frequencies, plus success rates, cannot be validated. Unresolved.
Confidence intervals No Monte Carlo output exists. Interval width and budget reliability cannot be assessed. Unresolved.
Backtesting --eval was not run because collection failed before any training or validation sample existed. Forecast calibration is unknown. Deferred until usable samples exist.
Chart compatibility Current forecast samples expose date and are nested by workflow, while the trend template expects flat samples with run_started_at and workflow_id. Current projections use weekly_monte_carlo.*_projected_aic, while the distribution template reads projection.weekly_aic_*. Even valid current-schema data would fail the trend chart or render misleading zero forecast bars without normalization. Unresolved; the ASCII fallback avoids fabrication.
Collection and reconciliation evidence
  • Initial collection: generated at 2026-11-08 10:05:03 UTC, exit code 1, zero-byte forecast.json.
  • Targeted recovery: gh-aw forecast --repo github/gh-aw --days 30 --sample 100 --period month --timeout 10 --json; exit code 1 and zero-byte forecast-rerun.json.
  • MCP recovery: requested recent info and usage artifacts over the same 30-day window; failed at run listing due to the installation API rate limit.
  • Workflow totals: unavailable because no workflow objects were emitted.
  • Sampled-run counts: unavailable; there are zero available sample records, not an observed count of zero executions.
  • Historical spending: cannot be calculated from run_samples[].aic because run_samples is absent.
  • History-window coverage, success rates, per-run percentiles, weekly/monthly projections, and reliability flags: all unavailable.
  • Structured recovery diagnostics are preserved alongside the forecast artifact as recovery-diagnostics.json.
Assumptions and scope
  • Repository: github/gh-aw.
  • Forecast date: 2026-11-08 UTC.
  • Intended history: the 30 days ending at the initial collection timestamp.
  • Intended sample cap: 100 completed runs per workflow.
  • No AIC, run count, frequency, percentile, success-rate, or workflow-activity value was inferred from filenames, schedules, or missing output.
  • A projection should be treated as budget-supporting only after each material workflow has at least 10 non-zero AIC observations, the forecaster marks its interval reliable, and backtesting supports calibration.

Next actions

  1. Retry authenticated forecast and usage-artifact collection after the GitHub installation rate limit recovers; retain the resulting JSON even when individual workflows fail.
  2. Make collection failures emit a structured JSON error envelope so an exit code cannot leave an ambiguous zero-byte artifact.
  3. Normalize chart inputs: add workflow_id to each flattened run sample, use the sample date field, and read the 10th-, 50th-, and 90th-percentile values from weekly_monte_carlo.
  4. Once usable samples exist, run --eval, compare actual AIC with the central 80% forecast interval, and reject budget guidance for workflows with fewer than 10 non-zero AIC observations or materially broad intervals.

Reference: §42467952498

Generated by 📈 Daily Spending Forecast · codex · routed: gpt56 max · ~116 AIC · ⊞ 347 · ◷

  • expires on Nov 15, 2026, 2:28 AM UTC-08:00

Activity

  1. github-actions commented on Nov 10, 2026

    @github-actions
    ContributorAuthor

    This issue is being closed as outdated. A newer issue has been created: #76475

    View newer issue


    This action was performed automatically by the Daily Spending Forecast workflow.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions