Vega — research subscription
AF

Vega Research

Stale

northwind.onmicrosoft.com / sub-2290-vega · Azure AI Foundry · sub-2290-vega · eastus

Vega Research Copilot

Stages complete

3 / 3

Records read

4,120

Elapsed

18m 26s

Cost of discovery

$0.00

What stage A actually read

Shown before anything downstream commits to it

LOGGED
Records read
4,120 of 41,200
Window
2026-06-20 → 2026-07-25
Input tokens
~8.2M
Output tokens
~494k

A deliberate statistical sample, taken by hash of request id — not the first N rows. Cost figures are not sampled; they come from the billing export.

Models seen

  • gpt-4.125,544
  • gpt-4.1-mini11,124
  • gpt-4.1-ft-vega014,532

billing $31,800 LOGGED

Stages A to C

Each stage shows what it was handed as well as what it produced — the input is usually the part nobody can see.

  1. Input

    Source
    Azure AI Foundry
    Account
    sub-2290-vega
    Window
    2026-06-20 → 2026-07-25
    Records available
    41,200

    Output

    Records read
    4,120LOGGED
    Redaction
    managed
    Residual identifiers
    none found

    • A.1Data-loss-prevention availability checkDECISION

    • A.2Read model-provider logsAUTO

    • A.3Read billing exportAUTO

    • A.4Redact promptsAUTO

    • A.5Deep residual personal-data scanAUTO

    • A deliberate statistical sample, not the first N rows.
discovery outputarchived
[A]ingest  connector=vega-research  account=sub-2290-vega  region=eastus
[A][A.2] read 4,120 of 41,200 logged calls
[A]        PARTIAL CORPUS — sampled by hash of request_id
[A]        cost figures are NOT sampled — they come from the billing export
[A][A.3] billing export → 2026-06-20 → 2026-07-25
[B][B.1] cluster + measure → 7 workloads across 4 applications
[B][GATE.approved] curation approved
[B][B.3] coverage 58% → continue
[C][C.1] rules fired on 7 workloads; each rule cites its measurement
[C][C.4] tiers and confidence assigned
[C][done] snapshot snap-2026-08-27-9d04 sealed