ds_41_features.py

"""VynFi DataSynth 4.1.x — exercise the new config subsections end-to-end.

DS 4.1.x / VynFi API 4.1.x added seven new config surfaces to
``POST /v1/generate``. This example wires them all into a single
generation job, waits for completion, and inspects the archive to confirm
that each feature produced observable output.

Sections exercised:

- ``analyticsMetadata`` — prior-year benchmarks + industry peers +
  management reports + drift events
- ``audit`` — workpapers + IT controls + review delay model
- ``complianceRegulations`` — legal documents + opinions
- ``accountingStandards`` — IFRS 16 leases, fair value, framework
  differences
- ``interconnectivity`` — vendor tiers / customer segments / industry
  metadata (Scale+)
- ``templates.packId`` — link a user-uploaded template pack (optional;
  commented out because it needs a pre-created pack)
- ``llm.enrichmentEnabled`` — switch on LLM-backed enrichments in the
  generation pipeline

Tier notes:
- ``interconnectivity.*`` requires Scale+.
- ``complianceRegulations.legalDocumentsEnabled`` requires Team+.
- ``llm.enrichmentEnabled`` requires Scale+ and a configured portal LLM.
"""

from __future__ import annotations

import os

import vynfi

client = vynfi.VynFi(api_key=os.environ["VYNFI_API_KEY"], timeout=120.0)

config = {
    "sector": "manufacturing",
    "country": "CH",
    "accountingFramework": "ifrs",
    "rows": 400,
    "companies": 2,
    "periods": 2,
    "periodLength": "quarterly",
    "processModels": ["o2c", "p2p", "h2r", "audit"],
    "exportFormat": "json",
    "fraudPacks": ["vendor_kickback"],
    "fraudRate": 0.03,
    # ── DS 4.1: analytics metadata ──────────────────────────────────
    "analyticsMetadata": {
        "enabled": True,
        "priorYear": True,
        "industryBenchmark": "manufacturing_ch_midsize",
        "managementReports": ["cash_flow", "opex_by_function"],
        # Synthetic drift events let anomaly detectors learn to ignore them.
        "driftEvents": [
            {"period": 2, "type": "cost_inflation", "magnitude": 0.08},
        ],
    },
    # ── DS 4.1: audit workpapers + IT controls ──────────────────────
    "audit": {
        "enabled": True,
        "generateWorkpapers": True,
        "itControls": True,
        "minTeamSize": 3,
        "maxTeamSize": 6,
        "workpapersAveragePerPhase": 12,
        "averageReviewDelayDays": 2,
    },
    # ── DS 4.1: compliance regulations (Team+) ─────────────────────
    "complianceRegulations": {
        "legalDocumentsEnabled": True,
        "legalOpinionProbability": 0.15,
    },
    # ── DS 4.1: accounting standards ───────────────────────────────
    "accountingStandards": {
        "enabled": True,
        "framework": "ifrs",
        "leasesEnabled": True,  # IFRS 16
        "fairValueEnabled": True,  # IFRS 13
        "generateDifferences": True,  # US GAAP ↔ IFRS reconciliation
    },
    # ── DS 4.1: interconnectivity (Scale+) ─────────────────────────
    "interconnectivity": {
        "vendorNetworkEnabled": True,
        "customerSegmentationEnabled": True,
        "industrySpecificEnabled": True,
    },
    # ── DS 4.1: templates (needs a pre-uploaded pack) ──────────────
    # "templates": {"packId": "<uuid of a pack you own>"},
    # ── DS 4.1: LLM enrichment (Scale+) ────────────────────────────
    "llm": {
        "enrichmentEnabled": True,
        "modelId": "anthropic/claude-sonnet-4.5",
    },
}

print("Submitting DS 4.1 feature-matrix job …")
job = client.jobs.generate_config(config=config)
print(f"  job {job.id}")
done = client.jobs.wait(job.id, timeout=420)
print(f"  status {done.status}")
if done.status != "completed":
    raise SystemExit(done.error_detail or "Generation did not complete.")

# ── Group the archive paths by top-level directory to see what landed ───────

files = client.jobs.list_files(job.id)
grouped: dict[str, list[str]] = {}
for f in files.files:
    top, sep, rest = f.path.partition("/")
    group = top if sep else "(root)"
    grouped.setdefault(group, []).append(rest if sep else top)

print(f"\nArchive contents ({files.total_files} files):")
for group, entries in sorted(grouped.items()):
    label = "(root)" if group == "(root)" else f"{group}/"
    print(f"  {label} ({len(entries)} files)")
    for entry in sorted(entries)[:4]:
        print(f"    {entry}")
    if len(entries) > 4:
        print(f"    … and {len(entries) - 4} more")

# ── Bonus: use the aggregated audit-artifacts endpoint ────────────────────

try:
    art = client.jobs.audit_artifacts(job.id)
    n_op = len(art.audit_opinions or [])
    n_kam = len(art.key_audit_matters or [])
    print(f"\naudit_artifacts(): {n_op} opinions, {n_kam} KAMs")
except vynfi.NotFoundError:
    print(
        "\naudit_artifacts(): no audit/*.json materialized on this run — "
        "the audit pipeline is best-effort on DS 4.1.x and only emits "
        "when findings warrant."
    )