BI-P014's security requirement is the whole prompt: DuckDB is a local engine
with `read_csv('/etc/passwd')` built in. Two layers.
LAYER 1, the mount registry: a caller names ARTIFACT IDS, never paths. The
registry resolves an id, checks the artifact belongs to the caller's tenant,
and realpath()s the result to confirm it is still inside the mount root once
symlinks are followed. A symlink to /etc/passwd, a traversing relative path, an
absolute path, and a sibling directory sharing the root's prefix
(/mnt/artifacts-evil for root /mnt/artifacts) are each refused. A cross-tenant
read is AUTHORIZATION_DENIED, deliberately not NOT_FOUND — an operator reading
the audit trail needs to see an attempted cross-tenant read as one.
LAYER 2, THE FINDING:
`SET allowed_directories=[...]` ALONE IS A NO-OP.
Measured on DuckDB 1.5.5. With allowed_directories set to a single mount
directory and nothing else, `read_csv('/etc/passwd')` RETURNED THE FILE and an
https:// URL was fetched over the network. Only when `enable_external_access
= false` is ALSO set does the allow-list bind; the same reads then fail with
"Permission Error: Cannot access file". `lock_configuration=true` afterwards
stops a query re-opening either.
So the obvious spelling of this control — set the allow-list and move on —
yields a setting that appears in duckdb_settings(), looks like a sandbox, and
permits everything. That is `trivy --exit-code 0` in a different costume. The
bootstrap is therefore emitted as ONE unit, `assertSessionSandboxed` refuses a
session missing any part of it, and the measurement itself is PINNED by a test
that never skips: if someone simplifies the bootstrap, the pinned evidence
(`read_etc_passwd_succeeded: true` for the naive spelling) keeps the finding
attached to the build rather than to a comment that would rot.
Twelve bypass attempts run against a real DuckDB with LAYER 1 REMOVED — sent
straight to the engine, registry not in the picture. Zero succeeded. The
counterfactual runs in the same session and is asserted: the mounted file IS
readable, so "nothing succeeded" is not a sandbox that blocks everything, which
would be an outage rather than a control. A locked session is then shown unable
to re-open its own sandbox, all three ways.
DoD, against real fixtures: Parquet aggregates correctly (SE 394.5, NO 105.5),
CSV round-trips a value containing a quote, and an ICEBERG SNAPSHOT reads via
iceberg_scan on an exact metadata JSON — which is not a workaround: DuckDB
refuses to glob for the latest version because globbing can read uncommitted
data, and an exact metadata file is what "snapshot" means. Arrow is reported
UNPROVEN by `readiness()` with the reason: DuckDB's arrow extension is a
network install the offline environment could not download.
THREE DEFECTS IN MY OWN TESTS, found and fixed rather than tidied away:
* a bypass-corpus entry was `SELECT 1 FROM (SELECT 1) t` — not an attack at
all. It succeeded and was reported as a leak. A corpus entry that cannot
fail is as useless as a check that cannot. Replaced with real ATTACH and
COPY attempts.
* the harness ran the session bootstrap outside its try block, so a statement
that correctly REFUSED to re-open a locked sandbox killed the subprocess
and surfaced as an unreadable "Command failed" instead of as the meaningful
refusal it was.
* the Iceberg fallback was `expect(evidence.iceberg ?? "unbuilt").toBeDefined()`
— a check that can never fail. It let a run without the fixture ERASE the
recorded Iceberg result from the evidence file while both suites stayed
green. It now rebuilds the fixture when pyiceberg is available and, when it
is not, asserts against the previously recorded result and marks it as
carried forward.
The Iceberg fixture is deliberately NOT committed: pyiceberg writes absolute
paths into its metadata and Avro manifests, so a committed table resolves only
on the machine that built it, and a fixture that silently fails everywhere else
is worse than none. tests/fixtures/duckdb/README.md says so and gives the
rebuild command; the recorded result is the artefact instead.
Heavy artifacts are refused with CAPABILITY_UNAVAILABLE and pointed at N0D,
per BI-P014 — silently running a 40 GB scan in the API process is how a preview
engine takes down a control plane.
35 unit tests, 12 live tests. Evidence: evidence/bi-p014/duckdb-sandbox.json.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
0pt — your data
Data services on demand for the Nordics + Europe. Compose a data job, pay 50 % to start, watch production live, validate with full provenance, deliver anywhere. Owned by SIAX Technology AB. Honest accuracy is the moat: every field carries source, freshness and confidence; nothing is guessed.
Live: https://0pt.app · Contract: CLAUDE.md + docs/
Architecture (verified against code, not aspirational):
docs/architecture/ARCHITECTURE.md. It documents the real API surface, the real persistence story, and the known security gaps — including several places where this README is optimistic. Read it before trusting any capability claim below. Repo identity + estate classification:siax.repo.json+siax.config.json. Compliance state:docs/compliance/(INVENTORY, RELATIONS, OSS-INVENTORY, GOLDEN-STANDARD-REPORT, BACKLOG, SLUTRAPPORT).
Two planes (see docs/02-architecture.md)
- Edge (Vercel, this app) — marketing, catalog, the order lifecycle, customer portal,
admin portal, and the API route handlers. Defers to the data engine when
ENGINE_URLis set. - Engine (
services/api, runs on a VPS) — FastAPI orchestrator + worker pipeline doing the compliant sourcing (official API → bulk → authorized web → licensed → else level 0). Never on Vercel (long jobs). Bypass/stealth/proxy-rotation/fake-account paths are excluded by design (CLAUDE.md §1).
Routes
| Area | Routes |
|---|---|
| Marketing | / (hero, filosofi, katalog, motorn, visualiseringar, täckning, editorial, bevis, priser, footer) |
| Catalog | /katalog, /katalog/[id] |
| Sourcing map | /kallor |
| Order lifecycle | /bestall → /bestall/forskott → /order/[id] → /order/[id]/validera |
| Customer portal | /portal, /portal/subscriptions, /portal/files, /portal/integrationer, /portal/settings |
| Admin portal | /admin, /admin/sources, /admin/jobs, /admin/customers |
| API | /api/{health,estimates,catalog,sources,orders,orders/[id],subscriptions,admin/metrics,uploads,auth/*,payments/*,webhooks/stripe,connections/db-grant} |
Design system
Mono/editorial, light, radius 0, Inter + JetBrains Mono, single red #e40014, animated grain,
scroll-reveal motion. No photography — imagery is rendered data (DataArtifact). Every fact
shows a provenance square (ProvenanceBadge). Full spec: docs/05-design-system.md.
Run locally
pnpm install
pnpm dev # Next.js on http://localhost:3000
# data engine (separate, optional)
cd services/api && pip install -r requirements.txt && uvicorn orchestrator.main:app --reload
Deploy
- Config target (as of 2026-08-09):
siax.config.jsondeclaresdeploy.target: "coolify", matching every other SIAX Cloud OS repo (n0d, st0re, aud0, ...) — a Dockerfile "for Coolify" (already committed, seeDockerfile) plus.gitea/workflows/cd.yml(renamed fromdeploy.yml2026-08-10) that POSTs a restart to a Coolify app UUID onCOOLIFY_TOKENpush. Actually live right now: the only running instance ofSee0pt.appis still on Vercel (project0pt, scope Simon Axelsson's projects) — no Coolify app has been provisioned for this repo yet, sodeploy.yml's app UUID is a placeholder that fails the job on purpose until someone fills it in. Flipping the config declares intent; it does not move traffic.docs/architecture/ARCHITECTURE.md§5/§8 for the full "Vercel or fleet" state and what's still open (Coolify app provisioning + DNS cutover — a separate decision from this config change, since the live Vercel deployment is not being touched).- CORRECTED 2026-09-06 — the bullet above is no longer true. The Coolify deploy hook is LIVE.
The repo secret
COOLIFY_TOKENexists (created 2026-08-31T11:34:34Z) and the repo Actions variableCOOLIFY_APP_IDholds a real 24-character UUID, not theTODO-no-coolify-app-provisioned-yetliteral. Both guard conditions in.gitea/workflows/cd.ymltherefore pass. Measured on the push tomainat1ff34410(2026-09-06): jobDeploy to Coolify / deploy(run 11695, job 26295) finished success withcoolify deploy HTTP 200in its log. Every merge intomainnow fires a real redeploy againstcloud.siax.io— know what you are shipping before you merge. Probed 2026-09-06 08:00 UTC:0pt.appstill answers from Vercel (server: Vercel,x-vercel-id: arn1::…) and returns HTTP 402DEPLOYMENT_DISABLED— "Payment required" on both/and/api/health. So the public domain is down, DNS has not been cut over to Coolify, and the Coolify app — whatever state it is in — serves no0pt.apptraffic. Two separate facts, both true at once: the deploy hook fires for real, and the product domain is dark behind an unpaid Vercel deployment. Still not established: whether the Coolify app itself builds and starts — the deploy endpoint returning 200 only means the request was accepted, not that a container came up healthy. - Engine + DB + embeds → one VPS via
infra/docker-compose.yml(Postgres+pgvector, Redis, api, worker, NocoDB, Metabase, Caddy). - Env inventory:
.env.example. - Auth on
/api/0pt/*is fail-closed: withAUTH_ENFORCEunset the middleware REQUIRES a verified Zitadel Bearer token (401 without one). Only the literalAUTH_ENFORCE=falsedisables it, and that is refused whenNODE_ENV=production(the process throws at startup rather than serving anonymously). Caveat (live-verified 2026-08-10): the deployed instance still answersGET /api/0ptwith 200 anonymously — the middleware appears not to run in production (gap S1, root cause unknown). Do not rely on this being enforced until an anonymous request gets a real 401. - Fonts: self-hosted via
@fontsource-variable/inter+@fontsource-variable/jetbrains-mono(no Google Fonts /next/font/google). - API contract (SSOT):
contracts/openapi.yaml.
Status
Frontend, full order lifecycle, customer + admin portals, edge API, payment/auth/embed infrastructure, the compliant data engine, schema and infra are all in place and building green. Integrations are env-guarded and activate when their keys/DB are provided:
- Database — Vercel → Storage → Postgres (sets
POSTGRES_URL) → durable persistence (schema insql/001_init.sql). - Stripe + Resend keys → payments + email.
- VPS for the engine → set
ENGINE_URL. AUTH_SECRET+DEV_LOGIN_KEY→ login (full OAuth via Auth.js lands with the DB).
© 2026 SIAX Technology AB. Not affiliated with Protosell.