Menu

Integrations & Connectivity Architecture

Least-privilege connections with explicit permission boundaries

When connecting to external ad platforms, storage vaults, and production infrastructure, ADSOMI never markets third-party brands as partners. Every connection is governed by server-side authorization, explicit user consent, and our 7-day media quarantine download rule.

Integration Categories

Permission boundaries and availability across five connectivity layers

Each integration category documents what data it reads, which actions require explicit confirmation, and how tenant isolation is enforced:

INT-01Read-Scoped Analytics & Gate-Controlled Publishing

Ad & Social Account Insights

Reads authentic date ranges and attribution windows from connected ad accounts to map creative performance inside Creative DNA. Live publishing or ad spend never triggers without explicit user authorization.

Authorization & Permission Boundary: Read-only performance metrics are strictly separated from write permissions that schedule posts or authorize ad spend.

Enforced Security Safeguards
  • Never displays simulated or estimated ROI charts for unconnected accounts.
  • OAuth tokens are encrypted server-side and never exposed to client bundles.
  • Platform connectivity operates strictly for technical data access and never implies commercial sponsorship.
INT-02Active Core Infrastructure

Private Object Storage & 7-Day Quarantine Vault

Uploaded source product/venue photos, SHA-256 digests, approved intermediate frames, and 3:4 (1440×1920) 2K master deliverables are isolated inside S3-compatible private object storage.

Authorization & Permission Boundary: Every media object is accessible exclusively via short-lived signed URLs scoped to its owning tenant and brand workspace.

Enforced Security Safeguards
  • Temporary outputs on external production infrastructure are downloaded into ADSOMI private quarantine storage within 7 days.
  • Raw provider CDN URLs are never exposed to end users, portfolios, or browsers.
  • Identifiable likeness and voice references remain strictly isolated per brand workspace.
INT-03Active Core Infrastructure

Network-Isolated Local Media Mastering (FFmpeg Worker)

True vector logo and verified price layer compositing, smooth frame transitions, audio normalization, and 3:4 (1440×1920) 2K Lanczos + CAS mastering run inside our local media worker.

Authorization & Permission Boundary: Isolated worker boundary closed to outbound public internet traffic, accepting only verified jobs from the internal queue.

Enforced Security Safeguards
  • C2PA Content Credentials and mandatory provenance markers are preserved across exports.
  • 3:4 2K masters and 9:16 / 4:5 social derivatives are processed with separate safe-zone profiles.
  • Open-source license notices (FFmpeg LGPLv2.1+ / GPL) are documented on our legal transparency page.
INT-04Scope-Locked Review & Signed Delivery

External Client Approval Portal & Webhook Notifications

Enables agencies and brand teams to collect side-by-side source-vs-output approvals (`approved` / `rejected`) from external stakeholders and track state transitions via signed event logs.

Authorization & Permission Boundary: External review links expose only the selected project candidates and never grant access to broader workspace settings or other brands.

Enforced Security Safeguards
  • Email and notification links never bypass server-side authorization checks.
  • Rejected candidate frames are locked from advancing to video or final export.
  • All approval and revision decisions are recorded in a timestamped audit log.
INT-05Advisory-Default Controlled Gateway

OMI Tool Gateway & Context Boundary

The server gateway governing how OMI accesses workspace data—protecting against prompt injection, cross-brand context leakage, and unapproved cost execution.

Authorization & Permission Boundary: Read-only context tools (brief and screen state) are strictly separated from mutation tools (project creation, render queueing).

Enforced Security Safeguards
  • Mutation tools require explicit user intent, RBAC verification, and cost preview.
  • Secrets, API keys, and BYOK credentials never enter chat context or logs.
  • No silent provider switching occurs during outages; transparent status notices apply.

Key & Data Governance

Four non-negotiable security rules across all integrations

External connections and infrastructure adapters operate under the following binding security contract:

SEC-01

Server-Only Secret Storage

No infrastructure key, OAuth secret, or signing credential is ever placed in client code, browser bundles, or public environment variables (`NEXT_PUBLIC_*`).

SEC-02

Write-Only BYOK & Preserved Safeguards

When an enterprise tenant supplies its own key (BYOK), the credential is write-only, never readable back, and scoped strictly to that tenant. BYOK never waives AUP, written consent, or C2PA rules.

SEC-03

No Silent Provider Switching

If an underlying production route changes, the corresponding attribution, privacy/subprocessor record, and operation scope update together; undocumented background rerouting is prohibited.

SEC-04

Actual Recipient Transparency (KVKK Art. 9)

User data is transmitted only to active integrations for the selected workflow as disclosed with purpose, transfer mechanism, and retention window on `/en/legal/subprocessors`.