Skip to content
SialFlow

PROGRAMMABLE MEDIA INFRASTRUCTURE

One workflow for every asset.

Process, transform, understand and deliver images, video and documents through one programmable and auditable platform.

Explore the platformView API documentation
  • Image
  • Video
  • Document
  • AI ops
SialFlow Asset Flow: source assets through the flow engine to secure, optimized delivery

Platform

Stop stitching media tools together.

Images, video, documents and AI operations should not live in disconnected systems. SialFlow brings them into one workflow engine, one API model and one operational view.

One orchestration layer

Image, video, document and AI-adjacent jobs share the same control plane instead of disconnected tools.

Shared API and events

Bearer auth, project scope, async jobs and outbox events follow one model across capabilities.

Tenant-aware execution

Quota, concurrency and fair scheduling keep multi-tenant workloads predictable.

Complete processing history

Job timelines, retries and audit-oriented event trails stay attached to every run.

Execution

See a flow in action.

Image pipelines with URL transform grammar and async job patterns — from ingest through variants on one control plane.

Example pipeline

Input
master
Operation
background remove
Policy
format: png|webp
Output
png alpha
State
Running
Duration
~2s

Background remove (example)

# Async job → new asset
POST /v1/images/{id}/bg-remove
{"format":"png"}

# Sync CDN path (alpha formats only)
/i/{org}/{image}/w400-e-bgremove-f-png/preview.png

# Poll: GET /v1/images/jobs/{job_id}

Image

Image workflows at catalog scale.

Turn repetitive image operations into durable, versioned and observable pipelines.

  • Smart crop (g-auto entropy)
  • Background removal (async + sync)
  • Responsive variants
  • Bulk processing jobs
  • Metadata on objects
  • Versioned transform profiles
  • Policy-scoped delivery
  • Multi-layer URL overlays
Explore image →

Input → nodes → variants

hero.jpg · ingest

Background remove · Running

Smart crop · Queued

w800 / w1200 / avif · Queued

enrich / moderate · roadmap

Timeline · encode ladder

scene markersencode ladderthumbnails

Encoding, scene markers, and batch jobs share the same job and telemetry model. Transcription as a Flow step is still ahead of the shipped first vertical.

Video

One execution model for video operations.

Run VOD, live and batch media processing through a shared orchestration and telemetry layer.

  • VOD encoding jobs
  • Batch media processing
  • Thumbnail generation
  • Clip-oriented workflows
  • Provider-neutral worker pools
  • Shared job telemetry
  • Content-aware encode options
  • Live-to-VOD paths (product direction)
Explore video →

Document Processing

Documents transformed into structured outputs.

Convert document pipelines into searchable, classifiable and machine-readable workflows on the same control plane.

  1. 01Original document
  2. 02Extraction / OCR
  3. 03Classification
  4. 04Metadata
  5. 05Structured output
  6. 06Review state
  • OCR / searchable PDF
  • Conversion & merge ops
  • Structured extraction paths
  • Page-async workers
  • Control-plane job model
  • Project-scoped execution
Explore Document Processing →Secure Documents →

Secure Documents

Sensitive files, shared with the people who need them.

Store the pack once, share it with your team, or send an outside recipient a link that expires. They open it under your rules — and you keep a clear record of the delivery.

  • Folders and versions
  • Share with people and groups
  • Links that expire
  • Delivery receipts
Explore Secure Documents →

AI & moderation

Understand assets before they are published.

Background remove is live. Broader enrichment and moderation queues stay on the roadmap — we keep AI-adjacent work on the same tenancy and Job Center, without pretending GenAI is fully shipped.

Background intelligence

Async background remove for heavy jobs, plus sync delivery transforms when you need an immediate alpha PNG or WebP.

Same control plane

AI-adjacent jobs share tenancy, quotas, events, and Job Center visibility with image work. Enrichment and moderation queues come later.

Honest scope

We ship bg-remove and image pipelines first. Broader GenAI enrichment is direction, not a pretend live product.

AI & Moderation →

Architecture

Built for workflows, not isolated API calls.

A layered control plane: clients hit one API model, policy and tenant scheduling decide execution, workers process, storage stays at the boundary, events close the loop.

  1. Client / API

    Bearer auth | project context | idempotent creates

  2. Workflow control plane

    Jobs | pipelines | outbox events | versioned defs

  3. Policy + tenant scheduler

    Quota | concurrency | fair share | retries

  4. Worker capacity

    Image · AI · Video · Document — isolated execution

  5. Object / customer storage

    Upload sessions | derived variants | connectors

  6. Events + telemetry + audit

    Outbox | job timelines | Governance audit explorer

  • Cross-media orchestration
  • Durable job execution
  • Versioned Flow definitions
  • Tenant quota management
  • Concurrency control
  • Fair scheduling
  • Provider-neutral workers
  • Customer-oriented storage connectors
  • Audit visibility via events
  • Per-job telemetry
  • Flow evidence on runs
  • Common authentication
  • Shared event / outbox model

Multi-tenant

Predictable execution across tenants.

Quota, concurrency and fair scheduling keep shared infrastructure from becoming a noisy neighbor problem.

  • Tenant quota
  • Concurrency limits
  • Fair scheduling
  • Per-tenant isolation
  • Job prioritisation
  • Retry policy
  • Resource visibility
  • Cost attribution signals

Tenant execution matrix

TenantRunningQueuedScheduler note
Tenant A1240fair share
Tenant B48within concurrency
Tenant C819retry backlog

Sample UI for quota, concurrency, and fair scheduling - not production counters.

Storage & providers

Keep control of storage and processing providers.

Connectors and worker pools are designed so storage ownership and compute providers stay at explicit integration boundaries — without locking you to a single vendor name.

  • Customer-owned storage direction
  • Object storage connectors
  • Provider-neutral worker execution
  • Controlled data movement
  • Integration boundaries
  • Credential isolation
  • Policy-based routing

Solutions

Operational models by industry.

Each solution maps to a workflow shape — not a generic icon grid.

E-commerce

Operational problem
Catalog and campaign assets sprawl across disconnected image tools.
Workflow model
Ingest → moderate → transform → variants → deliver
Primary output
PDP-ready variants, overlays and auditable preparation jobs.
View solution →

Agencies

Operational problem
Multi-client delivery needs reusable pipelines without losing audit.
Workflow model
Templates → brand policy gates → scheduled / bulk runs
Primary output
Client-scoped workspaces with shared execution patterns.
View solution →

Publishers

Operational problem
Editorial image, video and document work lives in separate systems.
Workflow model
Shared tenancy across Library, Delivery, and Flow first vertical
Primary output
One operational view from ingest through publish-ready assets.
View solution →

Developers

A consistent execution model across the platform.

Auth, tenancy headers, async jobs and events stay aligned so image, video and document work feels like one platform.

  • Bearer authentication
  • Project context
  • Workspace / org scope
  • Idempotent creates
  • Async jobs
  • Resource identifiers
  • Outbox events
  • Polling until push webhooks ship
  • Consistent error shape
  • Audit metadata
  • Retry state
  • Workflow version
DocumentationAPI reference

curl · background remove (shipped)

curl -X POST https://api.sialflow.com/v1/images/$IMAGE_ID/bg-remove \
  -H "Authorization: Bearer $SIALFLOW_API_KEY" \
  -H "X-SialFlow-Project-ID: $PROJECT_ID" \
  -H "Content-Type: application/json" \
  -d '{"format":"png"}'
# 202 → job kind=bg_remove · poll GET /v1/images/jobs/{id}

Observability

Every job leaves an operational record.

Illustrative telemetry for node status, retries, provider pool and audit events — not a live product screenshot.

  • Execution timeline
  • Node-level status
  • Retry history
  • Error context
  • Duration
  • Worker capacity
  • Workflow version
  • Tenant context
  • Audit events
  • Cost signals

Job record

Running
  • WorkflowImage pipeline v3
  • StepTransform · Running
  • Retries0 / 3
  • PriorityStandard
  • ScopeOrganization · Project
  • AuditEvent recorded

Progress, retries, and audit stay attached to every job on the control plane.

Control boundaries

Platform guarantees without vanity badges.

Trust is expressed as architectural controls — not customer logos or unverified certifications.

Auditable execution

Jobs and outbox events leave an operational record you can inspect.

Tenant isolation

Org, workspace and project context scopes work and credentials.

Provider-neutral architecture

Worker pools are treated as interchangeable execution capacity.

Customer-controlled storage

Connectors keep storage ownership at the integration boundary.

Observable workflows

Status, retries and duration stay visible at the job and node level.

Versioned definitions

Pipeline changes can be tracked as versioned workflow definitions.

Flow evidence

Published Flows keep allow/deny evidence and lineage on each run — separate from the outbox.

Policy-driven processing

Transforms and bg-remove jobs run under project tenancy and quota, with Flow gates where you publish a definition.

Start building

Request portal access

Create a workspace for the developer portal, then run image, video and document jobs on one control plane.

Build your next media workflow on one platform.

Start with an image pipeline, a video job or a document flow, then connect them as your operation grows.

Explore the platformRead the documentation