Documentation

Polysync Overview

Polysync orchestrates data and AI workflows across clouds from one control plane. It works with the platforms and jobs you already have: it connects to your existing pipelines, notebooks, and functions, and schedules, chains, and monitors them, with no orchestration code to write. It supports 23 services across Microsoft Azure, Google Cloud, and AWS (see the full catalogue).

Polysync is a multi-tenant service that you subscribe to through the Azure Marketplace, and your team signs in with their existing Microsoft work accounts. To understand the problem it solves, see Why Polysync.

How it works

  1. Connect — Register your Azure, Google Cloud, and AWS platforms, then point Polysync at your secret vault: Azure Key Vault, AWS Secrets Manager, Google Cloud Secret Manager, or HashiCorp Vault. If you would rather not link a vault, Polysync can hold the credential in its own managed Key Vault.
  2. Discover and import — Polysync reads each connected platform and imports its pipelines, notebooks, and functions, parameters included, as reusable job blueprints. Add your own inputs and outputs, with default values, for the tasks built on top.
  3. Orchestrate and schedule — Build tasks from those blueprints, draw dependencies across platforms, and attach a schedule. Run a task by hand first to prove it works. All of it is configuration, so there is nothing to compile, package, or redeploy.
  4. Monitor — Every run, retry, and outcome lands in one execution history, alongside analytics dashboards and an inline dependency graph that shows where each execution sits. Turn on the AI Monitoring Agent and it watches those runs for you, explains each failure, and retries what is safe only within the limits you set. See Monitoring and The AI Monitoring Agent.

Your engines stay where they are. Polysync triggers the jobs in your own cloud accounts and tracks their outcome; it does not move your pipelines or your data, and the services it orchestrates keep billing to your own cloud accounts exactly as they do today.

Core concepts

  • Secret Vault — Connection to a cloud secret manager (Azure Key Vault, AWS Secrets Manager, Google Cloud Secret Manager, HashiCorp Vault). Stores the credentials referenced by Platforms.
  • Platform — Connection to an external execution platform (Azure Data Factory, Databricks, Synapse, Functions, Logic Apps, Google Cloud Functions, …). Authenticated via a Secret Vault entry, a Managed Identity, or the Polysync Service Principal.
  • Job — A reusable template describing what can be executed on a Platform (a pipeline, notebook, function, etc.) plus its parameter schema. A Job can be created manually or by importing pipelines from a Platform.
  • Task — A configured instance of a Job: it pins parameter values, schedules, dependencies, retry policy, and contention profiles. Tasks are what actually run.
  • Contention Profile — A reusable concurrency / rate-limit policy that caps how many tasks may dispatch on a Platform at the same time, or how often they may dispatch.

Typical setup flow

  1. Create a Secret Vault.
  2. Create a Platform that authenticates via that vault (or a Managed Identity).
  3. Import or define one or more Jobs.
  4. Create Tasks from those Jobs, assign schedules, and link contention profiles.
  5. Use Run Once to validate a Task before its schedule fires.

Using this documentation

  • One guide per connection — there is a guide for each of the 23 services and each of the 4 secret vaults, plus one for every job type behind them. They cover the authentication options, the exact permissions to grant, and the parameters each job accepts.
  • Searchable — full-text search across the library returns the matching passage, not just a page name. Every topic has its own address, so you can send a colleague the exact page.
  • Linked from the product — Setup screens link straight to the relevant page, and the in-app Polysync Copilot reads from the same library, so an answer in chat matches the documentation.

Where to go next