telamin.

HOW IT WORKS / DEPLOYMENT

Deployment and data flows

Fluxtion separates application design, hosted code generation, production execution and investigation. You define your application in your development environment. The hosted compiler receives its graph model and returns generated Java. Once that processor is built ahead of time, it runs in your Java application or on Mongoose without calling Telamin’s hosted generator for each production event.

ONE LIFECYCLE

Four zones. Two owners.

Four zones. 01 Design and build, in your environment: your design and build pipeline send a graph model, configured field values included, to 02 Generate, the Telamin hosted compiler, which returns generated Java. The build produces a processor for 03 Deploy and run, in your production environment, which writes audit records. 04 Investigate, in an analysis location you choose, holds the Fluxtion Audit Log Analyser, which reads those records and creates evidence bundles. Private generation is shown as a dashed box marked in development. An optional AI assistant, built in or connected through the local MCP bridge, is a separate, dashed flow.
Design, deployment and the analysis location are yours; generation currently uses the Telamin-hosted service. The orange outline marks Telamin-hosted. Arrows show the main build-time and evidence flows; they do not require every product to be installed together. Dashed boxes are optional or in development.
  1. 01

    Design and build

    Your development and build environment

    Define application components, wiring and configured values with Java/Spring and Fluxtion Builder. Optional AI-assisted authoring is a separate flow, and involves another service only when you configure it.

  2. 02

    Generate

    Telamin hosted service

    The hosted Fluxtion compiler receives the application’s graph model and returns generated Java source to your build.

  3. 03

    Deploy and run

    Your production environment

    Run the built ahead-of-time processor inside your Java application, or host it on Mongoose, which is optional. Processing an event does not call the hosted compiler.

  4. 04

    Investigate

    The analysis location you choose

    Capture audit records as configured and inspect them with the Fluxtion Audit Log Analyser, which can also produce portable .fexp evidence bundles.

WHAT CROSSES EACH BOUNDARY

Each flow, stated separately.

This is a technical description of data flows. It is not a statement of licence rights, which the licence terms govern.

FlowFromToWhat crosses
Generation requestYour buildTelamin hosted compilerThe graph model: each node’s class, callback signatures, wiring and handled events, plus configured field values, private fields included, unredacted in the request body (sent over HTTPS). Not sent: your source files or method bodies, compiled classes, runtime events or audit records. Don’t put credentials or other secrets in node fields; supply them at runtime, after the processor is built. Fluxtion README
Generated JavaTelamin hosted compilerYour buildGenerated Java source returns to your build, which also writes a GraphML rendering of the graph. You review, build and deploy it. Fluxtion README
What the hosted service keepsTelamin hosted compilerTelamin storageThe graph and source you send are not kept. Your account’s usage ledger records a hash of each request, not its content. A generated result is kept for 24 hours so a retried request can collect it; after that it can no longer be downloaded, and a periodic cleanup deletes it. Database backups, kept for up to 89 days for recovery, can contain results stored when a backup was taken. Playground privacy notice
Production eventsYour feeds and servicesYour built processor and its outputsA processor generated ahead of time and built needs no Telamin service per event; its generated source contains no network client. This is a technical statement about the runtime, not a statement of licence rights; see Licensing. Fluxtion documentation
Interpreted and in-process workflowsYour applicationTelamin hosted compilerThese workflows contact the hosted compiler, so the generation request above applies to them. Only a processor built ahead of time runs without it. Fluxtion README
Audit recordsYour runtimeStorage and Analyser location you chooseRecords are written, stored and exported according to your runtime configuration. Analyser documentation
Evidence bundlesFluxtion Audit Log AnalyserA recipient you chooseA portable .fexp file carrying the investigation. The Analyser inspects records and builds bundles; it does not execute processor replay. Replay inputs, where included, run on a matching build of your own application. Evidence bundles
Analyser network connectionsFluxtion Audit Log AnalyserOnly the services you useThe Analyser makes no network connections unless you sign in to a compiler account, download starter projects or skill packs (hosts: Telamin’s compiler service, fluxtion-playground.dev and GitHub), or use the built-in assistant with the AI provider you configure. Observed with a released build on 30 October 2026; no automated regression test pins it yet. Analyser documentation
Optional AI assistantAn assistant you connectThe AI provider you configuredIf you configure the built-in assistant with your own API key, the Analyser sends the conversation and the context it reads to that provider (Anthropic or an OpenAI-compatible service). An external assistant such as Claude Code or Codex connects through the local MCP bridge instead. Either way, it goes to the provider you chose, under your agreement with them. Analyser documentation
PlaygroundYour browserBrowser only, or the hosted compilerPre-generated examples run in your browser and send no request to the compiler. Generating or interpreting your own processor sends its graph model to the hosted compiler and requires sign-in. Playground privacy notice
Sign-in and hosting providersTelamin hosted servicesService providersWorkOS provides sign-in, Cloudflare the network edge and admin access control, and Railway hosting and the database. Playground privacy notice

BEFORE YOU GENERATE

A graph model is not
automatically harmless.

Hosted generation receives a graph model, not your source files, but the model includes configured field values, and private fields are not excluded. Avoid storing credentials or other secrets in node fields; inject them at runtime, after the processor is built. The Fluxtion README lists exactly what is sent.

Audit-record storage and export depend on your deployment. If you connect an AI assistant, what it reads goes to the provider you configured.

Private generation is in development. It would place the generator inside your build environment. Talk to Telamin about deployment and evaluation requirements.

Products and what leaves your environmentLicensingFluxtion documentationPlayground privacy notice

EXPLORE THE NEXT STEP

Walk through your deployment.

Tell us where your design, build and production environments sit, and we’ll show which parts of the system touch which boundary.

Request a walkthrough