ingestia.ioingestia.io
Architecture

The path your data travels

Seven stages between the source system and the decision. Click a stage to see the screen that runs it and the page that explains it.

The platform diagram

The same architecture you see inside the product

Sources, ingestion, orchestration, a layered datalake and consumption — with governance across everything. Click any piece to see where it is explained.

Orchestration and scheduling

Decides WHEN each load runs, reruns what failed and logs everything.

drives block 2 (ETL/ELT) and the transformations in block 4 (Datalake)

Data sources

Systems where the data lives today.

ETL/ELT

Brings data from outside into your datalake.

Triggered by: DAGs (block 3)

Datalake

Storage and processing, in three layers.

Transformations triggered by the DAGs (block 3)

Consumption and intelligence

Where processed data becomes decisions.

Every piece is a link: it takes you to the site page or the documentation section that explains that part.

Per-customer isolation

Dedicated datasets per workspace — nothing crosses between customers.

Cost upfront

Estimate before running, charged afterwards — with a cap per query.

Audit

Runs, exports and personal data reveals are logged.

  1. 1

    The source

    The system where the data is born: ERP, CRM, database, spreadsheet, file or cloud service. Nothing is installed inside it — ingestia.io reads through the access you authorize.

    Read the documentation
    The source
  2. 2

    The connection

    The credential is stored and tested before it counts. If the test fails, the connection does not go live — and the reason is written out.

    Read the documentation
    The connection
  3. 3

    The scheduled load

    You choose the frequency and the time. On every run the platform records what it brought in, how long it took and how much it cost.

    Read the documentation
    The scheduled load
  4. 4

    Bronze — the data as it arrived

    A faithful copy of the source, with a date. It serves as proof: you can go back and check where each number came from, even months later.

    Read the documentation
    Bronze — the data as it arrived
  5. 5

    Silver — the data tidied up

    Types fixed, names standardized, duplicates resolved, keys matched. This is where ERP and CRM start speaking the same language.

    Read the documentation
    Silver — the data tidied up
  6. 6

    Gold — the data that answers

    Tables shaped like the business question, with the measures already defined. BI and AI read from here, never from the source.

    Read the documentation
    Gold — the data that answers
  7. 7

    Consumption

    Dashboards, questions in plain language, agents, Power BI, Excel and the API — all over the same Gold layer, with the same access control.

    Read the documentation
    Consumption

Fictitious demo data.

Where it runs

Google Cloud infrastructure in São Paulo

Every step above runs on a managed Google Cloud service, in Brazil. With your own cloud, it runs in your company's project.

  • Layers on BigQuery

    Bronze, Silver and Gold in each customer's own datasets, isolated by the environment prefix.

  • Landing on Cloud Storage

    Files and extracts arrive in an area exclusive to the environment before becoming tables.

  • Processing on Cloud Run

    Loads run in on-demand containers. Notebooks run on Spark clusters you switch on and off.