Orchestria Guide
No-Code Data Transformation

1 Create your first transformation

The video above shows the whole journey. The key points:

• Simple mode: a guided wizard (name, formats, samples, instructions, summary, generation). Expert mode: everything in a single window.
• Formats: JSON, XML, CSV, text or PDF as input and output. For CSV, set the separator and whether there is a header row.
• Samples: a real example of your input data and the expected result. This is what lets the AI understand the transformation.
• Instructions: describe in a few sentences what the transformation must do. Refer to fields by name, using dot notation for nested objects ("customer.address.city").
• Generation: when you save, the AI generates the code; you then test the result and refine it through the chat.

2 Call your transformation through the API

The video above shows the whole journey. The key points:

• Call URL: the "eye" icon in the list shows the transformation's unique, permanent URL. Calls are POST requests with your data in the body.
• API key: create it under Developer → API Keys and send it in the Authorization header ("Bearer" followed by the key). Never share it: it gives access to your transformations.
• History: every call appears in the Logs, with its status and result.

curl -X POST https://orchestria.io/api/flows/YOUR_FLOW_UID/execute \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: text/csv" \
  --data-binary @orders.csv

3 Automate with routing

The video above shows the whole journey. The key points:

• Input: an SFTP connector watches a folder and picks up new files automatically (with optional archiving); an HTTP connector polls an API regularly. Every poll, even an empty one, counts toward your quota: choose a suitable frequency.
• Output: the result is sent to a webhook (a URL in your system) or dropped as a file on an SFTP server.
• Routing: the Routing page links a transformation to its input and output connectors. A connector is created once and reused across transformations.
• Authentication: none, Bearer Token, Basic Auth or Sign-In (a prior login that returns a token).

4 Chain transformations

The result of one transformation becomes the input of the next: for example, a CSV order converted to XML, then sent to billing (XML → JSON) and to logistics (XML → CSV). The Routing page diagram shows the chain, a banner flags a chain with no output, and the Logs keep track of each hand-off.

5 Virtual DNS

A virtual DNS gives you a stable address (for example "production" and "staging") that points to the transformation of your choice. You test a new version on the staging address, then switch the production address to it: calling systems change nothing, and rolling back is instant.

6 Processing PDF documents

The video above shows the whole journey. The key points:

• Extraction: AI Vision reads each PDF and fills a JSON template (number, date, issuer, amounts, lines…). The template is suggested from your example; you can rename, remove or add fields, start from a predefined template (Invoice, Purchase order, Quote, Delivery note) and check it with "Test the template".
• Transformation: the extracted JSON is then transformed into your output format, like any transformation.
• Validation: the "PDF Validations" page shows the PDF and the data side by side, the score and the detail of each check, and lets you fix the JSON, validate or reject, one by one or in bulk.

The 3 validation modes

• 🚀 Automatic: runs right after extraction, with no human check. For standardized, reliable documents.
• ⚖️ Controlled: a consistency score is computed for the document type (totals, VAT, number, date…). Above the threshold (95% by default) it runs automatically; below, it waits for validation.
• 👁️ Manual: every PDF waits for a human validation before running.

Send a PDF through the API

curl -X POST https://orchestria.io/api/flows/YOUR_FLOW_UID/execute \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/pdf" \
  -H "X-Filename: invoice.pdf" \
  --data-binary @invoice.pdf

In Controlled or Manual mode, the response is 202 with "status": "pending_validation" until the document is validated.

7 Alerts, logs and diagnosis

The video above shows the whole journey. The key points:

• Alert rules: ask the transformation's chat in plain language ("alert me if the amount including tax exceeds €10,000"), or set them at creation.
• Emails: tick "Email alerts" on the transformation. You get an email for each execution that triggers a rule, a ❌ email when an execution fails, then a ✅ when everything works again.
• Logs: the Logs page lists every execution (filters by transformation, status, dates); the link in an alert email opens the execution directly.
• Diagnosis: an execution's details give a readable cause and the technical details; the "Explain" button (Pro plans and above) adds an AI analysis.

8 Your account and your tools

Each plan sets monthly quotas, shown in the quota bar at the top of the application and in "My Account":

• Transformations: how many transformations you can create.
• Executions: every call counts, whether it succeeds or fails.
• AI tokens: used at creation, by the chat and by PDF extraction.

The video also covers account settings (language, time zone, AI provider), changing plan, a transformation's PDF documentation and API documentation (Swagger), and locking.

9 AI Provider Choice

Orchestria supports multiple artificial intelligence providers. You can choose the one that best suits your needs in terms of quality and cost.

Available providers

How to change provider

Go to My Account → Settings to select your AI provider. The change applies immediately to all your operations: transformation creation, AI chat modifications, and PDF extraction.

10 Security and Privacy

Protecting your data is our absolute priority. Orchestria is designed to ensure total confidentiality of your business data.

AI and data separation

Artificial intelligence is only used during transformation creation to generate transformation code. It only has access to the samples you provide, never to your real production data. Your execution payloads never transit through AI services.

No data storage

Data you send for execution is processed in memory then immediately deleted. Orchestria stores no payloads, which makes execution replay impossible but guarantees total GDPR compliance. Only metadata (status, timestamp, user) is retained for logs.

Sample protection

The data samples you provide when creating a transformation are securely stored in your account and never shared with third parties. These samples are necessary for the transformation to function and can be modified by editing the transformation.

AI and PDF documents

When creating or editing a PDF transformation, the document is sent to the AI for processing. Only "PDF Vision" mode uses AI for data extraction at each execution. This mode consumes more resources and should be used knowingly for complex or scanned documents.

11 Technical limits

To ensure optimal performance and service stability for everyone, Orchestria applies certain technical limits to processed payloads.

Maximum payload size

The maximum request size (input payload) is 10 MB. This limit applies uniformly to all formats (JSON, XML, CSV, text) and all execution methods (REST API, webhooks, connectors). Beyond this limit, you will receive a 413 error (Request Entity Too Large).

Capacity by format

A 10 MB limit allows processing approximately 500,000 simple JSON objects, 100,000 to 200,000 CSV rows, or 20,000 complex XML elements. This capacity covers 95 to 98% of standard integration use cases. For larger volumes, we recommend splitting your data into multiple calls or contacting us to discuss a solution adapted to your needs.