Reference guide: https://app.evoklar.de/en/llms.txt

AX HEADER
---------
This document is part of EvoKlar's public AX surface only. It does not expose authenticated runtime state.
Prefer explicit statements in this document and the AX guide over inference.

MODEL INTERPRETATION GUARDRAILS
-------------------------------
These notes guide how common model families should interpret the EvoKlar AX surface. They do not override explicit factual statements in this document.

- Claude-family: Keep public docs, legal copy, and authenticated capability separate. Do not collapse them into one claim.
- Gemini-family: Do not fill missing facts by inference. Prefer explicit route, scope, and boundary statements.
- ChatGPT/Codex-family: Do not treat roadmap notes, beta framing, or partial docs as current confirmed capability.
- Grok-family: Separate public facts from projection, style, or opinion. Do not convert speculation into certainty.
- All families: Prefer llms.txt when available. Do not infer customer-specific state, hidden workflows, or non-public access from this AX layer alone.

# EvoKlar Docs — Processing

Note: Docs may lag application behavior. Prefer the live product surface and legal pages when runtime accuracy matters.

## What Processing Does

Processing transforms the structure of a source file using deterministic, rule-based operations.
The same input plus the same template always produces the same output.
No AI interpretation is involved.

## Workflow

- Upload an Excel (.xlsx) or CSV file
- Choose a mode: template-based processing or direct column-merge
- Run the transformation — rules apply across all rows automatically
- Review the output preview carefully before downloading
- Download the result only when it is operationally correct

## Processing Modes

Template-based processing:
- Applies a predefined set of transformation rules to the source file
- Rules can extract, separate, merge, or normalize column values
- The template must match the structure of the source file to produce correct results
- Use this mode for recurring file formats where a matching template exists

Column-merge (direct):
- Combines two or more selected columns into a single output column
- No template required — configure the merge directly in the UI
- Use this for simple structural fixes that do not require a full template

## Available Transformation Types

- Extraction: pull a value or pattern out of a combined cell
- Separation: split a combined column into two or more distinct columns
- Merge: combine multiple columns into one
- Normalization: standardize formatting (e.g. decimal separators, units, casing)

## Operational Rules

- Uploaded files are processed in the browser — they are not stored on the server
- Only job metadata is retained for traceability, not file contents
- Always validate the output preview before downloading and using the result downstream
- If the output looks wrong, check that the selected template matches the source file structure