Skip to main content

🗄️ AI@Flows Data Files (Admin)

This page is for administrators who produce the analysis files and implementation reports, or who manage the shared folders. End users do not need it.

The full contract is in the specification templates/IF_IMPLEMENTATION_SPEC.md, delivered with the workbench. This page is a summary.

All files are UTF-8 JSON. The workbench only reads the analysis files and implementation reports; it writes only to the Review output folder: review comment files at the top level, settings and sign-offs in the _config subfolder. Review files are described in templates/IF_REVIEW_SPEC.md.


Interface analysis files​

Stored in the Interface Analysis reports folder. The workbench reads two formats and converts format 1.0 to 1.1 when it loads the file.

ItemFormat 1.0Format 1.1
Headersystem (no system_id)type = INTERFACE_ANALYSIS, format_version = 1.1, system, system_id (SID of the analysed system)
Status per interfaceLegacy status fieldstatus: Evidence found · Partial evidence · Not found
Source inventory rowLegacy inventory fieldsource_list (tab, row, fields, feedback expected)
ATC linksATC IDs only, read as cycle CYCLE-01atc_entries: "<cycle> <ATC ID>", any cycle

Optional fields an analysis file can set explicitly (they win over the workbench's keyword rules):

WhereKeyValues
InterfacedomainA code from Config › Business domains, e.g. P2P
InterfacetagsList of tag names
Flowtech_classIDOC · RFC · WS · FILE · ARCH · PRINT · OTHER
FlowmodeAutomated · Manual
FlowboShort business objects, e.g. ["PO", "GR"]
Custom code object / findingcomplexityC1 · C2 · C3 – complexity of the linked ATC entry in the AI@CustomCode analysis
custom_code.countscomplexity, max_complexityNumber of distinct linked ATC entries per level, e.g. {"C1": 2, "C2": 1, "C3": 0}, and the highest level. Without them the workbench counts the levels from the objects and findings

One analysis file per SAP system​

Write one analysis file per SAP system (ECC sandbox, development, quality, production; S/4HANA systems when analysed), with the same interface IDs (App_xxx) in every file. system_id is the SID of the application server, not of the database.

After an in-place conversion ECC and S/4HANA keep the same SID: write the ECC system as <SID>-ECC and the S/4HANA system as <SID> – in system_id, in the file name and in Config › SAP systems.

Application instances​

The Application landscape reads, per interface, the application instances connected to the SAP system of the file:

KeyContent
app_instancesList of {instance, tier, host, connection, evidence} – e.g. {"instance": "WMS DEV", "tier": "Development", "host": "wms-dev"}. tier uses the tiers of Config › SAP systems
app_instances[].single_saptrue = the instance can be connected to one SAP system only: its ECC connection ends when an S/4HANA system is plugged in (shown as ends at the re-plug)

The re-plug to the S/4HANA systems is not in the analysis files: it comes from the implementation reports of the S/4HANA systems (the instance used by the ECC system of the same tier).


Implementation reports​

Stored in the Interface Implementation reports folder. One file per target system and cycle:

IF_Implementation_<SID>_<CYCLE>.json
e.g. IF_Implementation_S4D_CYCLE-02.json
  • The cycle ID (cycle) is set by the AI application that writes the report.
  • A newer report for the same system and cycle replaces the file. If several versions are in the folder, the workbench uses the one with the newest generated_on and ignores the others.
  • Before writing a report, read the previous cycle, the latest sign-offs (_config/IF_Signoffs.json) and the processed implementation reviews.
  • Older reports with an iteration number are still read.

Header (main keys): type = INTERFACE_IMPLEMENTATION, format_version = 1.1, project, system (<SID> / <client>), system_id (must match a SID in Config › SAP systems), cycle, cycle_name, based_on (the analysis file used), generated_by, generated_on, summary, interfaces, and optionally demo.

Per interface (main keys):

KeyContent
id, nameSame ID as in the analysis file (e.g. App_001). This links the Analysis and Implementation pages
statusOne of the implementation statuses
owner, middleware, to_be_pattern, summary, next_stepsOwnership and status text
changedfalse if nothing changed since the previous cycle (older files: changed_in_iteration)
flowsAs-is → to-be per flow, change type, status, test result, evidence
configurationArea, name, source value, target value, status (OK · Changed · Missing · Not needed), note
custom_codeATC open / fixed per priority, objects with change status and transport
transports, tests, issuesTables as shown on the Implementation page
Configuration trace

Use the same area and name for the same item in every system's report. The workbench ignores the SID inside the name when it matches items, so MW_GATEWAY_S4D and MW_GATEWAY_S4Q are traced as one item.


Review comment files​

Written by the workbench at the top level of the Review output folder; the AI application answers each closed round with a resolution file.

ReviewReview file (one per save)Resolution file (AI application)
AnalysisIF_Review_<analysis>_R<nn>_<timestamp>.jsonIF_Review_<analysis>_R<nn>_resolution.json
ImplementationIF_ImplReview_<SID>_<CYCLE>_R<nn>_<timestamp>.jsonIF_ImplReview_<SID>_<CYCLE>_R<nn>_resolution.json
  • The latest version of a round is the one with the highest version number. Earlier versions are never changed (audit trail).
  • A round with status = Closed is the input for the AI application.

Files written by the workbench​

Settings and sign-offs, all in <Review output>/_config/:

FileContent
IF_Systems.jsonSAP systems (SID, client, tier, product, landscape, status)
IF_Tags.jsonTags
IF_Domains.jsonBusiness domains
IF_Assignments.jsonDomain and tags per interface ID
IF_Signoffs.jsonSign-off log: system, cycle, interface, role, user, decision, comment, timestamp, report version signed
IF_Reviewers.jsonReviewer list (encoded, PINs as salted hashes)

Each time a file is saved, the replaced version is kept as <name>_previous.json. If a file is damaged, the Config page says so; restore it from the _previous file or from the SharePoint version history.

Keep data out of the app and out of Git

The HTML file holds code only. Keep the analysis, implementation and review folders out of source control. They contain project data.