← Grimoire
v0.0.0 The Alchemist
// Tome III · The Alchemist

Anmerkung Processor

Upload · select forwarder · transmute · download
✦ ᚦ ✦

I — Select Forwarder

II — Upload Scroll

Click to browse or drag & drop your .xlsx file

III — Transmute & Download

Values below this absolute difference are treated as "no error" per forwarder. Saved locally in your browser.
default 0.08
default 0.09
default 0.04
default 0.09
Dry-run Preview
SheetRowStatusProposed AnmerkungReason
0
rows scanned
0
anmerkung filled
0
skipped (stat≠10)
0
no trigger (empty)
Trigger Breakdown
Upload multiple .xlsx files and process them all at once with the selected forwarder. Each document gets its own download button, or download all as a ZIP.
Click to browse or drag & drop multiple .xlsx files
    Type values for the cells the selected forwarder reads, hit Evaluate, and see exactly what Anmerkung the rules would produce. No file needed — pure what-if.
    // presets:
    Result — select a forwarder above and fill in any fields to test
    Drop two processed workbooks and Kiro will train the engine against them. Put your tool's output in A, the hand-corrected ground truth in B. Every disagreement becomes a labeled training case — wrong missed overfired — with the input cells the rules read and the trigger trace. Ship it as CSV/JSONL, or click ✦ on any row to open it in the Rule Tester and iterate on the rule against that exact case. For AI-driven rule updates: grab the ✦ AI Bundle (ZIP) — it packages the training set with a self-describing rule spec and a prompt template so any AI assistant can propose patches to assets/anmerkung.js without reading the source. Need a wider corpus? Use Bulk below to compare many predicted/expected pairs in one pass.
    A — Predicted (tool output)
    Click or drop an .xlsx
    B — Expected (ground truth)
    Click or drop an .xlsx
    Bulk — compare many pairs at once wider training corpus
    Drop a set of predicted workbooks (A) and the matching expected workbooks (B). Files are auto-paired by name — role words like predicted / expected are ignored — and every pair is compared in one pass. All rows feed the same training breakdown, filters, and exports below, so you get a far wider look than a single file pair. Each exported row carries a source_file so you know which pair it came from.
    A — Predicted set (tool output)
    Click or drop .xlsx files
    B — Expected set (ground truth)
    Click or drop .xlsx files