FYSK O
Document reading
Reads each accounting document and returns its content as structured data.
FYSK O overviewAI models for Swiss accounting.
FYSK is our family of AI models, trained in Delémont. One reads your documents, the other does the accounting classification and makes the entry. On your premises, with Kokonut Intelligence.
Two models, from document to entry.
FYSK C
Accounting classification
Does the accounting classification of each document read and makes the entry.
FYSK C overviewOn your premises or in the cloud.
On your premises with Kokonut Station, or in the cloud through the provider of your choice.
1
Kokonut Drop
Creates the accounting profile for the AI.
.kdf file
2
Kokonut Wave
Uses the profile to pre-process each document, then sends it to the AI.
.kdf
Profile loaded
each document
each document
3
Kokonut Station
offline
Kokonut Island
FYSK O
FYSK C
Ollama / LM Studio
AI of your choice
or
3
In the cloud
The AIs on the market, your choice
- ChatGPT
- Claude
- Gemini
- Mistral AI
- DeepSeek
- Infomaniak
contextualised entry
4
Your accounting software
Kokonut Fid
or another compatible software
3200
documents processed per hour
Measured on a DGX Spark
With Wave and Drop, give your AI the rigour of an accountant.
Whatever the AI, this duo anchors its results in each client's context and checks what comes out.
For fiduciaries-
Kokonut Drop
The real profile of the file
Each client file has its profile, in a single .kdf file.
-
Kokonut Wave
A standard for every AI
Wave keeps each AI in line with this profile and standardises the entry that comes out.
-
The same result
Whatever the AI
Invoice from Salty Swiss SA
- Kokonut Island (FYSK models) 6510
- Infomaniak or Mistral AI 6510
6510 Telephone, Internet
Compatible with RTX Spark chips
and other Blackwell, Hopper and Ada chips
SDK coming soon
Your software, connected to Kokonut Wave.
Add accounting AI to your software in a few lines of code, without building or training a single model.
For software publishers// npm install @kokonut/wave
import { Wave } from '@kokonut/wave';
const wave = new Wave({ url: 'https://wave.local' });
// Un lot de pièces pour un dossier client ;
// Wave a déjà son Kokonut Drop File
const batch = await wave.batches.create({
clientFile: 'exemple-sa',
documents: ['facture-0412.pdf', 'quittance-0413.pdf'],
});
const entries = await batch.entries();// composer require kokonut/wave
// .env : KOKONUT_WAVE_URL=https://wave.local
use Kokonut\Wave\Facades\Wave;
// Un lot de pièces pour un dossier client ;
// Wave a déjà son Kokonut Drop File
$batch = Wave::batches()->create(
clientFile: 'exemple-sa',
documents: ['facture-0412.pdf', 'quittance-0413.pdf'],
);
$entries = $batch->entries();# pip install kokonut-wave
from kokonut_wave import Wave
wave = Wave(url="https://wave.local")
# Un lot de pièces pour un dossier client ;
# Wave a déjà son Kokonut Drop File
batch = wave.batches.create(
client_file="exemple-sa",
documents=["facture-0412.pdf", "quittance-0413.pdf"],
)
entries = batch.entries()// dotnet add package Kokonut.Wave
using Kokonut.Wave;
var wave = new WaveClient(url: "https://wave.local");
// Un lot de pièces pour un dossier client ;
// Wave a déjà son Kokonut Drop File
var batch = await wave.Batches.CreateAsync(
clientFile: "exemple-sa",
documents: ["facture-0412.pdf", "quittance-0413.pdf"]
);
var entries = await batch.EntriesAsync();# Un lot de pièces pour un dossier client ;
# Wave a déjà son Kokonut Drop File
curl -X POST https://wave.local/v1/batches \
-F client_file=exemple-sa \
-F documents[]=@facture-0412.pdf \
-F documents[]=@quittance-0413.pdfExample: create a batch, provisional API