B2B Enterprise Framework

Smart Kitchen AssistantNext-Gen Edge AI

The definitive open-source methodology for offline appliance intelligence. Featuring Acoustic Shields, 100% Visual Ground Truth, and Zero-Latency Hardware Integration.

B2B Licensing

Experience Live Edge AI

Select a scenario below to see the exact B2C user experience in real-time.

Step 1: Test Vision AI

Select Scenario. Extract Recipe.

Experience the Yemek AI Vision Technology. Select any food scenario below and watch the AI extract a complete recipe instantly.

Select a scenario to run Vision AI simulation:

Heart-Shaped Eggs

Currently Selected

Spaghetti Bolognese

Run Extraction Simulation →

Turkish Adana Kebab

Run Extraction Simulation →

Step 2: Edge Voice AI

Easy5 mins

Heart-Shaped Sunny-Side Up Eggs

A visually appealing breakfast classic formed perfectly using a silicone heart mold.

Required Commands

"Next""Back""Repeat""Stop"

No Cloud Required. Disconnect your Wi-Fi to test offline Kitchen Mode.

1. The HFSCA Architecture

Hands-Free Semantic Culinary Assistant methodology completely decouples intelligence from the cloud, eliminating latency and server costs.

Acoustic Ignore Shield

Solves the "echo loop" in high-reflectivity kitchens without hardware interrupts. By locking the isSpeakingRef, the system safely ignores its own voice output.

Hardware Persistence

Bypasses native Garbage Collection (GC) via immutable memory boundaries (utteranceRef), maintaining session state even during prolonged listening gaps.

0 USD Inference Cost

Running entirely on the client-side appliance NPU or browser layer, preventing linear server scaling costs for manufacturers. Zero API dependencies.

Vision Pipeline

Reverse-Engineering Reality

Standard AI hallucinates by generating text first, then trying to force an image to match. Yemek AI completely flips the generative hierarchy.

1
Visual Ground TruthThe high-fidelity image is generated OR captured by the fridge camera first. It is the absolute physical reality.
2
Multimodal ScanningThe Vision-LLM scans the image pixel by pixel to deduce ingredients, textures, and plating techniques.
3
Recipe HarmonizationThe NLP engine writes instructions strictly based on the visual evidence. 100% Zero-Hallucination output.

pipeline_execution.log

> INITIALIZING REVERSE PIPELINE...

graph TD
  A[Camera Input] --> B(Visual Ground Truth)
  B --> C{Vision Layer Analysis}
  
  C --> D[Identify: Pan-Seared Salmon]
  C --> E[Identify: Asparagus Base]
  
  D & E --> F[Recipe Harmonization]
  F --> G(Zero-Hallucination Output)

> PIPELINE COMPLETED. 0 ERRORS.

Dual-Licensing Strategy

Our core architecture is open-source under AGPL-3.0. This mandates that any enterprise embedding our system in proprietary smart appliances must open-source their entire stack.

Enterprise / Commercial SDK

To bypass the AGPL-3.0 mandate and securely integrate the Smart Kitchen Assistant into closed-source IoT firmware (e.g., Oven MCU, Smart Fridge OS), tech manufacturers must acquire a B2B Commercial License.