Skip to main content
FKFK Trading
OnlineFK

Local AI · Offline Mode

On-device assistance only where the specific browser and hardware can run the downloaded model.

Mode: Cloud OpenRouter
Local runtime
Rule-based
Probing runtime...
Cloud status
Online
OpenRouter mock ready
Model download
Not cached
Consent required (12 MB RAM)
Storage available
Probing...
OS dependent

On-device assistant

Rule-based offline
Rahim ke 10 bag Shah cement dao
Only locally available product lookup, intent extraction, and draft preparation. No remote inference while offline.

AI capability status

Product searchAvailable locally
Sale intent extractionRule-based offline
Complex cloud analyticsCloud OpenRouter
Cloud image visionCloud OpenRouter
OpenRouter callsMock adapter ready

Model management (Consent & SHA-256)

Measured Device Matrix & Model Families
FK Tiny Intent ONNX/WASM Pack v1.212 MB RAM
Recommended for Android & iOS PWA: tiny WASM-compatible lexicon/weight classifier (~1.9 KB artifact, ~12 MB runtime RAM), auto-fails over to rule parser if storage or RAM constrained.
SmolLM2 360M Q4F16 (WebLLM / WebGPU)520 MB RAM
High memory cost (~380 MB weights + ~520 MB VRAM/RAM). Unsupported on low-RAM mobile browsers or iOS WebKit memory caps; benchmarked for desktop WebGPU only.
Pending physical iOS/Android UAT