Update Dockerfile

Update Dockerfile

Update Dockerfile

Update Dockerfile

Update Dockerfile

🔥 GOD MODE ACTIVATED: Subprocess monitoring, Context Guardrails, and No-MMAP Turbo Boost

🚦 TRAFFIC JAM HACK: Added Parallel Processing & Invisible Keep-Alive Heartbeats

🧠 Expand Context Window to 8192 tokens for long documents

🔧 Fix llama-server flag: specify --flash-attn on

🚀 PERFECT OPTIMIZATION: KV Cache 8-bit Quantization & Flash Attn for max CPU speed

🔧 Remove unsupported --image-max-pixels flag

🔧 Fix OOM crash: reduce memory usage and image bounds

🔥 FINAL ULTRA SYSTEM

🔥 SPEED OPTIMIZATION V3

Final performance tuning: threads 2/4 + mlock

Use Python script for reliable model download

Switch to reliable huggingface-cli for model download

Fix flash-attn syntax and revert to stable llama.cpp

Deploy nuclear performance optimizations

Revert to working Q5_K_XL model with optimized performance settings

Ultimate speed optimization: Q4_K_M model + tuned threads and cache

Revert to official llama.cpp (fixes mmproj gemma4v error) + keep speed optimizations

Remove unsupported -ctxcp argument for ik_llama.cpp

Apply major performance optimizations for model and API

Add CA certificates and enable CURL in llama.cpp build

Enable SSL support in llama-server for HTTPS image downloads

Switch to manual CMake build of llama-server for stability

Fix llama-server startup command

Switch to llama-cpp-python and fix build

Add all project files for Gemma 4 deployment