Lib TQ_Engine new version

Lib TQ_Engine new version

Lib TQ_Engine new version

Lib TQ_Engine new version

Lib TQ_Engine new version

Lib TQ_Engine new version

Refactor: sync optimized search core and python bridge

feat: integrate ultra-optimized TQ SIMD core with prefetching, LTO, mimalloc and memory reuse

Speed: Fixed duplicate tasks and cleaned up parallel Groq analysis

Speed: Parallelized Groq analysis for 3-4x faster simulation

Fix: Robust context preparation to avoid AttributeError

Fix: Defined missing batch_start variable in process_batch

Cleanup: Removed redundant session loading logic from chat endpoint

Optimize: Simulation mode now uses system-only scope for speed and stability

Fix: Optimized search scope and parameters for Simulation mode

Fix: Embedding translated queries instead of raw Vietnamese

Fix: Explicitly check stream is False for simulation mode

Fix: Restored prompt variables and added robust stream extraction

Fix: Final streaming fix with metadata sync and key stripping

Security: Fixed user/session isolation in DynamicBatcher

Fix: Final prompt NameError fix and context sync

Fix: Pass context-aware prompt to AI and optimize streaming padding

Fix: Strip whitespace from Groq API keys to prevent illegal header errors

Fix: Truly bypass LLM Answer generation in Simulation mode

Optimize: Migrated analysis to Groq via httpx with mandatory batching and key rotation

Final: Fully unified batching architecture and fixed prompt NameError

Cleanup: Unified chat flow to always use batcher and removed redundant code

Optimize: Query analysis now uses HF Inference API for better stability

LLM: Set OpenRouter 120B as primary provider with HF/Groq fallbacks

Speed: Use HF Inference API for ultra-fast LLM responses

Fix: Correct DB migration timing after cloud sync

Fix: Robust database migration for messages table

Optimize: Unify single and batch requests through DynamicBatcher for maximum stability

Optimize: Implement query batching (10 queries per prompt) for simulation

Optimize: Add semaphore to limit concurrent Groq requests

Security: Isolate chat history by user_id and session_id

Fix: Resolve syntax error in chat endpoint

Optimize: Reduce Groq timeout and fix real latency display

Fix: Final architecture for immediate streaming and robust LLM fallback

Fix: Explicitly disable proxy buffering with headers

Fix: Add padding to stream to bypass proxy buffering

Fix: Force Groq client to bypass HF proxies and IP blocks

Log: Add detailed Groq error tracking and increase timeout

Optimize: Standardize Groq client and analysis mode

Fix: Correct syntax error and finalized search hydration logic

Fix: Resolve hydration failure by preventing source labeling conflict and simplifying main loop

Log: Standardize query analysis logs (Mode & Key)

Cleanup: Remove misleading Native SIMD warning

Optimize: Fix Rust build and improve large ID database retrieval

Debug: Add file size check for system database