feat: enhance RAG intelligence - intent-first protocol, /no_think for Qwen3, smart rerank fallback, context word cap

Push requested version of llm.py with monologue suppression and history optimization

Hard revert of llm.py to state 034bc73 as requested

Reverting to previous stable state after failed specificity experiment

Enforce high specificity and strict question answering in system prompt

Optimize token usage to resolve 413 Payload Too Large error

Strip internal reasoning monologues and increase token headroom

Optimize LLM conciseness and fix response cutoff

Fix Groq 413 Payload Too Large by strictly capping context and tokens

Refine for Extreme Conciseness: enforce word economy and dense summarization

Optimize for Intelligent Completeness: 10 context chunks and dense system prompt

Fix latency: switch to bge-small and bge-reranker-base for CPU optimization

Fix bot behavior: handle greetings, strip reasoning monologues, and sanitize history

Fix Groq 413 Payload Too Large by reducing max_tokens and context chunks

Update Groq Qwen model ID to qwen/qwen3-32b

Switch entirely to qwen-2.5-32b due to DeepSeek-R1 string deprecations

Update DeepSeek-R1 model alias to deepseek-r1-distill-qwen-32b

Update decommissioned DeepSeek-R1 model alias to specdec

Fix Groq 400 Bad Request for DeepSeek-R1 by adjusting temperature

Fix ui_gradio LLMService initialization and update README

Add rank_bm25 to requirements

Implement DeepSeek-R1, BGE-Large, and Hybrid Search

Update SYSTEM_PROMPT to use bullet points conditionally

Fix LLM hallucinations with strict prompt and 0 temperature

Surgical Formatting Fix: increased heading boldness and removed raw '*' symbols from chat display

Reverted UI changes to restore original alignment and icon sizing

Fixed widget alignment: corrected icon sizing, implemented CSS reset to prevent theme leaks, and ensured proper positioning

Revamped Chat UI: Premium glassmorphism design, smooth animations, and proper HTML/Markdown rendering for bold and lists

Implemented HTML Bolding protocol and strictly prohibited hallucinated HR categories

Implemented Precision Protocol: removed headers and enforced strict, bullet-only responses to maximize conciseness

Implemented the Goldilocks Protocol: optimized for answers that are both complete and concise

Implemented Strict Ground Truth protocol and synonym-aware retrieval to prevent hallucinations in HR answers

Upgraded RAG pipeline for ChatGPT-level depth and intelligence (Multi-query, larger context, smarter prompts)

Add LFS tracking for PDFs and retry push

Pushing all files including docs and index from backup

Upload 47 files

wipe space

Restore and push every single file from the final backup

Implement Clarification Intelligence: Ambiguity detection and multi-turn clarification protocol

Fix RAGPipeline instantiation in ui_gradio.py

Implement Deep Intelligence: Query Expansion, Reranking, and High-Precision Chunking

Fix Admin sidebar sync and increase relevance threshold to 0.45

Update assistant identity and push

Peak Synthesis Protocol and Universal Multi-User Name Fetching

Peak Intelligence, fix truncation, and 5s name-fetch delay

Hard ban 'handbook' and 'documents' words from AI

Enhance name discovery with sessionStorage and push

Fix truncation and enforce strict 30-word limit

Fix assistant title to Martechsol Assistant and push

Enforce strict 30-word limit and intelligent summarization