Update LLM system prompt: Add FORMAT E for structured multi-level topics

Replace all em dashes with hyphens in llm.py

Fix acceptability retrieval; shorten redirect responses

Fix retrieval for prorata queries

Improve bot: self-identity, non-English redirect, privacy-tone for personal data, no hallucination on people/org, varied fallbacks

fix(llm): stop AI from addressing the user by name

fix(backend): allow renaming sessions before first message is sent

fix: resolve admin refresh logout issue by checking credentials dynamically

Hardening admin security: moving credentials to environment variables

fix: implement session merging to allow multiple logins for the same user without splitting history

Fix: Broad HR queries + Anti-hallucination hardening (no fake careers page)

Fix: Conversational Intelligence - Natural greetings, varied responses, no robot loop

Production Intelligence: Graceful Guidance + Full Intent Coverage + Urdu Support

Upgrade: Hybrid Intelligence (LLM Query Rewriter + Local Expansion)

Upgrade: Intelligence & Reasoning (Added Format D + Expanded Intent Map)

Fix follow-up typo handling (then/them) + prevent specific leave type listing other types

Add conversational follow-up resolver: pronouns (them/it/this) now resolve to last topic for accurate retrieval

Hardening RAG Intelligence: Restoration of perfect format decision table, optimized thresholds, and query expansion logic for GPT-4o-mini

feat: institutional-grade intelligence upgrade with hybrid expansion, logical deduction, and core reference data

Restore: Revert codebase to stable May 14th 6 PM state

feat: hardcode leave entitlements in system prompt for guaranteed completeness

feat: reach 100% intelligence with broader retrieval and mandatory deduction rules

feat: improve list completeness and add conclusion logic

feat: further intelligence boost via multi-intent matching and relaxed rerank thresholds

feat: increase HR assistant intelligence via logical deduction and broader retrieval

Add OpenAI integration while preserving stable intelligence - 2026-05-15 18:05

Restore to latest backup from 13 May - 2026-05-15 17:56

Finalize Export Chat button text and remove debug logs

Enhance sidebar export with modal, radio options, and ZIP bundling

UI: Further decrease export dropdown hide delay to 300ms

UI: Decrease export dropdown hide delay to 800ms

UI: Improve export dropdown hover behavior with 1.5s delay and smooth transitions

Polish: Add step numbers to structured list format

Intelligence: Remove all hardcoded format rules, let LLM choose format by intent

Generic: Replace hardcoded intent map with universal keyword extractor, fix 'can i resign'

Fix: Disable stale cache, add ul/li CSS reset, unify post-processing with _clean_html

Conciseness: Enforce 3-point limit for leaves and ignore secondary clauses

Intelligence: Concise-first prompt, context filtering, smart format selection

Refactor: Remove hardcoded topic defaults - make LLM fully context-driven

Intelligence: Add Format D for yes/no, consequence, and conversational questions

Fix: Apply br cleanup to OpenAI caller (active provider)

Fix: Remove stray <br> before </li> and collapse double <br> in output

Restore: Premium HTML list formatting for structured responses

Fix: Replace LLM expander with instant hybrid normalizer, fix re-index, fix prorata

Optimization: Lowered relevance threshold for better recall

Performance: Professional batched search and latency optimizations

Fix: Prioritize original query and enhance compound word analyzer for prorata

Fix: LLMService initialization TypeError and provider synchronization

Upgrade: Professional intelligent retrieval pipeline with robust tokenization and LLM query expansion

Fix: Add specific intent rules for 'half day' and 'unauthorized' leaves to prevent generic catch-all matching