feat: multi-region parallel search for top stories

feat: 100% live search-only top stories with multi-query parallelization

fix: guarantee 6 top stories with 30-day fallback and regional expansion

feat: guarantee 6 top stories with vector DB fallback

feat: refine search query and increase top stories limit to 6

fix: resolve Depends NameError and ensure logging integrity

feat: visual top stories, redis caching, and automated 4-hour refresh

feat: implement 'latest=3 days' logic and Jina content date extraction

feat: implement aggressive temporal weighting and 14-day default window for news relevance

feat: implement semantic TL;DR citations and live image support

fix: implement missing _get_history_text in RagChatUseCase to fix 500 error during stream

fix: remove Transfer-Encoding chunked (illegal in HTTP/2) from SSE response headers

fix: use snapshot_download to avoid OOMKilled during model pre-cache build step

fix: add X-Accel-Buffering header to fix ERR_HTTP2_PROTOCOL_ERROR on HF Spaces SSE stream

fix: generate real follow-up questions instead of q1/q2/q3 placeholders

chore: add .gitignore, remove __pycache__ and temp files from tracking

fix: structured output prompt with example + DDG timeout 4s

feat: structured LLM output + clean content for LLM

fix: live doc dedup + Jina boilerplate stripping

fix: NewsAPI query precision + non-news domain filtering

fix: temporal queries now prioritize live results end-to-end

fix: configure root logging so INFO logs appear in HF Space

feat: detailed step-by-step pipeline logging

feat: hybrid RAG pipeline upgrade

feat: enhance DuckDuckGo live search with smart Ethiopia filtering

fix(reranker+jina): auto-fallback to CrossEncoder + silence Jina noise

fix(3-bugs): Jina 401, reranker tokenizer, intent classifier v1 still running

feat(query-enhancement): upgrade LLM prompt + replace print with logger

feat(intent): add 4-provider fallback chain for intent classification

feat(intent): replace hard-coded rules with LLM-powered classifier v4

feat(intent): complete rewrite of intent classifier v3

feat(language-gate): detect non-English queries and return polite response

fix(critical): identity leak, intent misclassification, reranker tokenizer

fix(prompt): prevent hallucinated citations - enforce strict source index

fix(rag-api): session ID merging + FOLLOW_UP streaming leak

fix(rag-api): add msgpack and confluent-kafka to requirements, safe import

feat(rag-api): hybrid top stories — 3 from Kafka news.processed + 3 from DuckDuckGo

fix(rag-api): relax keyword overlap filter + cache DeBERTa intent results

feat: numbered citations [1][2][3], follow-up questions, focus modes

feat: add top stories API + fix reranker tokenizer

feat: integrate Jina Reader for full article extraction (71x content boost)

fix: Improve DuckDuckGo error handling + deployment tools

fix: Remove Qdrant order_by and upgrade reranker dependencies

fix(news): Add time-based filtering and cache-control to /latest endpoint

feat: Add query enhancements and flexible prompting (v2.1)

fix: handle None values in hybrid result deduplication

fix: correct DuckDuckGo API call and database search execution

fix: make _build_context and execute_chat async for hybrid search

fix: use await instead of asyncio.run for hybrid search

fix: add missing logger import in rag_chat_use_case