fix(ig): retry on connect timeout

fix(scheduler): use stdlib zoneinfo

fix(scheduler): rescheduling now works on HF Spaces

fix(scheduler): never block API event loop — mutex + thread pool

feat: post type respects enabled_formats

feat: trend filter, engagement metrics, duplicate protection, watermark consistency

fix: Cloudinary async upload, calendar drag reschedule endpoint

feat: HF keep-alive, IG token reminder, auto hashtags, budget alerts, hashtag suggestions

feat: config export/import seed endpoint

fix: content planner uses provider key system

feat: topic-aware image generation

fix: remove side-slice layout

fix: random music selection per post

fix: remove neon grid lines

feat: publish mode picker — normal post or reel

fix: carousel parser and prompt format

fix: carousel parser — strip bold labels, fix angle instruction leak

fix: carousel — fallback parser, count=1 bug, debug logging

feat: post variety — 6 content angles, random layouts, varied image styles

fix: carousel — robust slide parser + create_carousel_slides overlay

fix: respect saved model priority order for Anthropic

feat: per-category model priority — reorder text fallback queue from Settings

feat: configurable posting times (IST), reels only for carousel posts

fix: pin scheduler timezone to Asia/Kolkata (IST)

feat: granular agent sub-steps, immediate WS push, all backend fixes

feat: surface agent auto-improve steps to progress bar

diag: log exception type and IG API response to debug publish failure

fix: replace LangGraph with direct async LLM call for reliable timeout

fix: remove left green accent bar, rebrand watermark to dailywireup

fix: replace langchain with direct httpx call for reliable LLM timeout

fix: use async ainvoke so timeout actually cancels hung LLM calls

fix: per-statement transactions and 30s budget for triggers install

fix: set statement_timeout at connection level + add 60s migration budget

fix: set 30s statement_timeout per migration to override Supabase default

fix: remove vector ALTER TABLE from migrations (columns exist, causing timeouts)

fix: run each migration in its own transaction to survive statement timeouts

fix: reduce LLM timeout to fail faster and try next provider

feat: viral news-style prompts and dailywireup branding on overlays

fix: re-enable text overlay on text_image posts

fix: unpack _build_attempts dicts correctly in post_improver

fix: normalize api_base URLs to ensure https:// prefix

fix: add circuit breaker to embeddings service to silence 401 spam

fix: make column migrations non-fatal so server starts even on timeout

fix: add DB connection/command timeout for Neon cold start

fix: add DB connection timeout to prevent hanging on Neon cold start

fix: reset stuck generating posts on startup after server crash

Initial deployment to HF Spaces