Merge agent/sync-owner-files

Sync AGENTS.md and opencode.json with owner versions

Merge agent/replace-agents-md

Replace AGENTS.md with owner version (Tests, Before writing new code)

Replace AGENTS.md with owner version

Merge config-sentry-effort: subagents, sentry MCP, effort routing

Config: quick/deep subagents, sentry MCP, effort routing rules

Merge fix-banner-centering: optically center tracked banner titles

Banner: optically center tracked titles (letter-spacing compensation)

Merge fix-adversarial-wave1: adversarial fixes, hardened tests, comment cleanup

AGENTS.md: comment style rules (Russian, short, why-only)

Comment cleanup: short Russian why-comments, no AI boilerplate

Subagent re-review: fix confirmed misses, harden thin tests

Independent re-review: fix missed regressions and close gaps

Fix regressions found by adversarial review: routing, rendering, rate limits

Docs and tests wave 7: routing description matches reality, stronger assertions instead of or-chains, document the deliberate drop_pending_updates tradeoff

Cleanup wave 6: one shared SSE parser and one shared fallback chain, drop dead code and a silent list alias, cover the previously untested failure branches

Quality wave 5: summarizing history trim in streaming, LaTeX parity between render paths, rate limit before state/album downloads, parity tests

Medium wave 4: validated env numbers, Upstash backend and shutdown flush tests, deno check/lint in CI, reproducible pip step, docs for missing variables

Reliability wave 3: router skips quota-exhausted models, prune never drops a chat holding the lock

Security wave 2: gate deploy on push+own repo, 401 and denial logging on admin endpoints, FastAPI schema off, proxy body cap, ownerless pick confined to chat, loud warning about secrets derived from BOT_TOKEN

Fix chat-lock hangs: TTS async with timeout, stream handshake capped, route attempts bounded by remaining budget, one shared chat-lock limit

Split glued 3-item bullet lists in both render paths (prod 26.09.2026 threshold miss)

Continue feature: resume interrupted stream answers on bare continue request

Reliability: cap malformed-call retry by route deadline, service quota separation, album buffer cleanup

Security: validate pick before pop (foreign tap no longer burns token), SSRF guard on download hosts

UX batch: start examples, pick reissue on expired tap, TTS multi-part, reset doc fix

Prompt: informal ty covers verb forms too (prishli, not prishlite)

Claude-parity fixes: callback filters for pick buttons, inline numbered-list splitter in both render paths, prompt ban on glued lists

Luna audit round 2: review fixes (regex bounds, mixed albums, redaction guards)

Luna audit fixes: album lock, multilingual routing, streaming budget, quota and persistence hardening

Merge media-fail-honest: honest media errors and media entry log

Fix sticker test lang expectation

Media drops are honest: fetchable attachment failure replies error, entry log for media

Formatting like Claude: one list item per line, prompt rule plus post-fix

Review fixes: real audio suffix, no RU marker, transcript cap

Ideal routing: voice transcribed into scenario routing, OR qwen newcomer in light chain

stats: active chats counted over trailing 24h, total kept

Qwen audit fixes: aio Gemini client, snapshot-before-thread state flush, drop dead Fish check, utcnow fix

Merge album-parallel: download album extras in parallel

Stale text: drop dead /model ref from prompt, fix test counts and TTS order notes

Perf: download album extras in parallel, keep message order

Merge guest-hint: DM hint in all 25 langs instead of bare placeholder

Style: informal ty in Russian tiktok strings, regression test

Fix guest hint test: default chat lang is English

Guest TikTok: DM hint in all 25 langs instead of bare placeholder

TikTok music: sound-author handle stripped, performer is sound author, artist suffix trimmed

TikTok music: retry send_audio without thumbnail on failure

Merge titleswap fix

WIP thumb fallback