Spaces:
Paused
Paused
Commit History
-
Mozg odpowiada tylko na podpisane wywolania; bez placu zabaw; awaria to blad, nie pozegnanie
- respond(context_json, sig): HMAC-SHA256 z ARIA_BRAIN_SECRET, fail-closed, limity rozmiaru - _try_it usuniete (publiczne wywolania modelu na kredyt organizacji) - jezyk po granicach slow, honorowany wybor z Workera; role tylko user/assistant - blad inferencji rzuca gr.Error zamiast udawac wyjscie ARII - gradio i huggingface_hub przypiete; README bez hf_oauth8f9bde4
-
aria: block = silence + "can't reply" composer + cheeky unblock comeback; mute unchanged; raise limit to 50; drop usage countdown
Malina +1: Claude Opus 4.8- block: DO tracks blocked (loaded from D1 + WS block/unblock signals); stays silent incl. proactive while blocked; on unblock runs a short cheeky line (persona ODBLOKOWANIE, model-generated/varied). Client hides ARIA output when blocked and swaps the composer for "Nie możesz odpowiedzieć na tę konwersację". - mute unchanged (silent, no comeback on unmute). - budget 20 -> 50 messages / 5h for everyone. - remove the "x/5h left" presence indicator (just shows Active). - css/js v20, aria-settings v5. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>f4410c4 -
aria: profile-picture avatar; come-back-after-limit catch-up; email in usage + read-by-email views
Malina +1: Claude Opus 4.8- avatar: replace the "A" letter avatars (header, typing row, ARIA messages) with profile-picture-aria.png (resized 256px) via CSS background. - limit behaviour: while over budget ARIA is "away" — one goodbye, then silence, but the user's messages are STORED (answered=0, don't burn budget, capped). On her return (reconnect after the 5h window resets, or the next message) she runs a single catch-up turn (comeback=true) that apologises in her own words and answers the backlog. New DO methods: maybeCatchUp/runCatchUp/unansweredUserEvents. - DB: email column on aria_usage (kept fresh by the worker, backfilled) + views aria_conversations_by_email / aria_messages_by_email to read any user's chat by email. email now flows Pages → aria-worker (WS query + X-Aria-Email header). - persona: POWRÓT (comeback) section, in her own words, varied. - css/js v17. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>57b6530 -
aria: examples-are-examples (less robotic repeats); fix missing proactive pings; unify reply timing; drop ↩ from reply menu
Malina +1: Claude Opus 4.8- persona: strong "examples are only examples, not rules — vary, invent your own, never repeat the same line like a bot" note in the system prompt; de-anchor the user-reaction comment (was always "ej no to nie było aż tak śmieszne"). - proactive/idle fix: CISZA rule rewritten to actually re-engage when the topic is still open (varied wording), only staying silent when the topic was closed or she just pinged. IDLE_MAX_PER_TURN 1 → 2 (a second, later nudge). - timing: credit the model's thinking time against the first bubble's "typing" pause (total ≈ max(modelTime, pause), not modelTime + 1–8s) across normal/idle/react. - UI: reply menu item is just "Odpowiedz" (no ↩ glyph). js v16. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>8eee93b -
aria: interruptible delivery — adapt when the user cuts in mid-reply
Malina +1: Claude Opus 4.8If the user sends a new message while ARIA is still sending a multi-bubble reply, she now finishes the current bubble, holds the unsent remainder as a resumable thread, and (after the user's burst settles via the existing debounce) re-decides with the brain: answer the new message first, then resume that thread in the same turn or drop it — her call. - DO: emitBubbles is interruptible (pacedDelay/wakeDelay); returns the not-yet-sent bubbles, stored in this.unsent. Interruption during "thinking" holds the whole stale reply instead of sending it. unsent is carried into the next turn's context. - worker/orchestrator: pass unsent → ctx.interrupted. - persona: PRZERWANY WĄTEK section — model decides to weave the thread back (own words) or let it go; never explains the mechanic. - remove now-unused sleep() helper. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>84c5d94 -
aria: fix iOS WS dropout + reply-to stripping; double-tap heart; drop swipe arrow
Malina +1: Claude Opus 4.8- WS resilience (iOS "message sent, no reply" bug): client heartbeat with liveness watchdog (zombie socket → reconnect), reconnect on visibility/online/focus, and an outbox that resends unacked events. DO: idempotent receive (dedup + ack), echo to all sockets incl. sender so a resent message re-renders after a reconnect wipe. - reply-to fix: DO no longer strips a reply_to that points at the last user message (that silently killed "reply to my message" + emphasis quotes). Persona: explicit reply requests always set reply_to. - double-tap on ARIA's bubble → ❤️ reaction (toggles), iMessage-style. - remove the ↩ arrow glyph shown during swipe-to-reply. - cache-bust aria.css v12 / aria-chat.js v13. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>29bf7f3
-
aria: iOS keyboard viewport fix + more reactions/reply-to/bubbles + time awareness
Malina +1: Claude Opus 4.8- mobile: track visualViewport (height + offsetTop via translateY) instead of fighting iOS scroll; app glues above the keyboard like iMessage, messages stay bottom-pinned. Stop closeHold() on every viewport tick (no more self-closing picker). - persona: reactions ~every 2-3 fitting msg (❤️ most common); reply-to used naturally/often (quote a specific msg, return to a thread, emphasis); encourage splitting into 2-4 bubbles (up to 6). - time awareness (Europe/Warsaw): worker adds now_text + per-message send time; persona gains a CZAS section + gap detection so ARIA senses the clock, send times and long absences ("dawno cię nie było", "już po północy") without ever stating the time like a robot or mentioning it unprompted. - cache-bust aria.css v11 / aria-chat.js v12. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>8d1f0ef -
system prompt update
Verified signatureb3d60e7 -
aria: reconcile reaction set (❤️😆😮🥹😢😡) across worker+brain+client; instant keyboard fit; brb-suppression + model-judged idle; emotional-tone continuity; tip under input bar
Malina +1: Claude Opus 4.8- events.js ALLOWED_REACTIONS + build_user_frame contract now match user's new emoji set (schema.py/aria-chat.js already had it) - keyboard: replace 630ms pingViewport crawl with snap+rAF+single reconcile (instant) - DO: 'poczekaj/muszę lecieć' suppresses idle pings; idle now routed through the brain (mostly []), no canned 'jesteś?' - persona frame: hold emotional register (no empathy->jokes whiplash) + proactive caring support; model-judged silence block - aria.html/css: 'Tip - sprobuj przytrzymac wiadomosc hehe' under input bar; cache-bust v10/v11 Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>4a9c791 -
TEST persona update v2
Verified signature4eba183 -
TEST persona.py update
Verified signature945c87e -
inne emotki
Verified signatureef5a0f7 -
aria: smart clustering (wait for multi-bubble buildup), per-bubble relevance/threading, reply-to flow fixed, interchangeable XD/hahah + hmm/wtf/no wayyyy
Malina +1: Claude Opus 4.8- DO: adaptive debounce — question→~1.2s, lead-in/2nd-quick-bubble→wait up to 10s (reset per bubble) so 'dobra/już ci tłumaczę/no więc...' all land before reply; enforce reply-quote only when not answering the immediate last message; drop dumb 'no mów' fallback (model guesses); add ts→'Xs temu' per message in context - brain: schema now passes reply_to through (was dropped → quote NEVER showed); build_user_frame adds id-map + threading rules (recognize follow-ups of the user's OWN earlier msg, decide to reply or stay silent, set reply_to to the specific msg when not the last); SYSTEM_PROMPT: XD↔hahah, XDDDD↔hahaahahah, hmm/wtf/no wayyyy; 'byt'->'inteligencja'; per-turn frame moved into system msg (no more answering the instruction itself) Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>3f0378a -
aria: fix instruction-leak (frame+lang moved into system msg, convo ends on real user turn); react ~every 3rd msg; 'byt'->'inteligencja'; 400-char limit
Malina +1: Claude Opus 4.8- orchestrator: per-turn frame + language directive now live in the SYSTEM message instead of trailing user turns -> ARIA stopped answering the instruction itself ('no wiadomo że po polsku'); reaction-only turns get a tiny synthetic user turn - persona: reactions encouraged ~every 3rd emotional message (as an add-on, not instead of the reply); SYSTEM_PROMPT 'byt' -> 'inteligencja' - message length cap 100 -> 400 (server events.js, input maxlength, client) Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>2e61db0 -
Update brain/orchestrator.py
Verified signature18d0fde -
brain: switch to aya-expanse-32b (your 32B/gemma-4 weren't available; gemma-4 broke Aria entirely)
Malina +1: Claude Opus 4.8- model: CohereLabs/aya-expanse-32b (works, strong Polish; kept bill_to=NextNeural) - restore reactions in build_user_frame (SYSTEM_PROMPT untouched), de-emphasized so the model still replies with words; both directions (Aria reacts + comments user reactions) - reply context + silent gender (never verbalizes the assumption) - robust output parsing: accept content/text/message keys; split multi-sentence replies into separate lowercase no-period bubbles (chat style, model-agnostic) Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>5ea523e -
zmiana na google/gemma-4-31b-it
Verified signature11e098d -
fix zmiany modelu
Verified signature09c43be -
Model change to Qwen/Qwen2.5-14B-Instruct
Verified signatureb6af9cd -
Update sys prompt
Verified signature77903ac -
update prompt
Verified signatureed80fe0 -
Tescik 2
Verified signature113d2e6 -
Update prompt - role.
Verified signature3c17d5f -
Tescik
Verified signaturefec9e45 -
Update system promptu
Verified signaturec4fcbc1 -
Nowy prompt
Verified signature1a43bda -
aria: recognize user gender from their name + address by name sometimes
Malina +1: Claude Opus 4.8Pass the signed-in user's name down Pages->worker->DO->brain context. Persona infers gender from the name (Polish -a = fem etc.), uses correct gendered forms toward the user, and addresses them by name occasionally (skips weird non-name nicks). Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>94eb07c -
aria: warmer/more caring persona (not overdone); mobile chat full-bleed on all phones
Malina +1: Claude Opus 4.8- persona: less nonchalant, genuinely caring/attentive (listens, asks, supports); people-first on personal topics; explicitly not saccharine/clingy - css: .aria-main full width by default; the centered 430px phone frame applies only >=640px, so phones no longer show side background bars Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>10f3caa -
aria: stop leaking internals — event ids + mechanics no longer reach the user
Malina +1: Claude Opus 4.8- persona: ids are internal (target_event_id only, never in text); never mention JSON/format/bubbles/system prompt/'must reply in format'; react to unknown gibberish like a human, don't explain mechanics - schema: strip any leaked evt_/sys_/bub_/ac_ token (incl. quoted) from bubble text - persona: react a bit more often when there's emotion; use XD/lmao subtly (light topics only) - DO: never drop proactive/reaction bubbles in the echo filter (so idle nudges always persist to memory) Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>5fd78ff -
brain/persona: restore multi-bubble splitting (lost in the 'smarter' rewrite)
Malina +1: Claude Opus 4.8Re-added an explicit, strong rule to split replies into several short messenger bubbles (one thought per bubble, 2-4 for substantive answers) instead of one block. Verified: recursion -> 4 bubbles, smalltalk -> 2, still smart + playful. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>b213a83 -
brain/persona: playful BUT intelligent — stop playing dumb
Malina +1: Claude Opus 4.8Persona was forcing dismissive teen filler ('nwm'/'spoko', ultra-short), which kneecapped the model. New persona: genuinely smart/knowledgeable, gives accurate substantive answers, explains clearly, reasons — while keeping the casual lowercase messenger wit. Verified: solid recursion explainer, real Canberra history, correct riddle, still short on smalltalk. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>8ad3134 -
aria: message reactions — long-press ARIA bubble to react + ARIA auto-reacts to user
Malina +1: Claude Opus 4.8- DB: reaction column on aria_events (ARIA->user) and aria_bubbles (user->ARIA) - DO: handle {t:react} (user), apply 'reaction'-type brain output to user events, broadcast {t:reaction}; history includes reactions - contracts: 'reaction' bubble type + allowed emoji set (👍❤️😂😮😢😡), validated both sides - brain/persona: react sometimes (not always); ❤️ = 'ok/nice' not love; emoji only in reactions - UI: long-press picker w/ animation, animated reaction badge, render+persist+live-sync Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>89b1f9c -
brain: force reply language to match user's last message (was defaulting to Polish on English)
Malina +1: Claude Opus 4.8Add _lang(ctx) (last user message) and append an explicit per-turn directive ('Reply in ENGLISH' / 'Odpowiedz po polsku'); reuse it for signoffs. Verified: EN in -> EN reply, PL in -> PL reply. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>961d1cd -
aria: fix role mixup in rehydrated history; reliable language-matched idle nudges; anti-echo
Malina +1: Claude Opus 4.8- DO: rehydrate context by MERGING user events + ARIA bubbles in time order with correct roles (was loading only user events -> model couldn't tell who said what) - idle nudges generated locally from PL/EN pools, language = user's LAST message, no-repeat; always sends (model sometimes returned nothing) - drop bubbles that echo the user or repeat a prior line - persona: explicit role note + 'never copy the user'; signoff lang = last user msg Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>1308d69 -
aria: randomized two-bubble farewell on inference failure (12 leave + 11 bye per lang)
Malina +1: Claude Opus 4.8Out-of-credit/error now sends a casual two-bubble sign-off (leave line -> bye line), randomized and language-matched, e.g. 'muszę spadać' + 'siemanko' / 'gotta go' + 'catch you later'. DO mirrors it for Space-unreachable. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>98ca451 -
brain: bill inference to NextNeural org via bill_to (was charging personal token owner)
Malina +1: Claude Opus 4.8InferenceClient(bill_to='NextNeural') routes Inference Providers usage to the org instead of the personal account that owns HF_TOKEN. Also unblocks replies since the org has billing/credit. Configurable via ARIA_BILL_TO; requirements bumped for bill_to. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>988cdc2 -
aria: casual sign-off on inference failure instead of robotic error
Malina +1: Claude Opus 4.8On brain failure (esp. 402 no-credit) ARIA now bails like a person — 'muszę lecieć' / 'gotta go for now' (language-matched) — instead of 'ej coś mi się zcięło'. Same for Space-unreachable in the DO (casual bubble, not a red error). Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>466ef03 -
aria: 27B model, messenger style (lowercase, no period), 25s idle nudges, cluster tuning
Malina +1: Claude Opus 4.8- brain: switch to google/gemma-3-27b-it (vision, multilingual, ~3x cheaper than 72B) - persona+schema: human messenger style — lowercase start, no trailing period; idle produces a short varied ping (halo?/jesteś?/you there?), never repeats last msg - DO: idle 25s; inject system_idle into history so brain nudges (not duplicates); dedup idle bubble vs last reply; debounce 900ms so quick msgs cluster Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>8a87a60
-
fix(brain): use Qwen2.5-VL-72B-Instruct (7B-VL not offered by any provider)
Malina +1: Claude Opus 4.8The default model returned 'model_not_supported', producing the 'ej, coś mi się zcięło' fallback for every message. Switch to a vision-capable model that's actually enabled; log the real exception instead of swallowing it. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>8e49975
-
ARIA: event-driven multimodal agent (brain + Cloudflare backend + Messenger UI)
Malina +1: Claude Opus 4.8- brain/: persona, event/bubble contracts, vision orchestrator; app.py exposes api_name=respond - cloudflare/aria-worker: AriaConversation Durable Object, R2 media, usage (20 msgs/5h) - cloudflare/pages-integration: Messenger UI + _worker.js merge snippet - cloudflare/db: additive D1 migration - ARIA_ARCHITECTURE.md Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>4131875