Ship Assistent 0.11.6: opt-in vision, negative pass-through, generate from context.

Co-authored-by: Cursor <cursoragent@cursor.com>
2026-08-22 05:21:48 +03:00
parent 56e089d8da
commit 784426c9bf
11 changed files with 265 additions and 40 deletions
@@ -16,6 +16,8 @@
    const LS_AUTO_APPLY = 'swarm_assistent_auto_apply';
    const LS_AUTO_GENERATE = 'swarm_assistent_auto_generate';
    const LS_AUTO_CRITIQUE = 'swarm_assistent_auto_critique';
+    /** One-shot: turn off auto look_at / auto-critique (vision is button / /look / model look_at). */
+    const LS_VISION_OPTIN_MIG = 'swarm_assistent_vision_optin_v1';
    const LS_AUTO_DOWNLOAD = 'swarm_assistent_auto_download';
    /** When '1', unload chat LLM before Generate (frees VRAM; VL reload can take 1–2 min). Default off. */
    const LS_PARK_LLM = 'swarm_assistent_park_llm';
@@ -89,7 +91,7 @@
 /debug ask — то же + короткий ответ модели
 /why — сразу /debug ask
 /gen — Generate сейчас
-/look generate|refN — прикрепить окно к vision
+/look generate|refN — показать кадр модели (vision)
 /init /mask /clear — Init / Mask / Clear Init
 /interrupt — остановить генерацию
 /aspect 16:9 — размер из таблицы 1K
@@ -656,6 +658,43 @@
        el.dispatchEvent(new Event('change', { bubbles: true }));
    }

+    function liveNegativePrompt() {
+        return String(val('input_negativeprompt') || val('alt_negativeprompt_textbox') || '').trim();
+    }
+
+    function exactDefaultNegative() {
+        const { exact } = resolveExactBundle();
+        const n = exact?.generation?.negative ?? exact?.negative;
+        return n != null ? String(n).trim() : '';
+    }
+
+    function setNegativePrompt(text) {
+        const s = text != null ? String(text) : '';
+        if (document.getElementById('input_negativeprompt')) {
+            setVal('input_negativeprompt', s);
+        }
+        if (document.getElementById('alt_negativeprompt_textbox')) {
+            setVal('alt_negativeprompt_textbox', s);
+        }
+    }
+
+    /** Create / keep / pass-through negative so Generate never drops the Swarm box. */
+    function ensureNegativeForGenerate(patch) {
+        let neg = '';
+        if (patch && patch.negative != null && String(patch.negative).trim() !== '') {
+            neg = String(patch.negative).trim();
+        } else {
+            neg = liveNegativePrompt() || exactDefaultNegative();
+        }
+        if (neg) {
+            setNegativePrompt(neg);
+            if (patch && (patch.negative == null || String(patch.negative).trim() === '')) {
+                patch.negative = neg;
+            }
+        }
+        return neg;
+    }
+
    function isEmptyParamField(raw, { treatZeroEmpty = false } = {}) {
        if (raw == null) {
            return true;
@@ -875,7 +914,7 @@
    /** When the model wrote ### JSON Patch with no fence — build Apply+Generate from prose / last patch. */
    function synthesizePatchAfterEmptyFence(reply, userText, opts = {}) {
        const wants = !!(opts.userWantsGenerate || state.pendingSilentGen
-            || userAsksGenerate(userText) || userAsksContinue(userText));
+            || userAsksGenerate(userText) || userAsksContinue(userText) || userImpliesGenerate(userText));
        const missing = replyMissingJsonPatch(reply);
        if (!wants && !missing) {
            return null;
@@ -930,6 +969,14 @@
        if (patch.aspect) {
            keep.aspect = patch.aspect;
        }
+        if (patch.negative != null && String(patch.negative).trim() !== '') {
+            keep.negative = patch.negative;
+        } else {
+            const liveNeg = liveNegativePrompt() || exactDefaultNegative();
+            if (liveNeg) {
+                keep.negative = liveNeg;
+            }
+        }
        if (Array.isArray(patch.loras)) {
            keep.loras = patch.loras;
        }
@@ -947,9 +994,10 @@
            + '- Structure & front-load: subject → pose/action → body/wardrobe → setting → materials/textures → camera/framing → lighting → medium/mood.\n'
            + '- Expand thin ideas; fix anti-patterns; one coherent scene.\n'
            + '- Keep LoRA trigger phrases in English near the subject they affect.\n'
-            + '- Prefer positives over negatives; preserve meaning and NSFW level from SOURCE.\n'
+            + '- Always include JSON "negative": keep/supplement SOURCE+live negative, or use Exact default if empty. Never drop it.\n'
+            + '- Put “no blur / empty street” ideas as positives in prompt, not as a huge negative dump.\n'
            + '- One short ack in the user language max, then ONE fenced JSON merging these keys: '
-            + `${JSON.stringify(keep)} plus the new English "prompt".\n`
+            + `${JSON.stringify(keep)} plus the new English "prompt" and "negative".\n`
            + '- Include actions:["generate"] when an image was requested.\n\n'
            + `SOURCE:\n${patch.prompt}`
        );
@@ -965,6 +1013,9 @@
            ...base,
            ...effective,
            prompt: effective.prompt || base.prompt,
+            negative: (effective.negative != null && String(effective.negative).trim() !== '')
+                ? effective.negative
+                : (base.negative || liveNegativePrompt() || exactDefaultNegative() || undefined),
            actions: (Array.isArray(effective.actions) && effective.actions.length)
                ? effective.actions
                : (base.actions || ['generate']),
@@ -990,6 +1041,137 @@
        ).test(t);
    }

+    /** Button / /look / «посмотри на кадр» — not casual «смотри какая». */
+    function userAsksLook(text) {
+        const t = String(text || '').trim();
+        if (!t) {
+            return false;
+        }
+        if (/\b(look\s+at|critique|criticize|describe\s+(this|the|ref|image)|what\s+do\s+you\s+see)\b/i.test(t)) {
+            return true;
+        }
+        if (cyrTokenRe('критик[а-яё]*|что\\s+не\\s+так|разбери').test(t)) {
+            return true;
+        }
+        if (cyrTokenRe('опиши\\s+(это|эту|реф|изображ[а-яё]*|картинк[а-яё]*|кадр|результат|референс)').test(t)) {
+            return true;
+        }
+        return /(?:^|[^а-яёa-z0-9_])(посмотри|смотри|глянь)\s+(на\s+)?(результат|картинк[а-яё]*|изображен[а-яё]*|кадр|ген|реф)/i.test(t);
+    }
+
+    /**
+     * Scene / edit / «ещё» — Generate without the magic word «генерируй».
+     * Bare «давай» and trivia questions stay false.
+     */
+    function userImpliesGenerate(text) {
+        const t = String(text || '').trim();
+        if (!t || userAsksNoGenerate(t)) {
+            return false;
+        }
+        if (userAsksGenerate(t) || userAsksContinue(t) || isSameButAspectRequest(t)) {
+            return true;
+        }
+        if (t.length < 8) {
+            return false;
+        }
+        const wantsLook = userAsksLook(t);
+        const wantsRedraw = cyrTokenRe('поправь|исправь|перегенерир|перерисуй|улучши|переделай').test(t)
+            || /\b(fix|redo|redraw|improve)\b/i.test(t);
+        if (wantsLook && !wantsRedraw && !userAsksGenerate(t)) {
+            return false;
+        }
+        if (cyrTokenRe(
+            'что\\s+такое|как\\s+работает|зачем\\s+|какие\\s+(лор|модел|чекпоинт)|'
+            + 'список\\s+лор|где\\s+настрой|что\\s+значит',
+        ).test(t) && !cyrTokenRe('нарису|сгенер|картинк').test(t)) {
+            return false;
+        }
+        if (cyrTokenRe(
+            'нарису|сгенер|перерису|картинк|изображен|'
+            + 'хочу\\s+(увидеть|видеть)|покажи\\s+(её|ее|его|как)|'
+            + 'сделай\\s+(её|ее|его|мне)|пусть\\s+будет|'
+            + 'давай\\s+(её|ее|его|с\\s|в\\s|на\\s)|'
+            + 'в\\s+(студии|лесу|интерьер|постел)|'
+            + 'другой\\s+(ракурс|свет|наряд|поза)|'
+            + 'поменяй\\s+(позу|свет|одежд|фон)|добавь\\s+(свет|детал)|'
+            + 'ещё\\s+одн|еще\\s+одн|вариант',
+        ).test(t)) {
+            return true;
+        }
+        if (/\b(draw|paint|render|make her|make him|another one|new frame|in the (studio|forest))\b/i.test(t)) {
+            return true;
+        }
+        const isQuestion = /[?？]\s*$/.test(t);
+        if (isQuestion) {
+            return cyrTokenRe('нарису|сгенер|можешь\\s+(сделать|нарисовать)|можно\\s+(картинк|сгенер)').test(t);
+        }
+        return t.length >= 40 && cyrTokenRe(
+            'девушк|женщин|парн|мужчин|стоит|сидит|лежит|обнаж|поза|интерьер|студи',
+        ).test(t);
+    }
+
+    function packBlocksAutoGenerate(pack) {
+        const p = String(pack || '');
+        return p === 'describe_ref' || p === 'catalog_card' || p === 'author_persona';
+    }
+
+    /** If the turn is a frame, put actions:["generate"] on the patch even when the model omitted it. */
+    function ensureGenerateAction(patch, userText) {
+        if (!patch || typeof patch !== 'object') {
+            return patch;
+        }
+        if (userAsksNoGenerate(userText)) {
+            return patch;
+        }
+        if (Array.isArray(patch.actions) && patch.actions.map(String).includes('generate')) {
+            return patch;
+        }
+        const pack = $('sa_pack')?.value || '';
+        if (packBlocksAutoGenerate(pack) && !userAsksGenerate(userText) && !userAsksContinue(userText)) {
+            return patch;
+        }
+        const hasFrame = patch.prompt != null
+            || (Array.isArray(patch.variants) && patch.variants.length > 0);
+        const implied = userImpliesGenerate(userText);
+        if (!hasFrame && !implied) {
+            return patch;
+        }
+        const out = { ...patch };
+        const acts = Array.isArray(patch.actions) ? patch.actions.map(String).filter((a) => a && a !== 'generate') : [];
+        acts.push('generate');
+        out.actions = acts;
+        return out;
+    }
+
+    function packWantsVision(pack) {
+        const p = String(pack || '');
+        return p === 'critique_image' || p === 'describe_ref' || p === 'compose_scene' || p === 'inpaint_edit';
+    }
+
+    /** Honor model look_at only when asked, pack needs it, or look-only (no Generate this turn). */
+    function shouldHonorLookAt(patch, opts = {}) {
+        if (!patch || typeof patch !== 'object') {
+            return false;
+        }
+        if (patch.look_at == null && patch.vision_from == null && patch.vision_slots == null) {
+            return false;
+        }
+        if (opts.fromAutoCritique || opts.fromVisionHop) {
+            return true;
+        }
+        if (userAsksLook(opts.userText || '')) {
+            return true;
+        }
+        if (packWantsVision($('sa_pack')?.value)) {
+            return true;
+        }
+        // look_at + generate on a write turn stares at the old frame and delays the new one.
+        if (patchHasGenTrigger(patch)) {
+            return false;
+        }
+        return true;
+    }
+
    function stripGenerateAction(patch) {
        if (!patch || typeof patch !== 'object') {
            return patch;
@@ -1204,6 +1386,12 @@
                setVal(batchId, String(batch));
            }
        }
+        if (!liveNegativePrompt()) {
+            const neg = exactDefaultNegative();
+            if (neg) {
+                setNegativePrompt(neg);
+            }
+        }
    }

    function shouldSkipSessionRollback(key, patchValue) {
@@ -2274,6 +2462,9 @@
                promptBox.dispatchEvent(new Event('change', { bubbles: true }));
            }
            setVal('input_negativeprompt', params.negative != null ? String(params.negative) : '');
+            if (document.getElementById('alt_negativeprompt_textbox')) {
+                setVal('alt_negativeprompt_textbox', params.negative != null ? String(params.negative) : '');
+            }
            // Force-write numerics when present so prior chat values cannot stick.
            if (params.width != null) {
                setVal('input_width', String(params.width));
@@ -3965,9 +4156,6 @@
                box.dispatchEvent(new Event('input', { bubbles: true }));
                box.dispatchEvent(new Event('change', { bubbles: true }));
            }
-            if (patch.negative != null) {
-                setVal('input_negativeprompt', patch.negative);
-            }
            if (Array.isArray(patch.loras)) {
                for (const l of patch.loras) {
                    const triggers = l.triggers || (l.trigger_phrase ? [l.trigger_phrase] : []);
@@ -3980,6 +4168,13 @@
                }
            }
        }
+        if (doPrompt) {
+            if (patch.negative != null) {
+                setNegativePrompt(patch.negative);
+            } else if (patchHasGenTrigger(patch)) {
+                ensureNegativeForGenerate(patch);
+            }
+        }

        if (doLoras && Array.isArray(patch.loras) && typeof loraHelper !== 'undefined' && loraHelper) {
            try {
@@ -4611,6 +4806,7 @@
                await applyPatch(job.patch, 'all');
                syncLiveParamsBar();
            }
+            ensureNegativeForGenerate(job.patch);

            if (epoch !== state.chatEpoch) {
                break;
@@ -5332,6 +5528,18 @@
            }
            localStorage.setItem(LS_PACK_ORDINARY_MIG, '1');
        }
+        // One-shot: stop staring at every Generate. Vision = button / /look / model look_at.
+        if (!localStorage.getItem(LS_VISION_OPTIN_MIG)) {
+            localStorage.setItem(LS_AUTO_VISION, '0');
+            localStorage.setItem(LS_AUTO_CRITIQUE, '0');
+            localStorage.setItem(LS_VISION_OPTIN_MIG, '1');
+            if ($('sa_auto_vision')) {
+                $('sa_auto_vision').checked = false;
+            }
+            if ($('sa_auto_critique')) {
+                $('sa_auto_critique').checked = false;
+            }
+        }
        const base = localStorage.getItem(LS_BASE);
        const model = localStorage.getItem(LS_MODEL);
        const pack = localStorage.getItem(LS_PACK);
@@ -7865,19 +8073,18 @@
        }
        if (!effective && !fromVisionHop && !fromAutoCritique && !fromDebug
            && !opts.fromEmptyPatchRetry
-            && replyMissingJsonPatch(reply)
-            && (opts.userWantsGenerate || state.pendingSilentGen || userAsksContinue(opts.userText || ''))) {
-            appendSystemNote('Патч пустой — прошу модель дописать JSON.');
+            && userImpliesGenerate(opts.userText || '')) {
+            appendSystemNote('Нужен кадр — прошу JSON с prompt + generate.');
            await sendChat({
                skipSlash: true,
                skipAutoPack: true,
                fromEmptyPatchRetry: true,
                userWantsGenerate: true,
                forcedUserText:
-                    'Ты написал «JSON Patch» без fenced ```json```. '
-                    + 'Сейчас ответь ТОЛЬКО одним fenced JSON объектом: '
-                    + '{"prompt":"<English Krea prompt for Qwen3-VL>","actions":["generate"]}. '
-                    + 'prompt — только английский (skill prompting). Без прозы, без ### заголовков.',
+                    'Пользователь уже просит кадр (это следует из сообщения, даже без слова «генерируй»). '
+                    + 'Ответь ТОЛЬКО одним fenced JSON: '
+                    + '{"prompt":"<English Krea prompt>","negative":"<echo or short quality neg>","actions":["generate"]}. '
+                    + 'prompt — английский. Без прозы, без «скажи сгенерируй».',
            });
            return;
        }
@@ -7905,7 +8112,17 @@
                rememberLastPatch(effective);
            }
        }
-        if (effective && !fromVisionHop && !fromAutoCritique && !suppressGen) {
+        if (effective && !fromDebug && !suppressGen) {
+            effective = ensureGenerateAction(effective, opts.userText || '');
+            rememberLastPatch(effective);
+        }
+        if (effective && !fromVisionHop && !fromAutoCritique && !fromDebug && !shouldHonorLookAt(effective, opts)) {
+            effective = stripLookAt(effective);
+            if (effective) {
+                rememberLastPatch(effective);
+            }
+        }
+        if (effective && !fromVisionHop && !fromAutoCritique && !suppressGen && shouldHonorLookAt(effective, opts)) {
            const hopped = await maybeVisionHop(effective, opts.attachedSlotIds || []);
            if (hopped) {
                return;
@@ -7917,6 +8134,7 @@
            return;
        }
        const wantsGen = !suppressGen && !!(opts.userWantsGenerate || state.pendingSilentGen
+            || userImpliesGenerate(opts.userText || '')
            || (Array.isArray(effective?.actions) && effective.actions.map(String).includes('generate')));
        const willGen = !!(effective && !fromAutoCritique && !suppressGen
            && (wantsGen || $('sa_auto_generate')?.checked));
@@ -8385,7 +8603,7 @@
        if (!opts.fromAutoCritique && !opts.fromDownload && !opts.fromVisionHop && !opts.fromDebug) {
            state.lastUserParamIntent = userTextMentionsParams(text);
            state.lastUserControlIntent = userTextMentionsControls(text);
-            state.pendingSilentGen = userAsksGenerate(text) || userAsksContinue(text) || isSameButAspectRequest(text);
+            state.pendingSilentGen = userImpliesGenerate(text);
            if (userAsksNoGenerate(text)) {
                state.pendingSilentGen = false;
            }
@@ -8602,12 +8820,12 @@
            const prose = extractPatch(reply).prose || reply;
            state.history.push({ role: 'assistant', content: prose, persona, pack });
            persistHistory();
-            setBusyPhase(state.pendingSilentGen || userAsksGenerate(text) || userAsksContinue(text) ? 'silent_gen' : 'thinking');
+            setBusyPhase(state.pendingSilentGen || userImpliesGenerate(text) ? 'silent_gen' : 'thinking');
            try {
                await handleReplySideEffects(reply, civitaiResults, {
                    ...opts,
                    userText: text,
-                    userWantsGenerate: !!state.pendingSilentGen || userAsksGenerate(text) || userAsksContinue(text),
+                    userWantsGenerate: !!state.pendingSilentGen || userImpliesGenerate(text),
                    attachedSlotIds: visionSlots.map((s) => s.id),
                });
            } finally {
@@ -6,7 +6,7 @@ window.SA = window.SA || {};

 (function () {
    const PATCH_KEYS = [
-        'prompt', 'loras', 'width', 'height', 'steps', 'cfg', 'seed', 'sigma_shift', 'sampler',
+        'prompt', 'negative', 'loras', 'width', 'height', 'steps', 'cfg', 'seed', 'sigma_shift', 'sampler',
        'actions', 'search_query', 'civitai_query',
        'use_init_image', 'clear_init_image', 'init_creativity', 'denoise',
        'use_mask_image', 'clear_mask_image', 'mask_blur', 'mask_grow',
@@ -31,8 +31,8 @@ window.SA = window.SA || {};
            return false;
        }
        const cardish = !!(obj.kind || obj.triggers || obj.when || obj.prompt_hint);
-        const genish = !!(obj.prompt != null || obj.loras || obj.actions || obj.width || obj.height
-            || obj.steps || obj.cfg || obj.aspect || obj.seed != null
+        const genish = !!(obj.prompt != null || obj.negative != null || obj.loras || obj.actions
+            || obj.width || obj.height || obj.steps || obj.cfg || obj.aspect || obj.seed != null
            || obj.search_query || obj.civitai_query || obj.look_at || obj.controls);
        if (cardish && !genish && (obj.name || obj.triggers || obj.when)) {
            return true;