diff --git a/Assets/assistent.js b/Assets/assistent.js
index 98dfe8f..37d6323 100644
--- a/Assets/assistent.js
+++ b/Assets/assistent.js
@@ -1126,6 +1126,7 @@
             auto_generate: !!$('sa_auto_generate')?.checked,
             persona: $('sa_persona')?.value || localStorage.getItem(LS_PERSONA) || 'neutral',
             model_cards: [],
+            taste_profile: summarizeTaste(),
             ...initCtx,
         };
 
@@ -1149,20 +1150,48 @@
             }
         } catch (e) { /* ignore */ }
 
-        // Recommendation cards for current checkpoint + selected LoRAs only.
+        // Recommendation cards: checkpoint + selected LoRAs + other has_card entries (capped).
         const cardKeys = [];
+        const seenCard = new Set();
+        const addKey = (kind, name) => {
+            if (!kind || !name) {
+                return;
+            }
+            const key = `${kind}:${name}`;
+            if (seenCard.has(key)) {
+                return;
+            }
+            seenCard.add(key);
+            cardKeys.push({ kind, name });
+        };
         if (ctx.checkpoint?.name) {
-            cardKeys.push({ kind: 'checkpoint', name: ctx.checkpoint.name });
+            addKey('checkpoint', ctx.checkpoint.name);
         }
         for (const l of ctx.selected_loras || []) {
             if (l?.name) {
-                cardKeys.push({ kind: 'lora', name: l.name });
+                addKey('lora', l.name);
+            }
+        }
+        for (const l of inv.loras || []) {
+            if (l?.has_card && l?.name) {
+                addKey('lora', l.name);
+            }
+            if (cardKeys.length >= 14) {
+                break;
+            }
+        }
+        for (const c of inv.checkpoints || []) {
+            if (c?.has_card && c?.name) {
+                addKey('checkpoint', c.name);
+            }
+            if (cardKeys.length >= 16) {
+                break;
             }
         }
         for (const k of cardKeys) {
             const cached = state.modelCards[`${k.kind}:${k.name}`];
             if (cached) {
-                ctx.model_cards.push(cached);
+                ctx.model_cards.push(slimCardForContext(cached));
             }
         }
 
@@ -1197,6 +1226,42 @@
         return ctx;
     }
 
+    function slimCardForContext(card) {
+        if (!card || typeof card !== 'object') {
+            return null;
+        }
+        const out = {
+            kind: card.kind || null,
+            name: card.name || null,
+            triggers: Array.isArray(card.triggers) ? card.triggers.slice(0, 8) : undefined,
+            weight: card.weight != null ? card.weight : undefined,
+            when: card.when ? String(card.when).slice(0, 160) : undefined,
+            avoid: card.avoid ? String(card.avoid).slice(0, 120) : undefined,
+            prompt_hint: card.prompt_hint ? String(card.prompt_hint).slice(0, 160) : undefined,
+            notes: card.notes ? String(card.notes).slice(0, 200) : undefined,
+        };
+        const clean = {};
+        for (const [k, v] of Object.entries(out)) {
+            if (v != null && v !== '') {
+                clean[k] = v;
+            }
+        }
+        return clean;
+    }
+
+    function summarizeTaste() {
+        const t = state.taste || {};
+        if (!(t.styles?.length || t.likes?.length || t.avoid?.length || t.notes)) {
+            return null;
+        }
+        return {
+            styles: (t.styles || []).slice(0, 8),
+            likes: (t.likes || []).slice(0, 10),
+            avoid: (t.avoid || []).slice(0, 8),
+            notes: t.notes ? String(t.notes).slice(0, 240) : undefined,
+        };
+    }
+
     function slimInventoryLoras(list, limit) {
         const selected = new Set();
         try {
@@ -1206,26 +1271,60 @@
                 }
             }
         } catch (e) { /* ignore */ }
-        const rows = (list || []).map((l) => ({
-            name: l.name,
-            title: l.title || l.name,
-            trigger_phrase: l.trigger_phrase || null,
-            triggers: Array.isArray(l.triggers) ? l.triggers.slice(0, 8) : undefined,
-            architecture: l.architecture || null,
-            compat_class: l.compat_class || null,
-            has_card: !!l.has_card,
-            krea_likely: !!l.krea_likely,
-            blurb: l.blurb || l.usage_hint || null,
-            default_weight: l.default_weight || undefined,
-            tags: Array.isArray(l.tags) ? l.tags.slice(0, 6) : undefined,
-            _sel: selected.has(String(l.name || '').toLowerCase()),
-        }));
-        rows.sort((a, b) => (b._sel - a._sel) || (b.krea_likely - a.krea_likely) || String(a.name).localeCompare(String(b.name)));
-        return rows.slice(0, limit).map(({ _sel, ...rest }) => {
-            const out = {};
-            for (const [k, v] of Object.entries(rest)) {
-                if (v != null && v !== '' && !(Array.isArray(v) && !v.length)) {
-                    out[k] = v;
+        const softCap = Math.min(limit || 100, 80);
+        const rows = (list || []).map((l) => {
+            const sel = selected.has(String(l.name || '').toLowerCase());
+            const hasCard = !!l.has_card;
+            const krea = !!l.krea_likely;
+            const blurb = l.blurb || l.usage_hint || null;
+            return {
+                name: l.name,
+                title: l.title || l.name,
+                trigger_phrase: l.trigger_phrase || null,
+                triggers: Array.isArray(l.triggers) ? l.triggers.slice(0, 8) : undefined,
+                architecture: l.architecture || null,
+                compat_class: l.compat_class || null,
+                has_card: hasCard,
+                krea_likely: krea,
+                blurb,
+                default_weight: l.default_weight || undefined,
+                tags: Array.isArray(l.tags) ? l.tags.slice(0, 6) : undefined,
+                _score: (sel ? 1000 : 0) + (hasCard ? 200 : 0) + (krea ? 50 : 0) + (blurb ? 10 : 0),
+            };
+        });
+        rows.sort((a, b) => b._score - a._score || String(a.name).localeCompare(String(b.name)));
+        // Full detail for top tier; name+trigger only for the rest within softCap.
+        const fullDetail = 36;
+        return rows.slice(0, softCap).map((row, idx) => {
+            const out = { name: row.name, title: row.title };
+            if (row.trigger_phrase) {
+                out.trigger_phrase = row.trigger_phrase;
+            }
+            if (row.triggers) {
+                out.triggers = row.triggers;
+            }
+            if (row.krea_likely) {
+                out.krea_likely = true;
+            }
+            if (row.has_card) {
+                out.has_card = true;
+            }
+            const rich = idx < fullDetail || row._score >= 200;
+            if (rich) {
+                if (row.architecture) {
+                    out.architecture = row.architecture;
+                }
+                if (row.compat_class) {
+                    out.compat_class = row.compat_class;
+                }
+                if (row.blurb) {
+                    out.blurb = row.blurb;
+                }
+                if (row.default_weight) {
+                    out.default_weight = row.default_weight;
+                }
+                if (row.tags) {
+                    out.tags = row.tags;
                 }
             }
             return out;
@@ -1233,7 +1332,9 @@
     }
 
     function slimInventoryCheckpoints(list, limit) {
-        return (list || []).slice(0, limit).map((c) => {
+        const rows = (list || []).slice();
+        rows.sort((a, b) => ((b.krea_likely ? 1 : 0) - (a.krea_likely ? 1 : 0)) || ((b.has_card ? 1 : 0) - (a.has_card ? 1 : 0)) || String(a.name).localeCompare(String(b.name)));
+        return rows.slice(0, limit || 40).map((c) => {
             const out = {
                 name: c.name,
                 title: c.title || c.name,
@@ -1249,6 +1350,77 @@
         });
     }
 
+    function loadTaste() {
+        try {
+            const raw = localStorage.getItem(LS_TASTE);
+            if (!raw) {
+                return;
+            }
+            const parsed = JSON.parse(raw);
+            if (parsed && typeof parsed === 'object') {
+                state.taste = {
+                    styles: Array.isArray(parsed.styles) ? parsed.styles.slice(0, 12) : [],
+                    likes: Array.isArray(parsed.likes) ? parsed.likes.slice(0, 16) : [],
+                    avoid: Array.isArray(parsed.avoid) ? parsed.avoid.slice(0, 12) : [],
+                    notes: String(parsed.notes || '').slice(0, 400),
+                    updated: parsed.updated || 0,
+                };
+            }
+        } catch (e) { /* ignore */ }
+    }
+
+    function saveTaste() {
+        try {
+            localStorage.setItem(LS_TASTE, JSON.stringify(state.taste || {}));
+        } catch (e) { /* ignore */ }
+    }
+
+    function pushUnique(arr, value, max) {
+        const v = String(value || '').trim();
+        if (!v || v.length < 2) {
+            return;
+        }
+        const lower = v.toLowerCase();
+        const next = (arr || []).filter((x) => String(x).toLowerCase() !== lower);
+        next.unshift(v.slice(0, 80));
+        return next.slice(0, max);
+    }
+
+    function updateTasteFromPatch(patch, userText) {
+        if (!patch) {
+            return;
+        }
+        const taste = state.taste || { styles: [], likes: [], avoid: [], notes: '' };
+        if (Array.isArray(patch.loras)) {
+            for (const l of patch.loras) {
+                const name = l?.name || l;
+                if (name) {
+                    taste.likes = pushUnique(taste.likes, name, 16);
+                }
+            }
+        }
+        const aspect = patch.aspect || null;
+        if (aspect) {
+            taste.styles = pushUnique(taste.styles, `aspect ${aspect}`, 12);
+        }
+        if (patch.creativity) {
+            taste.styles = pushUnique(taste.styles, `creativity:${patch.creativity}`, 12);
+        }
+        const ut = String(userText || '').toLowerCase();
+        if (/фото|photo|photoreal|реализм|film grain/.test(ut)) {
+            taste.styles = pushUnique(taste.styles, 'photoreal / film', 12);
+        }
+        if (/аниме|anime|illustration|иллюстр/.test(ut)) {
+            taste.styles = pushUnique(taste.styles, 'illustration / anime', 12);
+        }
+        if (/без\s+3d|не\s+3d|no\s+3d|не\s+render/.test(ut)) {
+            taste.avoid = pushUnique(taste.avoid, '3D render look', 12);
+        }
+        taste.updated = Date.now();
+        state.taste = taste;
+        saveTaste();
+    }
+
     function isPatchObject(obj) {
         if (!obj || typeof obj !== 'object') {
             return false;
@@ -1967,11 +2139,12 @@
                 if (btn) {
                     btn.textContent = 'Downloaded';
                 }
-                refreshInventory(() => {
-                    if ($('sa_input')) {
-                        $('sa_input').value = `LoRA "${payload.name}" is now installed. Enable it with its triggers and improve the prompt.`;
-                    }
-                    sendChat({ fromDownload: true });
+                refreshInventory(async () => {
+                    await maybeWriteCardAfterDownload({
+                        kind: 'lora',
+                        name: payload.name,
+                        civitai: card,
+                    });
                 }, { rescan: true });
             } else {
                 setStatus(msg || 'Download failed');
@@ -2008,6 +2181,30 @@
         }
     }
 
+    async function maybeWriteCardAfterDownload({ kind, name, civitai }) {
+        const display = name || civitai?.file_name || civitai?.name || 'model';
+        appendSystemNote(`Downloaded ${display}. Writing a recommendation card…`);
+        setPackValue('catalog_card', { flash: true });
+        const meta = {
+            triggers: civitai?.triggers || [],
+            base_model: civitai?.base_model,
+            civitai_url: civitai?.url || civitai?.civitai_url,
+            version_id: civitai?.version_id || civitai?.modelVersionId,
+            name: display,
+        };
+        if ($('sa_input')) {
+            $('sa_input').value = '';
+        }
+        await sendChat({
+            forcedUserText: `LoRA "${display}" is now installed. Write a recommendation card (JSON) using its triggers/metadata. Then briefly suggest how to enable it in the next generate.`,
+            skipSlash: true,
+            skipAutoPack: true,
+            fromDownload: true,
+            fromCards: true,
+            cardTarget: { kind: kind || 'lora', name: display, meta },
+        });
+    }
+
     function wantsAutoVision() {
         return !!$('sa_auto_vision')?.checked;
     }
@@ -2429,22 +2626,47 @@
 
     async function prefetchActiveModelCards() {
         const keys = [];
+        const seen = new Set();
+        const add = (kind, name) => {
+            if (!kind || !name) {
+                return;
+            }
+            const key = `${kind}:${name}`;
+            if (seen.has(key)) {
+                return;
+            }
+            seen.add(key);
+            keys.push({ kind, name });
+        };
         try {
             const ck = resolveCurrentCheckpoint();
             if (ck?.name) {
-                keys.push({ kind: 'checkpoint', name: ck.name });
+                add('checkpoint', ck.name);
             }
         } catch (e) { /* ignore */ }
         try {
             if (typeof loraHelper !== 'undefined' && Array.isArray(loraHelper?.selected)) {
                 for (const l of loraHelper.selected) {
-                    const name = l?.name || l;
-                    if (name) {
-                        keys.push({ kind: 'lora', name });
-                    }
+                    add('lora', l?.name || l);
                 }
             }
         } catch (e) { /* ignore */ }
+        for (const l of state.inventory?.loras || []) {
+            if (l?.has_card) {
+                add('lora', l.name);
+            }
+            if (keys.length >= 14) {
+                break;
+            }
+        }
+        for (const c of state.inventory?.checkpoints || []) {
+            if (c?.has_card) {
+                add('checkpoint', c.name);
+            }
+            if (keys.length >= 16) {
+                break;
+            }
+        }
         await Promise.all(keys.map((k) => prefetchCard(k.kind, k.name)));
     }
 
@@ -2689,10 +2911,32 @@
         const { fromAutoCritique, fromVisionHop, fromCards } = opts;
         if (fromCards) {
             const card = extractCardJson(reply);
-            if (card && $('sa_card_json')) {
-                $('sa_card_json').value = JSON.stringify(card, null, 2);
-                setView('cards');
-                setCardStatus('Draft from Assistent — review & Save');
+            if (card) {
+                if ($('sa_card_json')) {
+                    $('sa_card_json').value = JSON.stringify(card, null, 2);
+                }
+                if (opts.fromDownload || opts.cardTarget) {
+                    const kind = card.kind || opts.cardTarget?.kind || 'lora';
+                    const name = card.name || opts.cardTarget?.name;
+                    if (name && typeof genericRequest === 'function') {
+                        genericRequest(
+                            'AssistentSaveCard',
+                            { kind, name, card, enqueue_wanted: false },
+                            (data) => {
+                                if (data?.path) {
+                                    state.modelCards[`${kind}:${name}`] = card;
+                                    setCardStatus(data.installed ? `Card saved → ${data.path}` : `Card draft → ${data.path}`);
+                                    setStatus(`Card saved for ${name}`);
+                                }
+                            },
+                            0,
+                            () => setCardStatus('Card draft ready — Save manually'),
+                        );
+                    }
+                } else {
+                    setView('cards');
+                    setCardStatus('Draft from Assistent — review & Save');
+                }
             }
             return;
         }
@@ -2715,6 +2959,7 @@
         }
         if (patch && $('sa_auto_apply')?.checked) {
             await applyPatch(patch, 'all');
+            updateTasteFromPatch(patch, opts.userText || '');
             if (!fromAutoCritique) {
                 const src = await runGenerateFromPatch(patch);
                 if (src) {
@@ -3018,6 +3263,7 @@
         // (rescans disk when inventory is older than ~20s or after downloads).
         setStatus('Refreshing inventory…');
         await ensureFreshInventory({ forceRescan: !!opts.fromDownload });
+        await prefetchActiveModelCards();
 
         if (!opts.fromAutoCritique && !opts.fromDownload && !opts.fromVisionHop) {
             state.critiqueHopUsed = false;
@@ -3104,6 +3350,7 @@
             stopBusyUi('Done');
             await handleReplySideEffects(reply, civitaiResults, {
                 ...opts,
+                userText: text,
                 attachedSlotIds: visionSlots.map((s) => s.id),
             });
         };
@@ -3350,6 +3597,7 @@
         }
         window.__swarmAssistentWired = true;
         loadSettings();
+        loadTaste();
         setView(state.view || 'chat');
         updateGate();
         ensureBoard();
diff --git a/Prompts/base_krea2.md b/Prompts/base_krea2.md
index c174d1f..8a76eb4 100644
--- a/Prompts/base_krea2.md
+++ b/Prompts/base_krea2.md
@@ -7,8 +7,10 @@ You are **Swarm Assistent**, a collaborative art director for **Krea 2** image g
 - Architecture: Krea 2 (12B DiT). Not FLUX, not SDXL, not FLUX.1-Krea.
 - Text encoder: Qwen3-VL 4B. VAE: Qwen Image VAE.
 - **Turbo** defaults: steps **8** (min 4), CFG **1** (never CFG 0 — broken output), sigma shift **1.15**, side ~**1024** (128–4096 OK).
-- **RAW / Base:** steps ~20–52, CFG ~4–4.5. Community tip: RAW + turbo LoRA ~0.6 often beats pure Turbo for photoreal — mention only if checkpoint looks RAW; do not invent workflows Swarm cannot run.
+- **RAW / Base:** steps ~20–52, CFG ~4–4.5. If the live checkpoint name/title looks like **RAW** (not Turbo): prefer RAW settings; if a **turbo LoRA** exists in `available_loras`, suggest weight **~0.6** for photoreal (1.0 ≈ full turbo). Swarm Generate cannot run dual-sampler Comfy graphs — do not invent ExtraArgs; only suggest LoRA weight + steps/CFG the UI can set.
 - LoRAs: **only Krea2-trained**. Never suggest FLUX/SDXL LoRAs.
+- `model_cards` in live context (when present) beat generic blurbs — follow `when` / `avoid` / `prompt_hint` / `triggers`.
+- `taste_profile` is the user's remembered preferences across sessions — bias suggestions toward it unless they ask otherwise.
 
 ## How to prompt (local Swarm, not krea.ai cloud)
 
@@ -52,7 +54,9 @@ A JSON block named "Live SwarmUI context" is attached. Treat it as ground truth
 
 - Use only LoRAs listed in `available_loras` (by exact `name`), or candidates from a Civitai search round.
 - Prefer listed `trigger_phrase` / `triggers` — **never invent** trigger words.
+- When present, use `blurb` / `usage_hint` / `tags` / `has_card` to pick the right LoRA.
 - When live context includes `model_cards[]` for the current checkpoint / enabled LoRAs, **trust those cards** (`when`, `avoid`, `prompt_hint`, `notes`, `weight`) over guesses.
+- `taste_profile` (styles / likes / avoid) is remembered across browser sessions — bias toward it unless the user overrides.
 - Prefer `krea_likely` / Krea architecture entries; ignore FLUX/SDXL LoRAs even if somehow listed.
 - `default_weight` is a starting LoRA weight when set.
 - `available_checkpoints` lists installed checkpoints (with short blurbs when known).
diff --git a/Prompts/fix_params.md b/Prompts/fix_params.md
index 2fc83f7..af5e375 100644
--- a/Prompts/fix_params.md
+++ b/Prompts/fix_params.md
@@ -5,7 +5,7 @@ Goal: adjust **generation parameters** for Krea 2 Turbo (or RAW if context says
 ## Guidelines
 
 - **Turbo:** steps 4–12 (default **8**), CFG **1** (never 0), sigma shift ~**1.15**.
-- **RAW/base:** steps 20+, CFG ~4–4.5 — only if checkpoint/context indicates Raw.
+- **RAW/base:** steps 20–52, CFG ~4–4.5 — only if checkpoint/context indicates Raw. If a turbo-distill LoRA is available, weight **0.6** is the usual photoreal compromise (UI LoRA only — no dual-sampler).
 - **Aspect:** prefer patch field `aspect` (`1:1`, `4:5`, `2:3`, `16:9`, `9:16`, `4:3`, `3:2`, `2.35:1`) — UI maps to official 1K sizes. Else set width/height near 1024.
 - **Batch:** `images` or `batch` (1–4 typical).
 - **Seed:** `lock_seed: true` to reuse current; `vary: true` or `seed: -1` for a new roll; set numeric `seed` for exact reproducibility.
diff --git a/README.md b/README.md
index 19b9746..b46246a 100644
--- a/README.md
+++ b/README.md
@@ -78,7 +78,9 @@ Restart / rebuild SwarmUI after clone.
 
 **Cards** subtab: edit/save `{stem}.assistent.json` next to weights; live chat gets cards for the current checkpoint + enabled LoRAs only. Uninstalled models can enqueue `.gpu-rent-wanted-models.yaml` for the next `up`.
 
-Live context (checkpoint, server inventory LoRAs + triggers, `model_cards`, wildcards, current params, persona) is injected every request.
+Live context (checkpoint, server inventory LoRAs + blurbs/triggers, `model_cards`, `taste_profile`, wildcards, current params, persona) is injected every request. Inventory refreshes each chat turn (disk rescan when stale / after download / `/inventory`). After Confirm download Assistent auto-writes a `.assistent.json` card.
+
+**Wanted queue:** Cards → Enqueue wanted → `/mnt/swarm_data/.gpu-rent-wanted-models.yaml`. Laptop: `gpu-rent capture wanted` (also on `capture models` / seed `up`).
 
 **Cloud-only Krea.ai features** (moodboards UI, Generative Sliders, Creativity Raw/Low/Medium/High) are **not** in Swarm. The assistant emulates them with prompt language + board refs. Optional patch fields `creativity` / `intensity` / `complexity` / `movement` guide the LLM only.