restructure backend

.gitignore

@@ -33,4 +33,5 @@ report.[0-9]_.[0-9]_.[0-9]_.[0-9]_.json
 # Finder (MacOS) folder config
 .DS_Store
 
-**/venv
+**/venv
+images

README.md

@@ -1,15 +1,43 @@
-# desktop
+# Bun Server
 
-To install dependencies:
+Install dependencies:
 
 ```bash
 bun install
 ```
 
-To run:
+Run:
 
 ```bash
-bun run index.ts
+bun run src/index.ts
 ```
 
-This project was created using `bun init` in bun v1.3.5. [Bun](https://bun.com) is a fast all-in-one JavaScript runtime.
+# Python Service
+
+Run:
+
+```bash
+venv/Scripts/Activate.ps1
+```
+```bash
+python python_service/app.py
+```
+
+# BACKEND DATA GENERATION
+
+Update index / embeddings / ids for (new) cards:
+
+1. update cards in `\images\cards`
+2. activate venv:
+
+```bash
+venv/Scripts/Activate.ps1
+```
+
+3. run python script:
+
+```bash
+ python encodeImages.py
+```
+
+4. great success

bun.lock

@@ -5,7 +5,7 @@
     "": {
       "name": "desktop",
       "devDependencies": {
-        "@types/bun": "latest",
+        "@types/bun": "^1.3.6",
       },
       "peerDependencies": {
         "typescript": "^5",
@@ -13,11 +13,11 @@
     },
   },
   "packages": {
-    "@types/bun": ["@types/bun@1.3.5", "", { "dependencies": { "bun-types": "1.3.5" } }, "sha512-RnygCqNrd3srIPEWBd5LFeUYG7plCoH2Yw9WaZGyNmdTEei+gWaHqydbaIRkIkcbXwhBT94q78QljxN0Sk838w=="],
+    "@types/bun": ["@types/bun@1.3.6", "", { "dependencies": { "bun-types": "1.3.6" } }, "sha512-uWCv6FO/8LcpREhenN1d1b6fcspAB+cefwD7uti8C8VffIv0Um08TKMn98FynpTiU38+y2dUO55T11NgDt8VAA=="],
 
     "@types/node": ["@types/node@25.0.3", "", { "dependencies": { "undici-types": "~7.16.0" } }, "sha512-W609buLVRVmeW693xKfzHeIV6nJGGz98uCPfeXI1ELMLXVeKYZ9m15fAMSaUPBHYLGFsVRcMmSCksQOrZV9BYA=="],
 
-    "bun-types": ["bun-types@1.3.5", "", { "dependencies": { "@types/node": "*" } }, "sha512-inmAYe2PFLs0SUbFOWSVD24sg1jFlMPxOjOSSCYqUgn4Hsc3rDc7dFvfVYjFPNHtov6kgUeulV4SxbuIV/stPw=="],
+    "bun-types": ["bun-types@1.3.6", "", { "dependencies": { "@types/node": "*" } }, "sha512-OlFwHcnNV99r//9v5IIOgQ9Uk37gZqrNMCcqEaExdkVq3Avwqok1bJFmvGMCkCE0FqzdY8VMOZpfpR3lwI+CsQ=="],
 
     "typescript": ["typescript@5.9.3", "", { "bin": { "tsc": "bin/tsc", "tsserver": "bin/tsserver" } }, "sha512-jl1vZzPDinLr9eUt3J/t7V6FgNEw9QjvBPdysz9KfQDD41fQrC2Y4vKQdiaUpFT4bXlb1RHhLpp8wtm6M5TgSw=="],
 

docker-compose.yml

@@ -0,0 +1,15 @@
+version: "3.9"
+
+services:
+  pkmtcg-backend:
+    build: .
+    container_name: pkmtcg-backend
+    ports:
+      - "3333:3000"
+    environment:
+      IMAGES_DIR: /images/cards
+      FAISS_DIR: /faiss
+    volumes:
+      - ./data/images:/images/cards   # host ./data/images => container /images/cards
+      - ./data/faiss:/faiss           # host ./data/faiss => container /faiss
+    restart: unless-stopped

dockerfile

@@ -11,14 +11,22 @@ RUN apt-get update && apt-get install -y \
 RUN curl -fsSL https://bun.sh/install | bash
 ENV PATH="/root/.bun/bin:${PATH}"
 
+# --- Upgrade pip ---
+RUN pip3 install --upgrade pip
+
 # --- Install Python dependencies ---
 RUN pip3 install --upgrade pip
-RUN pip3 install torch --index-url https://download.pytorch.org/whl/cpu
-RUN pip3 install --no-cache-dir fastapi uvicorn pillow open-clip-torch numpy faiss-cpu python-multipart
+RUN pip3 install fastapi uvicorn pillow torch open-clip-torch numpy faiss-cpu python-multipart
+
+# --- Set working directory ---
+WORKDIR /app
+
+# --- Create directories for external data ---
+RUN mkdir -p /images/cards /faiss
 
 # --- Copy project files ---
 WORKDIR /app
-COPY python_service ./python_service
+COPY python_service/app.py ./python_service/app.py
 COPY src ./src
 COPY package.json .
 COPY bun.lock .
@@ -26,6 +34,10 @@ COPY bun.lock .
 # --- Expose ports ---
 EXPOSE 3000
 
+# --- Environment variables for external paths ---
+ENV IMAGES_DIR=/images/cards
+ENV FAISS_DIR=/faiss
+
 # --- Start services ---
-# Use & to run Python worker in background, Bun frontend as main process
+# Python FAISS worker runs in background, Bun frontend as main process
 CMD python3 python_service/app.py & bun run src/index.ts

encodeImages.py

@@ -0,0 +1,71 @@
+import os
+import numpy as np
+import torch
+import open_clip
+import faiss
+from PIL import Image
+
+CARDS_FOLDER = "/images/cards"
+EMBEDDINGS_FILE = "/pythonService/embeddings.npy"
+IDS_FILE = "/pythonService/ids.npy"
+FAISS_FILE = "/pythonService/card_index.faiss"
+
+device = "cuda" if torch.cuda.is_available() else "cpu"
+model, _, preprocess = open_clip.create_model_and_transforms(
+    'ViT-L-14', pretrained='laion2b_s32b_b82k'
+)
+model = model.to(device).eval()
+
+# ---- load existing or initialize ----
+if os.path.exists(FAISS_FILE):
+    print("Loading existing FAISS index...")
+    index = faiss.read_index(FAISS_FILE)
+    embeddings = np.load(EMBEDDINGS_FILE)
+    ids = np.load(IDS_FILE)
+else:
+    print("Creating new FAISS index...")
+    embeddings = np.zeros((0, 1024), dtype='float32')  # 1024 for ViT-L-14
+    ids = np.array([], dtype='<U100')
+    index = faiss.IndexFlatIP(1024)
+
+# ---- find images not yet processed ----
+existing_ids = set(ids.tolist())
+new_files = [
+    f for f in os.listdir(CARDS_FOLDER)
+    if f.lower().endswith((".png", ".jpg")) and f.rsplit(".", 1)[0] not in existing_ids
+]
+
+print(f"Found {len(new_files)} new cards to add")
+
+new_embeddings = []
+new_ids = []
+
+for fname in new_files:
+    path = os.path.join(CARDS_FOLDER, fname)
+    img = Image.open(path).convert("RGB")
+    with torch.no_grad():
+        emb = model.encode_image(preprocess(img).unsqueeze(0).to(device))
+
+    new_embeddings.append(emb.cpu().numpy())
+    new_ids.append(fname.rsplit(".", 1)[0])
+    print("Encoded:", fname)
+
+if len(new_embeddings) > 0:
+    new_embeddings = np.vstack(new_embeddings).astype('float32')
+    faiss.normalize_L2(new_embeddings)
+
+    # add to FAISS
+    index.add(new_embeddings)
+
+    # append to numpy arrays
+    embeddings = np.vstack([embeddings, new_embeddings])
+    ids = np.concatenate([ids, np.array(new_ids)])
+
+    # save everything
+    np.save(EMBEDDINGS_FILE, embeddings)
+    np.save(IDS_FILE, ids)
+    faiss.write_index(index, FAISS_FILE)
+
+    print(f"Added {len(new_files)} cards. Total now:", index.ntotal)
+else:
+    print("No new cards found — nothing to update.")

package.json

@@ -7,7 +7,7 @@
     "dev": "bun run src/index.ts"
   },
   "devDependencies": {
-    "@types/bun": "latest"
+    "@types/bun": "^1.3.6"
   },
   "peerDependencies": {
     "typescript": "^5"

pythonScripts/build_index.py

@@ -0,0 +1,61 @@
+import os
+import numpy as np
+from PIL import Image
+import torch
+import open_clip
+import faiss
+
+# --- Configuration ---
+CARDS_FOLDER = "cards_old"
+EMBEDDINGS_FILE = "embeddings.npy"
+IDS_FILE = "ids.npy"
+FAISS_INDEX_FILE = "card_index.faiss"
+
+# --- Device ---
+device = "cuda" if torch.cuda.is_available() else "cpu"
+print("Using device:", device)
+
+# --- Load CLIP model ---
+model, _, preprocess = open_clip.create_model_and_transforms(
+    'ViT-L-14', pretrained='laion2b_s32b_b82k'
+)
+model = model.to(device).eval()
+
+# --- Helper: encode image ---
+def encode_image(path):
+    img = Image.open(path).convert("RGB")
+    with torch.no_grad():
+        emb = model.encode_image(preprocess(img).unsqueeze(0).to(device))
+    return emb.cpu().numpy()
+
+# --- Build embeddings ---
+embeddings = []
+ids = []
+
+for fname in os.listdir(CARDS_FOLDER):
+    if fname.lower().endswith((".jpg", ".png")):
+        path = os.path.join(CARDS_FOLDER, fname)
+        emb = encode_image(path)
+        embeddings.append(emb)
+        ids.append(fname)
+        print("Encoded:", fname)
+
+embeddings = np.vstack(embeddings)
+
+# --- Save embeddings & IDs ---
+np.save(EMBEDDINGS_FILE, embeddings)
+np.save(IDS_FILE, np.array(ids))
+print("Saved embeddings and IDs.")
+
+# --- Normalize embeddings ---
+faiss.normalize_L2(embeddings)
+
+# --- Build FAISS index ---
+d = embeddings.shape[1]  # embedding dimension
+index = faiss.IndexFlatIP(d)  # inner product = cosine similarity
+index.add(embeddings)
+print("FAISS index built with", index.ntotal, "cards.")
+
+# --- Save FAISS index ---
+faiss.write_index(index, FAISS_INDEX_FILE)
+print("FAISS index saved:", FAISS_INDEX_FILE)

pythonScripts/download_baseset.py

@@ -0,0 +1,65 @@
+import os
+import re
+import requests
+from bs4 import BeautifulSoup
+
+HTML_FILE = "baseset.html"
+DOWNLOAD_FOLDER = "baseset"
+
+SET_MAP = {
+    "Basis-Set": 1,
+}
+
+os.makedirs(DOWNLOAD_FOLDER, exist_ok=True)
+
+with open(HTML_FILE, "r", encoding="utf-8") as f:
+    soup = BeautifulSoup(f, "html.parser")
+
+entries = []
+
+for a in soup.find_all("a"):
+    url = a.get("href", "")
+    title = a.get("data-elementor-lightbox-title")
+
+    if not title or not url.lower().endswith(".jpg"):
+        continue
+
+    # We only want the original (no -427x600 etc)
+    if re.search(r"-\d+x\d+\.jpg$", url.lower()):
+        continue
+
+    # Parse title: "Abra 43/102 - Basis-Set"
+    m = re.match(r"(.+)\s+(\d+)\/(\d+)\s*-\s*(.+)", title.strip())
+    if not m:
+        print(f"Skipping unmatched title format: {title}")
+        continue
+
+    name, card, total, set_name = m.groups()
+    card = int(card)
+
+    if set_name not in SET_MAP:
+        print(f"Unknown set: {set_name}, please map it.")
+        continue
+
+    set_num = SET_MAP[set_name]
+    new_filename = f"base{set_num}-{card}.jpg"
+
+    entries.append((url, new_filename))
+
+print(f"Found {len(entries)} images to download.")
+
+for url, filename in entries:
+    filepath = os.path.join(DOWNLOAD_FOLDER, filename)
+    print(f"Downloading {filename} from {url}")
+
+    try:
+        r = requests.get(url, timeout=10)
+        r.raise_for_status()
+    except Exception as e:
+        print(f"  Failed: {e}")
+        continue
+
+    with open(filepath, "wb") as f:
+        f.write(r.content)
+
+print("Done!")

pythonScripts/download_cards.py

@@ -0,0 +1,56 @@
+import os
+import requests
+from time import sleep
+
+# --- Configuration ---
+TCGDEX_API = "https://api.tcgdex.net/v2/de/cards"
+OUTPUT_FOLDER = "cards"
+REQUEST_DELAY = 0.1  # seconds between requests to avoid rate limiting
+
+# Create output folder if not exists
+os.makedirs(OUTPUT_FOLDER, exist_ok=True)
+
+# Fetch card list from TCGdex
+print("Fetching card list...")
+resp = requests.get(TCGDEX_API)
+if resp.status_code != 200:
+    raise Exception(f"Failed to fetch card list: {resp.status_code}")
+cards = resp.json()
+print(f"Total cards fetched: {len(cards)}")
+
+# Download each card image
+for card in cards:
+    card_id = card.get("id", None)
+    image_base = card.get("image", None)
+
+    if not card_id:
+        print("Skipping card with missing ID:", card)
+        continue
+
+    if not image_base:
+        print(f"No image URL for {card_id}, skipping...")
+        continue
+
+    image_url = image_base + "/high.png"
+    output_path = os.path.join(OUTPUT_FOLDER, f"{card_id}.png")
+
+    # Skip if already downloaded
+    if os.path.exists(output_path):
+        print(f"Already exists: {card_id}")
+        continue
+
+    try:
+        r = requests.get(image_url, stream=True)
+        if r.status_code == 200:
+            with open(output_path, "wb") as f:
+                for chunk in r.iter_content(1024):
+                    f.write(chunk)
+            print(f"Downloaded: {card_id}")
+        else:
+            print(f"Failed to download {card_id}: HTTP {r.status_code}")
+    except Exception as e:
+        print(f"Error downloading {card_id}: {e}")
+
+    sleep(REQUEST_DELAY)  # small delay to be polite
+
+print("All done!")

pythonScripts/fetchNames.py

@@ -0,0 +1,37 @@
+import os
+import requests
+from time import sleep
+
+# --- Configuration ---
+TCGDEX_API = "https://api.tcgdex.net/v2/de/cards"
+OUTPUT_FOLDER = "names"
+REQUEST_DELAY = 0.1  # seconds between requests to avoid rate limiting
+
+# Create output folder if not exists
+os.makedirs(OUTPUT_FOLDER, exist_ok=True)
+
+# Fetch card list from TCGdex
+print("Fetching card list...")
+resp = requests.get(TCGDEX_API)
+if resp.status_code != 200:
+    raise Exception(f"Failed to fetch card list: {resp.status_code}")
+cards = resp.json()
+print(f"Total cards fetched: {len(cards)}")
+
+names = set()  # using a set avoids duplicates automatically
+
+for card in cards:
+    card_name = card.get("name")
+    if not card_name:
+        print("Skipping card with missing name:", card)
+        continue
+    if "◇" in card_name:
+        continue
+    names.add(card_name)  # set ignores duplicates
+
+output_path = os.path.join(OUTPUT_FOLDER, "name.txt")
+with open(output_path, "w", encoding="utf-8") as f:
+    for name in names:
+        f.write("'" + name + "',\n")
+
+print(f"Wrote {len(names)} unique names to {output_path}")

pythonService/app.py

@@ -23,7 +23,7 @@ BASE = os.path.dirname(os.path.abspath(__file__))
 FAISS_INDEX_FILE = os.path.join(BASE, "card_index.faiss")
 EMBEDDINGS_FILE = os.path.join(BASE, "embeddings.npy")
 IDS_FILE = os.path.join(BASE, "ids.npy")
-TOP_K = 5
+TOP_K = 3
 
 # --- Load CLIP model ---
 device = "cuda" if torch.cuda.is_available() else "cpu"

src/api.ts

@@ -0,0 +1,18 @@
+export async function getImage(req: Bun.BunRequest<"/api/cards/:id">) {
+    const { id } = req.params;
+    let file = Bun.file(`images/cards/${id}.png`);
+    let type = "png";
+
+    if (!(await file.exists())) {
+        file = Bun.file(`images/cards/${id}.jpg`);
+        type = "jpg";
+
+        if (!(await file.exists())) {
+            file = Bun.file("images/cards/placeholder.png");
+            type = "png";
+            console.error(`File for image ${id} does not exist, serving placeholder`);
+        }
+    }
+
+    return new Response(file, { headers: { "Content-Type": `image/${type}` }, status: 200 });
+}

src/embeddings.ts

@@ -1,21 +0,0 @@
-import type { Card } from "./types";
-
-// Placeholder cards array
-let cards: Card[] = [
-    { id: "swsh1-1", name: "Celebi V", set: "Swsh1", number: "1", imageUrl: "https://assets.tcgdex.net/de/swsh/swsh1/1/high.png" },
-    { id: "swsh12-001", name: "Bluzuk", set: "Swsh12", number: "001", imageUrl: "https://assets.tcgdex.net/de/swsh/swsh12/001/high.png" },
-];
-
-export function loadCards() {
-    // Placeholder: you can later load embeddings.npy + FAISS
-    console.log("Cards module loaded (currently empty)");
-}
-
-export function queryCardById(id: string): Card | null {
-    return cards.find(c => c.id === id) || null;
-}
-
-// Example placeholder: return top N matches
-export function queryCardByEmbedding(/* embedding */): Card[] {
-    return cards.slice(0, 5); // dummy top 5
-}

src/index.ts

@@ -1,6 +1,5 @@
 import { serve, spawn } from 'bun';
-import { queryCardByEmbedding, queryCardById } from './embeddings';
-import type { Card } from './types';
+import { getImage } from './api';
 
 const PYTHON_SERVICE = "http://localhost:5001/query";
 
@@ -27,29 +26,18 @@ const server = serve({
                 }
             },
         },
-        "/api/cards/:id": async (req) => {
-            const { id } = req.params;
-            const card = queryCardById(id);
-            if (!card) return new Response(JSON.stringify({ error: "Card not found" }), { status: 404, headers: { "Content-Type": "application/json" } });
-            return new Response(JSON.stringify(card), { headers: { "Content-Type": "application/json" } });
+        "/api/cards/image/:id": {
+            async GET(req) {
+                return getImage(req);
+                /* const { id } = req.params;
+                const card = queryCardById(id);
+                if (!card) return new Response(JSON.stringify({ error: "Card not found" }), { status: 404, headers: { "Content-Type": "application/json" } });
+                return new Response(JSON.stringify(card), { headers: { "Content-Type": "application/json" } }); */
+            }
         },
         "/*": async () => {
             return new Response("<h1>Pokemon Card Backend</h1>", { headers: { "Content-Type": "text/html" } });
         },
-
-        /* "/api/cards/query": {
-            async POST(req) {
-                try {
-                    const { embedding } = await req.json() as { embedding: number[] };
-                    if (!embedding) return new Response(JSON.stringify({ error: "Missing embedding" }), { status: 400, headers: { "Content-Type": "application/json" } });
-                    const results: Card[] = queryCardByEmbedding(embedding);
-                    return new Response(JSON.stringify(results), { headers: { "Content-Type": "application/json" } });
-                } catch (err) {
-                    console.error("Error querying card:", err);
-                    return new Response(JSON.stringify({ error: "Failed to query card" }), { status: 500, headers: { "Content-Type": "application/json" } });
-                }
-            },
-        }, */
     },
 
     development: process.env.NODE_ENV !== 'production' && {
@@ -61,4 +49,4 @@ const server = serve({
     },
 });
 
-console.log(`🚀 Server running at ${server.url}`);
+console.log(`Server running at ${server.url}`);

src/types.ts

@@ -1,10 +0,0 @@
-export interface Card {
-    id: string;
-    name: string;
-    set: string;
-    number: string;
-    rarity?: string;
-    variant?: string;
-    foil?: boolean;
-    imageUrl: string;
-}

tsconfig.json

@@ -1,7 +1,7 @@
 {
   "compilerOptions": {
     // Environment setup & latest features
-    "lib": ["ESNext"],
+    "lib": ["ESNext", "DOM"],
     "target": "ESNext",
     "module": "Preserve",
     "moduleDetection": "force",