first commit

.gitattributes

@@ -0,0 +1,36 @@
+*.7z filter=lfs diff=lfs merge=lfs -text
+*.arrow filter=lfs diff=lfs merge=lfs -text
+*.bin filter=lfs diff=lfs merge=lfs -text
+*.bz2 filter=lfs diff=lfs merge=lfs -text
+*.ckpt filter=lfs diff=lfs merge=lfs -text
+*.ftz filter=lfs diff=lfs merge=lfs -text
+*.gz filter=lfs diff=lfs merge=lfs -text
+*.h5 filter=lfs diff=lfs merge=lfs -text
+*.joblib filter=lfs diff=lfs merge=lfs -text
+*.lfs.* filter=lfs diff=lfs merge=lfs -text
+*.mlmodel filter=lfs diff=lfs merge=lfs -text
+*.model filter=lfs diff=lfs merge=lfs -text
+*.msgpack filter=lfs diff=lfs merge=lfs -text
+*.npy filter=lfs diff=lfs merge=lfs -text
+*.npz filter=lfs diff=lfs merge=lfs -text
+*.onnx filter=lfs diff=lfs merge=lfs -text
+*.ot filter=lfs diff=lfs merge=lfs -text
+*.parquet filter=lfs diff=lfs merge=lfs -text
+*.pb filter=lfs diff=lfs merge=lfs -text
+*.pickle filter=lfs diff=lfs merge=lfs -text
+*.pkl filter=lfs diff=lfs merge=lfs -text
+*.pt filter=lfs diff=lfs merge=lfs -text
+*.pth filter=lfs diff=lfs merge=lfs -text
+*.rar filter=lfs diff=lfs merge=lfs -text
+*.safetensors filter=lfs diff=lfs merge=lfs -text
+saved_model/**/* filter=lfs diff=lfs merge=lfs -text
+*.tar.* filter=lfs diff=lfs merge=lfs -text
+*.tar filter=lfs diff=lfs merge=lfs -text
+*.tflite filter=lfs diff=lfs merge=lfs -text
+*.tgz filter=lfs diff=lfs merge=lfs -text
+*.wasm filter=lfs diff=lfs merge=lfs -text
+*.xz filter=lfs diff=lfs merge=lfs -text
+*.zip filter=lfs diff=lfs merge=lfs -text
+*.zst filter=lfs diff=lfs merge=lfs -text
+*tfevents* filter=lfs diff=lfs merge=lfs -text
+*.pdf filter=lfs diff=lfs merge=lfs -text

.gitignore

@@ -0,0 +1,90 @@
+# ==============================
+# Python
+# ==============================
+__pycache__/
+*.py[cod]
+*.pyo
+*.pyd
+*.so
+*.egg-info/
+*.egg
+
+# ==============================
+# Virtual Environment
+# ==============================
+venv/
+.env/
+.env.*
+
+# ==============================
+# OS / Editor files
+# ==============================
+.DS_Store
+Thumbs.db
+.vscode/
+.idea/
+*.swp
+*.swo
+
+# ==============================
+# Logs & Runtime Files
+# ==============================
+*.log
+logs/
+*.pid
+
+# ==============================
+# Python build artifacts
+# ==============================
+build/
+dist/
+*.whl
+
+# ==============================
+# Jupyter / Notebooks
+# ==============================
+.ipynb_checkpoints/
+
+# ==============================
+# FAISS / Vector Store Artifacts
+# ==============================
+data/faiss/
+*.faiss
+*.pkl
+
+# ==============================
+# Temporary data / caches
+# ==============================
+.cache/
+.tmp/
+.tmp_cache/
+
+# ==============================
+# API keys / secrets
+# ==============================
+.env
+config/.env
+config/settings.local.py
+
+# ==============================
+# Gradio / UI cache
+# ==============================
+.gradio/
+
+# ==============================
+# OS-specific
+# ==============================
+*.DS_Store
+desktop.ini
+
+# ==============================
+# Test output
+# ==============================
+.coverage
+htmlcov/
+
+# ==============================
+# Ignore downloaded PDFs (optional)
+# Uncomment if you don’t want PDFs tracked
+# ==============================
+# data/pdfs/*

README.md

@@ -0,0 +1,12 @@
+---
+title: AccessoryIQ
+emoji: 🏃
+colorFrom: gray
+colorTo: red
+sdk: gradio
+sdk_version: 5.49.0
+app_file: ui/app.py
+pinned: false
+---
+
+Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference

app/agents/evidence_agent.py

@@ -0,0 +1,125 @@
+from app.rag.vector_store import VectorStore
+
+
+class EvidenceAgent:
+    """
+    RAG-FIRST Evidence Agent
+    - Uses RAG if possible
+    - Signals when search fallback is needed
+    """
+
+    def __init__(self):
+        self.vs = VectorStore()
+        self.vs.load()
+
+    def gather(self, category, brand, model, accessory_type):
+        rag_chunks = self.vs.search(
+            query=f"{brand} {accessory_type} compatible accessories",
+            category=category,
+            brand=brand
+        )
+
+        supported_models = self._extract_supported_models(rag_chunks)
+
+        # 🚨 MODEL NOT FOUND → TRIGGER SEARCH
+        if not self._model_supported(model, supported_models):
+            return {
+                "needs_search": True,
+                "reason": f"Model '{model}' not found in official documentation"
+            }
+
+        # Extract accessories from RAG
+        result = self._extract_accessories(rag_chunks, model, accessory_type)
+
+        if not result["recommended"]:
+            return {
+                "needs_search": True,
+                "reason": "No accessories found in official documentation"
+            }
+
+        return result
+
+    # ----------------------------
+    # MODEL EXTRACTION
+    # ----------------------------
+    def _extract_supported_models(self, chunks):
+        models = set()
+
+        for chunk in chunks:
+            lines = [l.strip() for l in chunk["text"].splitlines()]
+            in_section = False
+
+            for line in lines:
+                if "supported models" in line.lower():
+                    in_section = True
+                    continue
+
+                if in_section:
+                    if not line.startswith(("•", "-")):
+                        break
+                    models.add(line.lstrip("•- ").strip().lower())
+
+        return models
+
+    def _model_supported(self, model, supported_models):
+        model = model.lower()
+        return any(model in m or m in model for m in supported_models)
+
+    # ----------------------------
+    # ACCESSORY EXTRACTION
+    # ----------------------------
+    def _extract_accessories(self, chunks, model, accessory_type):
+        SECTION_MAP = {
+            "charging": ["charging", "power"],
+            "audio": ["audio"],
+            "cooling": ["cooling"],
+            "display": ["display"],
+            "connectivity": ["connectivity"],
+            "storage": ["storage"],
+            "controller": ["controller"]
+        }
+
+        valid_sections = SECTION_MAP.get(accessory_type.lower(), [])
+
+        recommended = []
+
+        for chunk in chunks:
+            source = chunk["source_pdf"]
+            lines = [l.strip() for l in chunk["text"].splitlines() if l.strip()]
+
+            current_section = None
+
+            for line in lines:
+                lower = line.lower()
+
+                for sec in valid_sections:
+                    if sec in lower:
+                        current_section = sec
+                        break
+
+                if not current_section:
+                    continue
+
+                if "compatible" in lower:
+                    continue
+
+                if line.startswith(("•", "-", "o")):
+                    recommended.append({
+                        "accessory": line.lstrip("•-o ").strip(),
+                        "reason": "Listed as a compatible accessory in official specifications",
+                        "sources": [source]
+                    })
+
+        return {
+            "recommended": self._dedupe(recommended),
+            "avoid": []
+        }
+
+    def _dedupe(self, items):
+        seen = set()
+        out = []
+        for item in items:
+            if item["accessory"] not in seen:
+                seen.add(item["accessory"])
+                out.append(item)
+        return out

app/agents/planner_agent.py

@@ -0,0 +1,43 @@
+class PlannerAgent:
+    """
+    PHASE 1 Planner: Deterministic Formatter (Clean Output)
+
+    - No LLM
+    - No hallucination
+    - Groups sources cleanly
+    """
+
+    def decide(self, evidence, product, use_case):
+        if "refusal" in evidence:
+            return f"REFUSAL:\n{evidence['refusal']}"
+
+        recommended = evidence.get("recommended", [])
+        avoid = evidence.get("avoid", [])
+
+        if not recommended:
+            return "REFUSAL:\nNo compatible accessories found in official specifications."
+
+        lines = []
+        lines.append("Recommended:")
+
+        # Collect unique sources
+        all_sources = set()
+
+        for item in recommended:
+            lines.append(f"- {item['accessory']}")
+            all_sources.update(item.get("sources", []))
+
+        if avoid:
+            lines.append("\nAvoid:")
+            for item in avoid:
+                lines.append(f"- {item['accessory']}")
+                all_sources.update(item.get("sources", []))
+
+        lines.append("\nSources:")
+        for src in sorted(all_sources):
+            lines.append(f"- {src}")
+
+        confidence = min(0.9, 0.6 + 0.1 * len(recommended))
+        lines.append(f"\nConfidence: {round(confidence, 2)}")
+
+        return "\n".join(lines)

app/agents/search_agent.py

@@ -0,0 +1,115 @@
+import requests
+from urllib.parse import urlparse
+from config.settings import settings
+import logging
+
+
+class SearchAgent:
+    """
+    Tiered Search Agent with confidence scoring.
+    """
+
+    OFFICIAL_DOMAINS = {
+        "apple.com",
+        "support.apple.com",
+        "developer.apple.com",
+        "samsung.com",
+        "sony.com",
+        "playstation.com",
+        "dell.com",
+        "hp.com",
+        "asus.com",
+        "nvidia.com",
+        "amd.com"
+    }
+
+    COMMUNITY_DOMAINS = {
+        "reddit.com",
+        "stackoverflow.com",
+        "superuser.com",
+        "ifixit.com"
+    }
+
+    def search(self, product, accessory):
+        logging.info(f"[SEARCH] Triggered for: {product} | {accessory}")
+
+        query = f"{product} {accessory} compatibility"
+
+        headers = {
+            "X-API-KEY": settings.SERPER_API_KEY,
+            "Content-Type": "application/json"
+        }
+
+        payload = {"q": query, "num": 10}
+
+        try:
+            r = requests.post(
+                "https://google.serper.dev/search",
+                headers=headers,
+                json=payload,
+                timeout=10
+            )
+            r.raise_for_status()
+            data = r.json()
+        except Exception as e:
+            logging.error(f"[SEARCH ERROR] {e}")
+            return self._refusal()
+
+        return self._process_results(data)
+
+    def _process_results(self, data):
+        organic = data.get("organic", [])
+
+        official = []
+        community = []
+
+        for r in organic:
+            url = r.get("link", "")
+            domain = urlparse(url).netloc.lower()
+
+            entry = {
+                "title": r.get("title", "Accessory"),
+                "snippet": r.get("snippet", ""),
+                "url": url
+            }
+
+            if any(d in domain for d in self.OFFICIAL_DOMAINS):
+                official.append(entry)
+            elif any(d in domain for d in self.COMMUNITY_DOMAINS):
+                community.append(entry)
+
+        if official:
+            return self._format(
+                official,
+                confidence=0.9,
+                label="Official documentation"
+            )
+
+        if community:
+            return self._format(
+                community,
+                confidence=0.65,
+                label="Community-verified sources"
+            )
+
+        return self._refusal()
+
+    def _format(self, results, confidence, label):
+        lines = ["Recommended:"]
+
+        for r in results[:3]:
+            lines.append(f"- {r['title']}")
+            lines.append(f"  Reason: {r['snippet']}")
+            lines.append(f"  Source: {r['url']}")
+
+        lines.append(f"\nConfidence: {confidence}")
+        lines.append(f"Source Type: {label}")
+
+        return "\n".join(lines)
+
+    def _refusal(self):
+        return (
+            "REFUSAL:\n"
+            "No reliable official or community evidence found.\n"
+            "Confidence: 0.0"
+        )

app/pipeline/main_pipeline.py

@@ -0,0 +1,28 @@
+from app.agents.evidence_agent import EvidenceAgent
+from app.agents.search_agent import SearchAgent
+from app.agents.planner_agent import PlannerAgent
+import logging
+
+
+def run_pipeline(category, model, accessory, use_case):
+    logging.info(f"[PIPELINE] {category=} {model=} {accessory=}")
+
+    evidence_agent = EvidenceAgent()
+    search_agent = SearchAgent()
+    planner = PlannerAgent()
+
+    evidence = evidence_agent.gather(
+        category=category,
+        brand=model.split()[0],
+        model=model,
+        accessory_type=accessory
+    )
+
+    # --- RAG FAILS → SEARCH ---
+    if evidence.get("needs_search"):
+        logging.info("[PIPELINE] RAG failed → invoking SearchAgent")
+        return search_agent.search(model, accessory)
+
+    # --- RAG SUCCESS ---
+    logging.info("[PIPELINE] RAG successful → Planner formatting")
+    return planner.decide(evidence, model, use_case)

app/prompts/planner_prompt.py

@@ -0,0 +1,45 @@
+PLANNER_SYSTEM_PROMPT = """
+You are a PRODUCT ACCESSORY INTELLIGENCE PLANNER.
+
+You are a STRICT DECISION LAYER.
+You do NOT retrieve data.
+You do NOT invent facts.
+
+========================
+MANDATORY RULES
+========================
+
+1. You MUST use ONLY the provided evidence.
+2. You MUST recommend ONLY PHYSICAL ACCESSORIES.
+   - Examples: charger, cable, adapter, hub, dock, stand
+   - NOT allowed: charging, audio, connectivity, display
+3. You MUST NOT guess or generalize.
+4. You MUST NOT merge speculative sources with official specifications.
+5. If official specifications (PDFs) are present, they take priority over web sources.
+6. If evidence is inconsistent, weak, or invalid, you MUST refuse.
+
+========================
+OUTPUT FORMAT (ONLY)
+========================
+
+Recommended:
+- <Accessory Name>
+  Reason: <Evidence-based reason>
+  Sources:
+    - <PDF name OR URL>
+
+Avoid:
+- <Accessory Name>
+  Reason: <Evidence-based reason>
+  Sources:
+    - <PDF name OR URL>
+
+Confidence: <number between 0 and 1>
+
+========================
+REFUSAL FORMAT
+========================
+
+REFUSAL:
+<Clear explanation of why evidence is insufficient or invalid>
+"""

app/rag/build_index.py

@@ -0,0 +1,26 @@
+import os
+from app.rag.pdf_loader import load_pdf
+from app.rag.chunking import chunk_text
+from app.rag.vector_store import VectorStore
+
+PDF_DIR = "data/pdfs"
+
+texts = []
+metadata = []
+
+for pdf in os.listdir(PDF_DIR):
+    if not pdf.endswith(".pdf"):
+        continue
+
+    pages = load_pdf(os.path.join(PDF_DIR, pdf))
+    chunks = chunk_text(pages)
+
+    for chunk in chunks:
+        texts.append(chunk["text"])
+        metadata.append(chunk)
+
+vs = VectorStore()
+vs.build(texts, metadata)
+vs.save()
+
+print("FAISS index built successfully.")

app/rag/chunking.py

@@ -0,0 +1,16 @@
+def chunk_text(pages, chunk_size=500, overlap=100):
+    chunks = []
+    for page in pages:
+        text = page["text"]
+        start = 0
+        while start < len(text):
+            end = start + chunk_size
+            chunks.append({
+                "text": text[start:end],
+                "page": page["page"],
+                "category": page["category"],
+                "brand": page["brand"],
+                "source_pdf": page["source_pdf"]
+            })
+            start += chunk_size - overlap
+    return chunks

app/rag/embeddings.py

@@ -0,0 +1,6 @@
+from sentence_transformers import SentenceTransformer
+
+model = SentenceTransformer("all-MiniLM-L6-v2")
+
+def embed_texts(texts):
+    return model.encode(texts, show_progress_bar=False)

app/rag/pdf_loader.py

@@ -0,0 +1,39 @@
+import os
+from pypdf import PdfReader
+
+def extract_metadata_from_filename(filename: str):
+    name = filename.lower()
+
+    if "iphone" in name or "galaxy" in name:
+        category = "mobile"
+    elif "tv" in name:
+        category = "tv"
+    elif "laptop" in name:
+        category = "laptop"
+    elif "playstation" in name:
+        category = "gaming"
+    else:
+        category = "unknown"
+
+    brand = name.split("_")[0]
+
+    return {
+        "category": category,
+        "brand": brand,
+        "source_pdf": filename
+    }
+
+
+def load_pdf(path: str):
+    reader = PdfReader(path)
+    filename = os.path.basename(path)
+    base_metadata = extract_metadata_from_filename(filename)
+
+    pages = []
+    for i, page in enumerate(reader.pages):
+        pages.append({
+            "text": page.extract_text(),
+            "page": i + 1,
+            **base_metadata
+        })
+    return pages

app/rag/vector_store.py

@@ -0,0 +1,54 @@
+import faiss
+import os
+import pickle
+from app.rag.embeddings import embed_texts
+from config.settings import settings
+
+
+class VectorStore:
+    def __init__(self):
+        self.index = None
+        self.metadata = []
+
+    def build(self, texts, metadata):
+        vectors = embed_texts(texts)
+        dim = vectors.shape[1]
+        self.index = faiss.IndexFlatL2(dim)
+        self.index.add(vectors)
+        self.metadata = metadata
+
+    def save(self):
+        os.makedirs(settings.VECTOR_STORE_PATH, exist_ok=True)
+        faiss.write_index(
+            self.index,
+            os.path.join(settings.VECTOR_STORE_PATH, "index.faiss")
+        )
+        with open(
+            os.path.join(settings.VECTOR_STORE_PATH, "meta.pkl"),
+            "wb"
+        ) as f:
+            pickle.dump(self.metadata, f)
+
+    def load(self):
+        self.index = faiss.read_index(
+            os.path.join(settings.VECTOR_STORE_PATH, "index.faiss")
+        )
+        with open(
+            os.path.join(settings.VECTOR_STORE_PATH, "meta.pkl"),
+            "rb"
+        ) as f:
+            self.metadata = pickle.load(f)
+
+    def search(self, query, category, brand, k=5):
+        vector = embed_texts([query])
+        distances, indices = self.index.search(vector, k * 3)
+
+        results = []
+        for idx in indices[0]:
+            meta = self.metadata[idx]
+            if meta["category"] == category and meta["brand"] == brand:
+                results.append(meta)
+            if len(results) >= k:
+                break
+
+        return results

app/scoring/evidence_scoring.py

@@ -0,0 +1,7 @@
+def score_accessory(spec_ok, review_hits, complaints):
+    if not spec_ok:
+        return 0.0
+
+    score = (review_hits * 0.3) + 0.5
+    score -= complaints * 0.4
+    return max(0.0, min(score, 1.0))

app/search/search_tool.py

@@ -0,0 +1,13 @@
+import requests
+from config.settings import settings
+
+def search_reviews(query):
+    headers = {"X-API-KEY": settings.SERPER_API_KEY}
+    payload = {"q": query, "num": 5}
+    response = requests.post(
+        "https://google.serper.dev/search",
+        headers=headers,
+        json=payload
+    )
+    response.raise_for_status()
+    return response.json().get("organic", [])

config/settings.py

@@ -0,0 +1,28 @@
+import os
+from dotenv import load_dotenv
+
+load_dotenv()
+
+class Settings:
+    # OpenRouter
+    OPENROUTER_API_KEY = os.getenv("OPENROUTER_API_KEY")
+    OPENROUTER_BASE_URL = "https://openrouter.ai/api/v1/chat/completions"
+
+    DEFAULT_MODEL = os.getenv(
+        "OPENROUTER_MODEL",
+        "meta-llama/llama-3-8b-instruct"
+    )
+    FALLBACK_MODEL = os.getenv(
+        "OPENROUTER_FALLBACK_MODEL",
+        "mistralai/mixtral-8x7b-instruct"
+    )
+
+    TEMPERATURE = 0.0
+
+    # Search
+    SERPER_API_KEY = os.getenv("SERPER_API_KEY")
+
+    # RAG
+    VECTOR_STORE_PATH = "data/faiss_index"
+
+settings = Settings()

data/pdfs/apple_iphone_2019_2024_specs.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f0ac0e0abcc3da17bc9332cf7225c7911b70f866e9d743517d40eebc9ada542d
+size 172101

data/pdfs/asus_laptops_2020_2024_specs.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f51fdea020ff5b581eac95feb12a82c452bfc861f10681bb851428f235cd96d4
+size 183398

data/pdfs/dell_laptops_2020_2024_specs.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c2fbf5ebe0c1d46b48a4ee34308a48bc59f4d11addc41993f81f1c86cead720a
+size 188906

data/pdfs/hp_laptops_2020_2024_specs.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6b9ca907dd8e5f739898797910cd7c1479bf62dfae24641ac2f9c7ffae486779
+size 185209

data/pdfs/lg_tv_2021_2024_specs.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:66b5a7b1d29629ef60fe55464ca01cfe1b7acd16e395f5dcb11c8c36a48bb0c0
+size 178286

data/pdfs/playstation_2013_2024_specs.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cb93ff8667a762bf1107a25944a61949224288debfa0da94719087d7388781ce
+size 184034

data/pdfs/samsung_galaxy_2020_2024_specs.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:62d520a0d912b1d0185345d70cae63a93561091e20c60298c054dd873d65f0b1
+size 172127

data/pdfs/samsung_tv_2021_2024_specs.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:558c89e698716d22e55f23301240598917a4f3791a96f6b246e9c8147e555e95
+size 181958

requirements.txt

@@ -0,0 +1,7 @@
+python-dotenv
+requests
+faiss-cpu
+sentence-transformers
+pypdf
+langchain
+gradio

ui/app.py

@@ -0,0 +1,100 @@
+import gradio as gr
+from app.pipeline.main_pipeline import run_pipeline
+
+import logging
+
+logging.basicConfig(
+    level=logging.INFO,
+    format="%(asctime)s | %(levelname)s | %(message)s"
+)
+
+# --------------------------------------------------
+# Accessory options derived from PDF scope
+# --------------------------------------------------
+ACCESSORY_MAP = {
+    "Mobile": ["Charging", "Audio", "Protection", "Connectivity"],
+    "Laptop": ["Charging", "Connectivity", "Docking", "Display", "Cooling", "Audio"],
+    "TV": ["Display", "Audio", "Mounting", "Power", "Connectivity"],
+    "Gaming": ["Controller", "Storage", "Audio", "Display", "Power"]
+}
+
+
+# --------------------------------------------------
+# Correct dropdown update (FORCED)
+# --------------------------------------------------
+def update_accessories(category):
+    if not category:
+        return gr.update(choices=[], value=None)
+
+    return gr.update(
+        choices=ACCESSORY_MAP[category],
+        value=ACCESSORY_MAP[category][0],  # force selection
+        interactive=True
+    )
+
+
+# --------------------------------------------------
+# Run pipeline (safe)
+# --------------------------------------------------
+def run(category, model, accessory, use_case):
+    if not category or not model or not accessory:
+        return "REFUSAL:\nPlease select Category, Model, and Accessory Type."
+
+    return run_pipeline(
+        category=category.lower(),
+        model=model.strip(),
+        accessory=accessory.lower(),
+        use_case=use_case.strip() if use_case else ""
+    )
+
+
+# --------------------------------------------------
+# UI Layout
+# --------------------------------------------------
+with gr.Blocks(title="AccessoryIQ") as demo:
+    gr.Markdown("## 🔌 AccessoryIQ – Evidence-Based Accessory Intelligence")
+
+    category = gr.Dropdown(
+        choices=["Mobile", "Laptop", "TV", "Gaming"],
+        label="Product Category",
+        interactive=True
+    )
+
+    accessory = gr.Dropdown(
+        choices=["Select category first"],  # IMPORTANT
+        label="Accessory Type",
+        interactive=True,
+        allow_custom_value=False
+    )
+
+    model = gr.Textbox(
+        label="Product Model",
+        placeholder="e.g., iPhone 15, Dell XPS 13, Samsung QN90C"
+    )
+
+    use_case = gr.Textbox(
+        label="Use Case (optional)",
+        placeholder="e.g., gaming, office work, travel"
+    )
+
+    category.change(
+        fn=update_accessories,
+        inputs=category,
+        outputs=accessory
+    )
+
+    output = gr.Textbox(
+        label="Accessory Recommendation",
+        lines=22,
+        interactive=False
+    )
+
+    submit = gr.Button("Get Recommendation")
+
+    submit.click(
+        fn=run,
+        inputs=[category, model, accessory, use_case],
+        outputs=output
+    )
+
+demo.launch()