Update app.py

app.py

@@ -1,59 +1,35 @@
-# influencer_match_app.py
+
 import os
-import re
 import random
 import pandas as pd
 import gradio as gr
 from sentence_transformers import SentenceTransformer, util
 
 # =========================
-# SYNTHETIC DATA (HF) CONFIG
+# ALWAYS-GENERATE (HF) DATASET
 # =========================
 CSV_PATH = "synthetic_influencers.csv"
-NUM_ROWS = 1200      # >= 1000 for the assignment
-SEED = 42            # reproducibility
+NUM_ROWS = 1200       # ≥1000 as required
+SEED = 42             # reproducibility
+
+def create_synthetic_influencer_dataset(n=1200, out_csv="synthetic_influencers.csv", seed=42):
+    """
+    Creates a synthetic dataset that mirrors your current schema:
+    Columns: Rank, Name, Followers, ER, Country, Niche, Reach, Source File, Source Path
+    Uses a Hugging Face text-generation model to generate influencer names.
+    """
+    random.seed(seed)
 
-# -------------------------
-# Hugging Face name generator
-# -------------------------
-def _load_name_pipeline():
+    # Lazy import to keep dependencies minimal in environments where transformers isn't preinstalled
     try:
         from transformers import pipeline
     except Exception as e:
         raise RuntimeError(
-            "Transformers not installed. Install with: pip install transformers torch --upgrade"
+            "Transformers not installed. Install with: pip install transformers torch"
         ) from e
-        
-    # small model for faster startup on HF Spaces (used for generating synthetic dataset)
-    return pipeline("text-generation", model="distilgpt2")
-
-NAME_PIPE = _load_name_pipeline()
-
-def generate_hf_name(niches_list, platform, country):
-    niche_phrase = ", ".join(niches_list)
-    prompt = (
-        "Generate ONE short, brandable influencer username.\n"
-        f"Context: niches = [{niche_phrase}], country = {country}, platform = {platform}.\n"
-        "Constraints: 1–3 words, letters/numbers only, no emojis, no spaces, no hyphens, "
-        "avoid the word 'Creator', avoid generic words like 'User' or 'Channel'. "
-        "Prefer CamelCase or concise compounds (e.g., FitAtlas, StyleLab, CodeBytes). "
-        "Return ONLY the username:"
-    )
-    out = NAME_PIPE(prompt, max_new_tokens=14, do_sample=True, temperature=0.9, top_p=0.92)[0]["generated_text"]
-    name = out.strip().splitlines()[-1].strip()
-    # keep alphanum, remove spaces/hyphens just in case
-    import re
-    name = re.sub(r"[^A-Za-z0-9]", "", name)
-    if not name or len(name) > 24:
-        # still simple: if the model slips, lightly trim
-        name = name[:24] or "BrandHandle"
-    return name
-
-# -------------------------
-# Synthetic dataset builder
-# -------------------------
-def create_synthetic_influencer_dataset(n=1200, out_csv=CSV_PATH, seed=SEED) -> pd.DataFrame:
-    random.seed(seed)
+
+    # Use a lighter model to keep HF Space startup snappy (swap to 'gpt2' if you prefer)
+    name_gen = pipeline("text-generation", model="distilgpt2")
 
     countries = [
         "USA","UK","Canada","Australia","Brazil","India","France","Germany","Italy","Spain",
@@ -61,27 +37,33 @@ def create_synthetic_influencer_dataset(n=1200, out_csv=CSV_PATH, seed=SEED) ->
     ]
     niches = [
         "Fashion","Travel","Food","Fitness","Tech","Beauty","Gaming","Music","Photography",
-        "Lifestyle","Education","Finance","Sports","Parenting","DIY", "Fashion & Travel", "Food & Fitness",
-        "Tech & Gaming", "Lifestyle & Travel", "Sports & Fitness"
+        "Lifestyle","Education","Finance","Sports","Parenting","DIY"
     ]
-    platforms = ["youtube", "instagram", "tiktok", "twitch", "x"]  # lowercase for file prefix
+    platforms = ["youtube", "instagram", "tiktok", "twitch", "x"]  # lowercase -> file prefix
 
     rows = []
     for rank in range(1, n + 1):
-        # --- Structured fields first (so name prompt can use them) ---
+        # --- Name via HF text generation ---
+        prompt = "Short, catchy influencer name:"
+        raw = name_gen(prompt, max_new_tokens=8, num_return_sequences=1, do_sample=True, temperature=0.9)[0]["generated_text"]
+        # cleanup
+        name = raw.replace(prompt, "").strip().split("\n")[0]
+        name = "".join(ch for ch in name if ch.isalnum() or ch in " -_.").strip()
+        if len(name) < 3 or len(name) > 40:
+            name = f"Creator_{rank}"
+
+        # --- Structured fields sampled to look realistic ---
+        followers = random.randint(5_000, 5_000_000)
+        er = round(random.uniform(0.5, 15.0), 2)  # %
         country = random.choice(countries)
         niche = random.choice(niches)
-        platform_token = random.choice(platforms)  # 'youtube', 'instagram', ...
-        followers = random.randint(5_000, 5_000_000)
-        er = round(random.uniform(0.5, 15.0), 2)  # percent
         reach = int(followers * random.uniform(0.25, 0.95))
+
+        platform_token = random.choice(platforms)  # e.g., 'youtube'
         region_hint = country.lower().replace(" ", "")
-        source_file = f"{platform_token}_data_{region_hint}.csv"   # first token remains the platform
+        source_file = f"{platform_token}_data_{region_hint}.csv"  # <- first token = platform
         source_path = f"synthetic/{source_file}"
 
-        # --- Now generate the name using those fields (HF model) ---
-        name = generate_hf_name(niche, platform_token, country)
-
         rows.append([
             rank, name, followers, er, country, niche, reach, source_file, source_path
         ])
@@ -92,8 +74,8 @@ def create_synthetic_influencer_dataset(n=1200, out_csv=CSV_PATH, seed=SEED) ->
     df_syn.to_csv(out_csv, index=False)
     return df_syn
 
-def build_and_load_synthetic() -> pd.DataFrame:
-    # Always rebuild to ensure HF-generated data is used
+def load_or_build_synthetic():
+    # Always build (or rebuild) to guarantee HF-generated data is used
     if os.path.exists(CSV_PATH):
         os.remove(CSV_PATH)
     _ = create_synthetic_influencer_dataset(n=NUM_ROWS, out_csv=CSV_PATH, seed=SEED)
@@ -101,13 +83,11 @@ def build_and_load_synthetic() -> pd.DataFrame:
     df_local.fillna("", inplace=True)
     return df_local
 
-# =========================
-# LOAD DATA (always synthetic)
-# =========================
-df = build_and_load_synthetic()
+# Build & load synthetic dataset
+df = load_or_build_synthetic()
 
 # =========================
-# FEATURE ENGINEERING (same logic)
+# FEATURE ENGINEERING (same as yours)
 # =========================
 # Extract platform name from Source File (first token before '_'), capitalize
 df['Platform'] = df['Source File'].astype(str).str.split('_').str[0].str.capitalize()
@@ -117,12 +97,12 @@ profile_fields = ["Name", "Platform", "Niche", "Country"]
 df["profile_text"] = df[profile_fields].agg(" - ".join, axis=1)
 
 # =========================
-# EMBEDDINGS & RECOMMENDER
+# EMBEDDINGS & RECOMMENDER (same as yours)
 # =========================
 model = SentenceTransformer("sentence-transformers/all-MiniLM-L6-v2")
 influencer_embeddings = model.encode(df["profile_text"].tolist(), convert_to_tensor=True)
 
-def recommend_influencers(brand_description: str):
+def recommend_influencers(brand_description):
     query_embedding = model.encode(brand_description, convert_to_tensor=True)
     cosine_scores = util.pytorch_cos_sim(query_embedding, influencer_embeddings)[0]
     top_indices = cosine_scores.topk(3).indices.tolist()
@@ -136,12 +116,12 @@ def recommend_influencers(brand_description: str):
             "Niche": row["Niche"],
             "Country": row["Country"],
             "ER": f"{row.get('ER', 'N/A')}",
-            "Followers": int(row["Followers"]),
-            "Reach": int(row["Reach"]) if str(row.get("Reach", "")).isdigit() else row.get("Reach", "")
+            "Followers": row["Followers"],
+            "Reach": row.get("Reach", "")
         })
     return recs
 
-def format_output(brand_input: str):
+def format_output(brand_input):
     recs = recommend_influencers(brand_input)
     html = ""
     for i, rec in enumerate(recs, 1):
@@ -151,8 +131,8 @@ def format_output(brand_input: str):
           <p style='margin:0.5em 0;'><strong>Niche:</strong> {rec['Niche']}</p>
           <p style='margin:0.5em 0;'><strong>Country:</strong> {rec['Country']}</p>
           <p style='margin:0.5em 0;'><strong>Engagement:</strong> {rec['ER']}%</p>
-          <p style='margin:0.5em 0;'><strong>Followers:</strong> {rec['Followers']:,}</p>
-          {f"<p style='margin:0.5em 0;'><strong>Reach:</strong> {int(rec['Reach']):,}</p>" if isinstance(rec['Reach'], int) else ""}
+          <p style='margin:0.5em 0;'><strong>Followers:</strong> {rec['Followers']}</p>
+          {f"<p style='margin:0.5em 0;'><strong>Reach:</strong> {rec['Reach']}</p>" if rec['Reach'] else ""}
         </div>
         """
     return html
@@ -176,9 +156,9 @@ iface = gr.Interface(
     article=(
         "**Project:** AI-Powered Influencer Recommender for Social Media Marketing\n\n"
         "**Models:**\n"
-        "- Hugging Face text-generation (distilgpt2) to create a 1,200-row synthetic influencer dataset at runtime\n"
+        "- text-generation (Hugging Face) for synthetic influencer names (dataset creation)\n"
         "- sentence-transformers/all-MiniLM-L6-v2 for semantic embeddings (recommendations)\n\n"
-        "**Dataset:** Synthetic dataset suitable for text modality of influencer data"
+        "**Dataset:** 1,200-row synthetic influencer dataset generated at runtime."
     ),
     examples=[
         ["Sustainable fashion campaign targeting eco-conscious millennials"],
@@ -192,3 +172,4 @@ iface = gr.Interface(
 
 if __name__ == "__main__":
     iface.launch(share=True)
+