Update app.py

app.py

@@ -1,93 +1,64 @@
+import os
 import gradio as gr
 from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
-import requests
-from bs4 import BeautifulSoup
+from openai import OpenAI
 from langdetect import detect
 
-MODEL_NAME = "facebook/mbart-large-50-many-to-many-mmt"
+# --- Hugging Face MT model ---
+MODEL_NAME = "Helsinki-NLP/opus-mt-en-pt"
 tokenizer = AutoTokenizer.from_pretrained(MODEL_NAME)
 model = AutoModelForSeq2SeqLM.from_pretrained(MODEL_NAME)
 
-def translate_text(word):
-    options = []
-
-    # ---- Step 1: Machine Translation with context ----
-    sentence = f"The meaning of '{word}' is:"
-    tokenizer.src_lang = "en_XX"
-    inputs = tokenizer(sentence, return_tensors="pt")
-    generated_tokens = model.generate(**inputs, forced_bos_token_id=tokenizer.lang_code_to_id["pt_XX"])
-    mt_translation = tokenizer.batch_decode(generated_tokens, skip_special_tokens=True)[0].strip()
-
-    if mt_translation and detect(mt_translation) == "pt":
-        options.append(("Machine Translation", mt_translation, "neutral"))
+# --- OpenAI client ---
+client = OpenAI(api_key=os.getenv("OPENAI_API_KEY"))
 
-    # ---- Step 2: Linguee ----
-    linguee_opts = lookup_linguee(word)
-    options.extend(linguee_opts)
+def literal_translation(text):
+    inputs = tokenizer(text, return_tensors="pt")
+    outputs = model.generate(**inputs, max_length=100)
+    return tokenizer.decode(outputs[0], skip_special_tokens=True)
 
-    # ---- Step 3: Dicio enrichment (for each PT candidate) ----
-    enriched = []
-    for src, trans, reg in options:
-        defs = lookup_dicio(trans)
-        if defs:
-            for d in defs:
-                enriched.append(("Dicio", f"{trans} → {d}", "formal"))
-    options.extend(enriched)
+def gpt_explanation(text):
+    prompt = f"""
+You are a bilingual translation assistant. 
+For the English expression "{text}":
+1. Provide a literal Brazilian Portuguese translation.
+2. Provide 2–3 idiomatic or natural Brazilian Portuguese equivalents that capture the real meaning.
+3. Explain in simple terms what the expression means in context.
+Answer clearly, structured as options.
+"""
+    response = client.chat.completions.create(
+        model="gpt-4o-mini",
+        messages=[{"role": "user", "content": prompt}],
+        temperature=0.7
+    )
+    return response.choices[0].message.content
 
-    # ---- Step 4: Fill with synonyms if < 5 ----
-    if len(options) < 5:
-        syns = lookup_sinonimos(options[0][1]) if options else []
-        for s in syns:
-            options.append(("Sinonimos", s, "varies"))
-            if len(options) >= 5:
-                break
-
-    # ---- Format ----
-    if not options:
-        return "❌ No translations found. Try with more context."
-
-    formatted = f"🔎 English: {word}\n\n"
-    for i, (src, trans, reg) in enumerate(options, start=1):
-        formatted += f"Option {i}: {trans}  (source: {src}, register: {reg})\n"
-    return formatted
+def translate_expression(expr):
+    options = []
 
-def lookup_linguee(word):
-    url = f"https://www.linguee.com/english-portuguese/search?source=english&query={word}"
-    r = requests.get(url)
-    if r.status_code != 200:
-        return []
-    soup = BeautifulSoup(r.text, "html.parser")
-    results = []
-    for hit in soup.select(".dictLink"):
-        text = hit.get_text().strip()
-        if text and text.isalpha() and len(text) > 2:
-            results.append(("Linguee", text, "varies"))
-    return results[:3]
+    # 1. Hugging Face literal MT
+    try:
+        lit = literal_translation(expr)
+        if lit and detect(lit) == "pt":
+            options.append(f"Literal (Hugging Face MT): {lit}")
+    except Exception as e:
+        options.append(f"(MT failed: {e})")
 
-def lookup_dicio(word_pt):
-    url = f"https://www.dicio.com.br/{word_pt.lower().replace(' ', '-')}/"
-    r = requests.get(url)
-    if r.status_code != 200:
-        return []
-    soup = BeautifulSoup(r.text, "html.parser")
-    defs = soup.select("p.significado")
-    return [d.get_text().strip() for d in defs[:2]]
+    # 2. GPT-4o-mini idiomatic + explanation
+    try:
+        gpt_out = gpt_explanation(expr)
+        options.append("GPT-4o-mini analysis:\n" + gpt_out)
+    except Exception as e:
+        options.append(f"(GPT failed: {e})")
 
-def lookup_sinonimos(word_pt):
-    url = f"https://www.sinonimos.com.br/{word_pt.lower().replace(' ', '-')}/"
-    r = requests.get(url)
-    if r.status_code != 200:
-        return []
-    soup = BeautifulSoup(r.text, "html.parser")
-    syns = soup.select("a.sinonimo")
-    return [s.get_text().strip() for s in syns[:3]]
+    return "\n\n".join(options)
 
 demo = gr.Interface(
-    fn=translate_text,
+    fn=translate_expression,
     inputs="text",
     outputs="text",
-    title="English → Brazilian Portuguese Translator",
-    description="Provides at least 5 PT-BR equivalents (MT + Linguee + Dicio + Synonyms)."
+    title="English → Brazilian Portuguese Translator (Hybrid)",
+    description="Gives literal + idiomatic PT-BR equivalents and explanations."
 )
 
 if __name__ == "__main__":