Deploy CGP Bot API with Gemini fallback

Dockerfile

@@ -0,0 +1,33 @@
+# Use a Python base image
+FROM python:3.9
+
+# Create a non-root user for security best practices
+RUN useradd -m -u 1000 user
+USER user
+
+# Set environment variable for user's local bin path
+ENV PATH="/home/user/.local/bin:$PATH"
+
+# Set environment variables to prevent Python from buffering stdout/stderr
+ENV PYTHONUNBUFFERED 1
+
+# IMPORTANT FIX: Set Hugging Face cache directory to a writable location
+ENV HF_HOME=/app/hf_cache
+
+# Set the working directory inside the container
+WORKDIR /app
+
+# Copy requirements.txt and install Python dependencies
+COPY --chown=user requirements.txt .
+RUN pip install --no-cache-dir --upgrade pip && \
+    pip install --no-cache-dir -r requirements.txt
+
+# Copy the application file
+# The app.py will download models from Hugging Face Hub, so no need to copy model dirs here.
+COPY --chown=user ./app.py /app/app.py
+COPY --chown=user ./cgpcorp_api_chat.py /app/cgpcorp_api_chat.py
+COPY --chown=user ./README.md /app/README.md
+
+# Command to run the application using uvicorn
+EXPOSE 7860
+CMD ["uvicorn", "app:app", "--host", "0.0.0.0", "--port", "7860"]

app.py

@@ -0,0 +1,18 @@
+from fastapi import FastAPI
+from pydantic import BaseModel
+from cgpcorp_api_chat import translate_text
+
+app = FastAPI(title="CGP Corp Chatbot API")
+
+class TranslationRequest(BaseModel):
+    text: str
+    direction: str  # "en-fr" or "fr-en"
+
+@app.get("/")
+def home():
+    return {"message": "✅ CGP Corp Bot API running on Hugging Face Spaces"}
+
+@app.post("/translate")
+def translate(req: TranslationRequest):
+    result = translate_text(req.text, req.direction)
+    return {"translated_text": result}

cgpcorp_api_chat.py

@@ -0,0 +1,45 @@
+import os
+import requests
+from transformers import AutoTokenizer, AutoModelForSeq2SeqLM, pipeline
+
+# Load Hugging Face models
+MODEL_EN_FR = "cgpcorpbot/cgp_model_en-fr"
+MODEL_FR_EN = "cgpcorpbot/cgp_model_fr-en"
+
+print("Loading EN->FR model...")
+tokenizer_en_fr = AutoTokenizer.from_pretrained(MODEL_EN_FR)
+model_en_fr = AutoModelForSeq2SeqLM.from_pretrained(MODEL_EN_FR)
+translator_en_fr = pipeline("translation", model=model_en_fr, tokenizer=tokenizer_en_fr)
+
+print("Loading FR->EN model...")
+tokenizer_fr_en = AutoTokenizer.from_pretrained(MODEL_FR_EN)
+model_fr_en = AutoModelForSeq2SeqLM.from_pretrained(MODEL_FR_EN)
+translator_fr_en = pipeline("translation", model=model_fr_en, tokenizer=tokenizer_fr_en)
+
+# Gemini API key (set as HF secret in the Space)
+GEMINI_API_KEY = os.getenv("GEMINI_API_KEY")
+
+def translate_text(text, direction="en-fr"):
+    """Translate using custom models, fallback to Gemini."""
+    try:
+        if direction == "en-fr":
+            return translator_en_fr(text, max_length=512)[0]['translation_text']
+        else:
+            return translator_fr_en(text, max_length=512)[0]['translation_text']
+    except Exception as e:
+        print("⚠️ Local model failed, falling back to Gemini:", e)
+        return gemini_translate(text, direction)
+
+def gemini_translate(text, direction="en-fr"):
+    """Fallback using Gemini API."""
+    if not GEMINI_API_KEY:
+        return "Gemini API key not configured."
+    target = "fr" if direction == "en-fr" else "en"
+    response = requests.post(
+        "https://api.gemini.com/translate",
+        headers={"Authorization": f"Bearer {GEMINI_API_KEY}"},
+        json={"q": text, "target": target}
+    )
+    if response.status_code == 200:
+        return response.json().get("translatedText", "Translation failed.")
+    return f"Gemini API Error: {response.text}"

requirements.txt

@@ -0,0 +1,7 @@
+fastapi
+uvicorn[standard]
+transformers
+torch
+requests
+python-dotenv
+datasets