Upload 2 files

app.py

@@ -0,0 +1,43 @@
+import os
+import gradio as gr
+import torch
+from transformers import AutoTokenizer, AutoModelForCausalLM
+
+MODEL_ID = os.getenv("MODEL_ID", "mistralai/Mistral-7B-Instruct-v0.3")
+HF_TOKEN = os.getenv("HF_TOKEN", None)
+
+tokenizer = AutoTokenizer.from_pretrained(MODEL_ID, token=HF_TOKEN)
+model = AutoModelForCausalLM.from_pretrained(
+    MODEL_ID,
+    token=HF_TOKEN,
+    torch_dtype="auto",
+    device_map="auto",
+)
+
+def draft_reply(subject, thread):
+    system = (
+        "You are an email assistant. Draft a reply email.\n"
+        "- Be clear and polite.\n"
+        "- Ask up to 2 clarifying questions if needed.\n"
+        "- Do not invent facts.\n"
+        "- Output ONLY the email body.\n"
+    )
+    user = f"Subject: {subject}\n\nEmail thread:\n{thread}\n\nWrite the reply now."
+
+    messages = [{"role": "system", "content": system}, {"role": "user", "content": user}]
+    prompt = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
+    inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
+
+    with torch.no_grad():
+        outputs = model.generate(**inputs, max_new_tokens=250, temperature=0.4, do_sample=True)
+
+    return tokenizer.decode(outputs[0][inputs["input_ids"].shape[1]:], skip_special_tokens=True).strip()
+
+demo = gr.Interface(
+    fn=draft_reply,
+    inputs=[gr.Textbox(label="Subject"), gr.Textbox(label="Email Thread", lines=10)],
+    outputs=gr.Textbox(label="Draft Reply", lines=12),
+    title="Email Reply Drafting Assistant",
+)
+
+demo.launch()

requirements.txt

@@ -0,0 +1,15 @@
+# torch
+# transformers>=4.45.0
+# accelerate>=0.33.0
+# safetensors>=0.4.3
+# huggingface_hub>=0.24.0
+# sentencepiece>=0.2.0
+# python-dotenv>=1.0.1
+
+gradio
+torch
+transformers>=4.45.0
+accelerate>=0.33.0
+safetensors>=0.4.3
+huggingface_hub>=0.24.0
+sentencepiece>=0.2.0