set zephyr model to be the main

app.py

@@ -29,58 +29,16 @@ ROLE_PRESETS = {
     ),
 }
 
-# ---------------- Examples per role ----------------
-ROLE_EXAMPLES = {
-    "Friendly Chatbot": [
-        "Explain embeddings in one paragraph.",
-        "What's a neat productivity trick for engineers?"
-    ],
-    "Dataset Auditor": [
-        "10k images (cats/dogs/birds). First-pass audit plan?",
-        "Class imbalance & duplicates: quick remediation checklist."
-    ],
-    "SQL Explainer": [
-        "Explain for a PM:\nSELECT u.name, COUNT(*) c FROM orders o JOIN users u ON u.id=o.user_id GROUP BY u.name HAVING COUNT(*)>5;",
-        "What does this CTE do and why use it?"
-    ],
-    "Code Reviewer": [
-        "Review Python for edge cases:\n\ndef top_k(nums, k):\n    return sorted(nums)[-k:]",
-        "Find race conditions in this multi-threaded write snippet."
-    ],
-    "Data Pipeline Doctor": [
-        "ETL fails randomly after 10k rows out of 2M. Where to look first?",
-        "Parquet load spikes memory; pragmatic fixes?"
-    ],
-    "Data Engineering Advisor": [
-        "Design a batch + streaming pipeline for clickstream analytics.",
-        "What’s the best way to partition a large parquet dataset for Athena?"
-    ],
-    "ML Dataset Preparer": [
-        "Steps to prepare a face recognition dataset from raw videos.",
-        "Best augmentation methods for small medical image datasets?"
-    ],
-    "Data Quality Analyst": [
-        "Detect anomalies in tabular data with numerical and categorical columns.",
-        "Checklist for ensuring schema consistency across multiple CSV files."
-    ],
-}
-
-# ---------------- Models ----------------
-DEFAULT_MODELS = [
-    "HuggingFaceH4/zephyr-7b-beta",
-    "meta-llama/Meta-Llama-3-8B-Instruct",
-    "mistralai/Mistral-7B-Instruct-v0.3",
-    "Qwen/Qwen2.5-7B-Instruct",
-    "google/gemma-2-9b-it",
-]
+# ---------------- Model ----------------
+MODEL_NAME = "HuggingFaceH4/zephyr-7b-beta"
 
 # ---------------- Client factory ----------------
-def get_client(model_name: str):
+def get_client():
     token = os.getenv("HF_TOKEN")  # optional
-    return InferenceClient(model=model_name, token=token)
+    return InferenceClient(model=MODEL_NAME, token=token)
 
 # ---------------- Chat backend (streaming) ----------------
-def stream_reply(message, history, role, system_message, model_name, max_tokens, temperature, top_p):
+def stream_reply(message, history, role, system_message):
     sys_msg = (system_message or "").strip() or ROLE_PRESETS.get(role, "")
     messages = [{"role": "system", "content": sys_msg}]
     for u, a in history:
@@ -90,16 +48,10 @@ def stream_reply(message, history, role, system_message, model_name, max_tokens,
             messages.append({"role": "assistant", "content": a})
     messages.append({"role": "user", "content": message})
 
-    client = get_client(model_name)
+    client = get_client()
     partial = ""
     try:
-        for event in client.chat_completion(
-            messages=messages,
-            stream=True,
-            max_tokens=max_tokens,
-            temperature=temperature,
-            top_p=top_p,
-        ):
+        for event in client.chat_completion(messages=messages, stream=True):
             delta = event.choices[0].delta.content or ""
             if delta:
                 partial += delta
@@ -108,23 +60,15 @@ def stream_reply(message, history, role, system_message, model_name, max_tokens,
         yield f"⚠️ Inference error: {e}"
 
 # ---------------- UI ----------------
-with gr.Blocks(title="HF Chat • Data Roles") as demo:
-    gr.Markdown("## 🤗 Hugging Face Chat (Data-focused Roles + Examples)")
+with gr.Blocks(title="HF Zephyr Chat • Data Roles") as demo:
+    gr.Markdown("## 🤗 Zephyr Chat (Data-focused Roles)")
 
-    with gr.Row():
-        role_dd = gr.Dropdown(
-            label="Role preset",
-            choices=list(ROLE_PRESETS.keys()),
-            value="Friendly Chatbot",
-            interactive=True,
-        )
-        model_dd = gr.Dropdown(
-            label="Model (type any HF model ID)",
-            choices=DEFAULT_MODELS,
-            value=DEFAULT_MODELS[0],
-            allow_custom_value=True,
-            interactive=True,
-        )
+    role_dd = gr.Dropdown(
+        label="Role preset",
+        choices=list(ROLE_PRESETS.keys()),
+        value="Friendly Chatbot",
+        interactive=True,
+    )
 
     system_tb = gr.Textbox(
         label="System message (auto-filled by role; you can edit)",
@@ -132,38 +76,14 @@ with gr.Blocks(title="HF Chat • Data Roles") as demo:
         lines=4,
     )
 
-    with gr.Accordion("Examples for selected role", open=False):
-        ex_radio = gr.Radio(
-            label="Pick an example to insert into the input",
-            choices=ROLE_EXAMPLES["Friendly Chatbot"],
-            interactive=True,
-        )
-        insert_btn = gr.Button("Insert example into input")
-
     chat = gr.ChatInterface(
         fn=stream_reply,
-        additional_inputs=[
-            role_dd,
-            system_tb,
-            model_dd,
-            gr.Slider(1, 2048, value=512, step=1, label="Max new tokens"),
-            gr.Slider(0.0, 2.0, value=0.7, step=0.1, label="Temperature"),
-            gr.Slider(0.0, 1.0, value=0.95, step=0.05, label="Top-p"),
-        ],
+        additional_inputs=[role_dd, system_tb],
     )
 
     def _on_role_change(role):
         return ROLE_PRESETS.get(role, "")
     role_dd.change(fn=_on_role_change, inputs=role_dd, outputs=system_tb)
 
-    def _examples_for_role(role):
-        items = ROLE_EXAMPLES.get(role, [])
-        return gr.update(choices=items, value=(items[0] if items else None))
-    role_dd.change(fn=_examples_for_role, inputs=role_dd, outputs=ex_radio)
-
-    def _insert_example(example_text):
-        return gr.update(value=example_text or "")
-    insert_btn.click(fn=_insert_example, inputs=ex_radio, outputs=chat.textbox)
-
 if __name__ == "__main__":
-    demo.launch()
+    demo.launch()