Create app.py

app.py

@@ -0,0 +1,180 @@
+#
+# SPDX-FileCopyrightText: Hadad <[email protected]>
+# SPDX-License-Identifier: Apache-2.0
+#
+
+import os
+from ollama import AsyncClient
+import gradio as gr
+
+async def playground(
+    message,
+    history,
+    num_ctx,
+    temperature,
+    repeat_penalty,
+    min_p,
+    top_k,
+    top_p
+):
+    if not isinstance(message, str) or not message.strip():
+        yield []
+        return
+
+    client = AsyncClient(
+        host=os.getenv("OLLAMA_API_BASE_URL"),
+        headers={
+            "Authorization": f"Bearer {os.getenv('OLLAMA_API_KEY')}"
+        }
+    )
+
+    messages = []
+    for item in history:
+        if isinstance(item, dict) and "role" in item and "content" in item:
+            messages.append({
+                "role": item["role"], 
+                "content": item["content"]
+            })
+    messages.append({"role": "user", "content": message})
+
+    response = ""
+    async for part in await client.chat(
+        model="gemma3:270m",
+        messages=messages,
+        options={
+            "num_ctx": int(num_ctx),
+            "temperature": float(temperature),
+            "repeat_penalty": float(repeat_penalty),
+            "min_p": float(min_p),
+            "top_k": int(top_k),
+            "top_p": float(top_p)
+        },
+        stream=True
+    ):
+        response += part.get("message", {}).get("content", "")
+        yield response
+
+with gr.Blocks(
+    fill_height=True,
+    fill_width=True
+) as app:
+    with gr.Sidebar():
+        gr.Markdown("## Ollama Playground by UltimaX Intelligence")
+        gr.HTML(
+            """
+            This space run the <b><a href=
+            "https://huggingface.co/google/gemma-3-270m" 
+            target="_blank">Gemma 3 (270M)</a></b> model from 
+            <b>Google</b>, hosted on a server using <b>Ollama</b> and 
+            accessed via the <b>Ollama Python SDK</b>.<br><br>
+
+            Official <b>documentation</b> for using Ollama with the 
+            Python SDK can be found 
+            <b><a href="https://github.com/ollama/ollama-python" 
+            target="_blank">here</a></b>.<br><br>
+
+            Gemma 3 (270M) runs entirely on <b>CPU</b>, utilizing only a 
+            <b>single core</b>. Thanks to its small size, the model can 
+            operate efficiently on minimal hardware.<br><br>
+
+            The Gemma 3 (270M) model can also be viewed or downloaded 
+            from the official Ollama website 
+            <b><a href="https://ollama.com/library/gemma3:270m" 
+            target="_blank">here</a></b>.<br><br>
+
+            While Gemma 3 has multimodal capabilities, running it on CPU 
+            with a relatively small number of parameters may limit its 
+            contextual understanding. For this reason, the upload 
+            functionality has been disabled.<br><br>
+
+            <b>Like this project? You can support me by buying a 
+            <a href="https://ko-fi.com/hadad" target="_blank">
+            coffee</a></b>.
+            """
+        )
+        gr.Markdown("---")
+        gr.Markdown("## Model Parameters")
+        num_ctx = gr.Slider(
+            minimum=512,
+            maximum=1024,
+            value=512,
+            step=128,
+            label="Context Length (num_ctx)",
+            info="Maximum context window size. Limited to CPU usage."
+        )
+        gr.Markdown("")
+        temperature = gr.Slider(
+            minimum=0.1,
+            maximum=2.0,
+            value=1.0,
+            step=0.1,
+            label="Temperature",
+            info="Controls randomness in generation"
+        )
+        gr.Markdown("")
+        repeat_penalty = gr.Slider(
+            minimum=0.1,
+            maximum=2.0,
+            value=1.0,
+            step=0.1,
+            label="Repeat Penalty",
+            info="Penalty for repeating tokens"
+        )
+        gr.Markdown("")
+        min_p = gr.Slider(
+            minimum=0.0,
+            maximum=1.0,
+            value=0.001,
+            step=0.001,
+            label="Min P",
+            info="Minimum probability threshold"
+        )
+        gr.Markdown("")
+        top_k = gr.Slider(
+            minimum=0,
+            maximum=100,
+            value=64,
+            step=1,
+            label="Top K",
+            info="Number of top tokens to consider"
+        )
+        gr.Markdown("")
+        top_p = gr.Slider(
+            minimum=0.0,
+            maximum=1.0,
+            value=0.95,
+            step=0.05,
+            label="Top P",
+            info="Cumulative probability threshold"
+        )
+    gr.ChatInterface(
+        fn=playground,
+        additional_inputs=[
+            num_ctx,
+            temperature,
+            repeat_penalty,
+            min_p,
+            top_k,
+            top_p
+        ],
+        chatbot=gr.Chatbot(
+            label="Ollama | Gemma 3 (270M)",
+            type="messages",
+            show_copy_button=True,
+            scale=1
+        ),
+        type="messages",
+        examples=[
+            ["Please introduce yourself."],
+            ["What caused World War II?"],
+            ["Give me a short introduction to large language model."],
+            ["Explain about quantum computers."]
+        ],
+        cache_examples=False,
+        show_api=False
+    )
+
+app.launch(
+    server_name="0.0.0.0",
+    pwa=True
+)