Skip to content

Commit 9d7534a

Browse files
nicholasdbradynicholasdbrady
committed
refactor: remove outdated content and update links in blog and README
1 parent a690206 commit 9d7534a

File tree

4 files changed

+20
-197
lines changed

4 files changed

+20
-197
lines changed

‎README.md‎

Lines changed: 4 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -16,11 +16,13 @@ Inside of this project, you'll find notebooks in my `examples` folder:
1616

1717
```text
1818
├── examples/
19-
│   ├── phi-3/
20-
| |── getting-started-with-phi-3.ipynb
19+
│   ├── deepseek/
20+
| |── deepseek-r1-with-azure-ai-foundry-and-gradio.ipynb
2121
| ├── maas-arena/
2222
| |── azure-ai-models-as-a-service-chatbot-arena.ipynb
2323
| |── maas_clients.py
24+
│   ├── phi-3/
25+
| |── getting-started-with-phi-3.ipynb
2426
| |── **more coming soon!**
2527
```
2628

‎examples/deepseek/deepseek-r1-with-azure-ai-foundry-and-gradio.ipynb‎

Lines changed: 15 additions & 190 deletions
Original file line numberDiff line numberDiff line change
@@ -8,12 +8,22 @@
88
"\n",
99
"## Introduction\n",
1010
"\n",
11-
"**DeepSeek R1** has gained widespread attention for its advanced reasoning capabilities, excelling in language processing, scientific problem-solving, and coding. With 671B total parameters, 37B active parameters, and a 128K context length, it pushes the boundaries of AI-driven reasoning ([Explore DeepSeek R1 on Azure AI Foundry](https://ai.azure.com/explore/models/DeepSeek-R1/version/1/registry/azureml-deepseek)). Benchmarking and evaluation results highlight its performance against other models, showcasing its effectiveness in reasoning tasks ([Evaluation Results](https://github.com/deepseek-ai/DeepSeek-R1/tree/main?tab=readme-ov-file#4-evaluation-results)). Building on prior models, DeepSeek R1 integrates Chain-of-Thought (CoT) reasoning, reinforcement learning (RL), and fine-tuning on curated datasets to achieve state-of-the-art performance. This tutorial will walk you through how to deploy DeepSeek R1 from [Azure AI Foundry's model catalog](https://ai.azure.com/explore/models/) and integrate it with [Gradio](https://www.gradio.app/) to build a real-time streaming chatbot specifically for thinking LLMs like **DeepSeek R1**.\n",
12-
"\n",
11+
"**DeepSeek R1** has gained widespread attention for its advanced reasoning capabilities, excelling in language processing, scientific problem-solving, and coding. With 671B total parameters, 37B active parameters, and a 128K context length, it pushes the boundaries of AI-driven reasoning ([Explore DeepSeek R1 on Azure AI Foundry](https://ai.azure.com/explore/models/DeepSeek-R1/version/1/registry/azureml-deepseek)). Benchmarking and evaluation results highlight its performance against other models, showcasing its effectiveness in reasoning tasks ([Evaluation Results](https://github.com/deepseek-ai/DeepSeek-R1/tree/main?tab=readme-ov-file#4-evaluation-results)). Building on prior models, DeepSeek R1 integrates Chain-of-Thought (CoT) reasoning, reinforcement learning (RL), and fine-tuning on curated datasets to achieve state-of-the-art performance. This tutorial will walk you through how to deploy DeepSeek R1 from [Azure AI Foundry's model catalog](https://ai.azure.com/explore/models/) and integrate it with [Gradio](https://www.gradio.app/) to build a real-time streaming chatbot specifically for thinking LLMs like **DeepSeek R1**."
12+
]
13+
},
14+
{
15+
"cell_type": "markdown",
16+
"metadata": {},
17+
"source": [
1318
"### DeepSeek R1 on Azure AI Foundry\n",
1419
"\n",
15-
"On **January 29, 2025**, Microsoft announced that **DeepSeek R1** is now available on **Azure AI Foundry** and **GitHub**, making it part of a growing portfolio of over **1,800 AI models** available for enterprise use. With this integration, businesses can deploy DeepSeek R1 using **serverless APIs**, ensuring seamless scalability, security, and compliance with Microsoft’s responsible AI principles. ([Azure AI Foundry announcement](https://azure.microsoft.com/en-us/blog/deepseek-r1-on-azure-ai-foundry))\n",
16-
"\n",
20+
"On **January 29, 2025**, Microsoft announced that **DeepSeek R1** is now available on **Azure AI Foundry** and **GitHub**, making it part of a growing portfolio of over **1,800 AI models** available for enterprise use. With this integration, businesses can deploy DeepSeek R1 using **serverless APIs**, ensuring seamless scalability, security, and compliance with Microsoft’s responsible AI principles. ([Azure AI Foundry announcement](https://azure.microsoft.com/en-us/blog/deepseek-r1-on-azure-ai-foundry))"
21+
]
22+
},
23+
{
24+
"cell_type": "markdown",
25+
"metadata": {},
26+
"source": [
1727
"### Benefits of Using DeepSeek R1 on Azure AI Foundry\n",
1828
"\n",
1929
"- **Enterprise-Ready AI:** DeepSeek R1 is available as a trusted, scalable, and secure AI model, backed by Microsoft's infrastructure.\n",
@@ -203,195 +213,10 @@
203213
"- Implementing real-time streaming responses.\n",
204214
"- Deploying the chatbot using Gradio.\n",
205215
"\n",
206-
"Get started today by visiting **[Azure AI Foundry](https://azure.microsoft.com/en-us/products/ai-services/ai-foundry)** and **[DeepSeek on GitHub](https://github.com/DeepSeekAI/DeepSeek-R1)**.\n",
216+
"Get started today by visiting **[DeepSeek R1 on Azure AI Foundry Model Catalog](https://ai.azure.com/explore/models/DeepSeek-R1/version/1/registry/azureml-deepseek)** and **[DeepSeek on GitHub Models](https://github.com/marketplace/models/azureml-deepseek/DeepSeek-R1)**.\n",
207217
"\n",
208218
"Happy coding! 🚀"
209219
]
210-
},
211-
{
212-
"cell_type": "code",
213-
"execution_count": null,
214-
"metadata": {},
215-
"outputs": [
216-
{
217-
"name": "stdout",
218-
"output_type": "stream",
219-
"text": [
220-
"* Running on local URL: http://127.0.0.1:7910\n",
221-
"\n",
222-
"To create a public link, set `share=True` in `launch()`.\n"
223-
]
224-
},
225-
{
226-
"data": {
227-
"text/html": [
228-
"<div><iframe src=\"http://127.0.0.1:7910/\" width=\"100%\" height=\"500\" allow=\"autoplay; camera; microphone; clipboard-read; clipboard-write;\" frameborder=\"0\" allowfullscreen></iframe></div>"
229-
],
230-
"text/plain": [
231-
"<IPython.core.display.HTML object>"
232-
]
233-
},
234-
"metadata": {},
235-
"output_type": "display_data"
236-
},
237-
{
238-
"name": "stdout",
239-
"output_type": "stream",
240-
"text": [
241-
"Gradio ChatMessage: {'role': 'user', 'metadata': {}, 'content': 'hey there', 'options': []}\n",
242-
"Converted to Azure Message: {'role': 'user', 'content': 'hey there'}\n",
243-
"Final Azure Messages: [{'role': 'system', 'content': 'You are a helpful assistant.'}, {'role': 'user', 'content': 'hey there'}]\n",
244-
"Using parameters - Temperature: 0.7, Top P: 1, Max Tokens: 2048\n",
245-
"Entering thought processing mode.\n"
246-
]
247-
}
248-
],
249-
"source": [
250-
"import os\n",
251-
"import gradio as gr\n",
252-
"from azure.ai.inference import ChatCompletionsClient\n",
253-
"from azure.ai.inference.models import SystemMessage, UserMessage, AssistantMessage\n",
254-
"from azure.core.credentials import AzureKeyCredential\n",
255-
"from gradio import ChatMessage\n",
256-
"from typing import Iterator\n",
257-
"\n",
258-
"###############################################################################\n",
259-
"# 1) Create the ChatCompletionsClient\n",
260-
"###############################################################################\n",
261-
"client = ChatCompletionsClient(\n",
262-
" endpoint=os.environ[\"AZURE_INFERENCE_ENDPOINT\"], # e.g. \"https://my-r1-endpoint.eastus2.inference.ai.azure.com\"\n",
263-
" credential=AzureKeyCredential(os.environ[\"AZURE_INFERENCE_CREDENTIAL\"])\n",
264-
" # If you're authenticating with Microsoft Entra ID, use DefaultAzureCredential() \n",
265-
" # or other supported credentials instead of AzureKeyCredential.\n",
266-
")\n",
267-
"\n",
268-
"###############################################################################\n",
269-
"# 2) Stream response function for Gradio\n",
270-
"###############################################################################\n",
271-
"def stream_response(user_message: str, messages: list, temperature: float, top_p: float, max_tokens: int) -> Iterator[list]:\n",
272-
" if not messages:\n",
273-
" messages = []\n",
274-
" \n",
275-
" # Convert Gradio chat history into Azure AI Inference messages\n",
276-
" azure_messages = [SystemMessage(content=\"You are a helpful assistant.\")]\n",
277-
" for msg in messages:\n",
278-
" print(f\"Gradio ChatMessage: {msg}\") # Debug print\n",
279-
" if isinstance(msg, ChatMessage):\n",
280-
" azure_msg = UserMessage(content=msg.content) if msg.role == \"user\" else AssistantMessage(content=msg.content)\n",
281-
" elif isinstance(msg, dict) and \"role\" in msg and \"content\" in msg:\n",
282-
" azure_msg = UserMessage(content=msg[\"content\"]) if msg[\"role\"] == \"user\" else AssistantMessage(content=msg[\"content\"])\n",
283-
" else:\n",
284-
" continue\n",
285-
" print(f\"Converted to Azure Message: {azure_msg}\") # Debug print\n",
286-
" azure_messages.append(azure_msg)\n",
287-
" \n",
288-
" # Ensure only serializable objects are sent to Azure\n",
289-
" azure_messages = [msg.dict() if hasattr(msg, \"dict\") else msg for msg in azure_messages]\n",
290-
" \n",
291-
" print(f\"Final Azure Messages: {azure_messages}\") # Debug print\n",
292-
" print(f\"Using parameters - Temperature: {temperature}, Top P: {top_p}, Max Tokens: {max_tokens}\")\n",
293-
" \n",
294-
" response = client.complete(messages=azure_messages, stream=True, temperature=temperature, top_p=top_p, max_tokens=max_tokens)\n",
295-
" \n",
296-
" # Initialize buffers\n",
297-
" thought_buffer = \"\"\n",
298-
" response_buffer = \"\"\n",
299-
" inside_thought = False\n",
300-
" \n",
301-
" for update in response:\n",
302-
" if update.choices:\n",
303-
" current_chunk = update.choices[0].delta.content\n",
304-
" \n",
305-
" if \"<think>\" in current_chunk:\n",
306-
" inside_thought = True\n",
307-
" print(\"Entering thought processing mode.\")\n",
308-
" messages.append(ChatMessage(role=\"assistant\", content=\"\", metadata={\"title\": \"🧠 R1 Thinking...\", \"status\": \"pending\"}))\n",
309-
" yield messages\n",
310-
" continue\n",
311-
" elif \"</think>\" in current_chunk:\n",
312-
" inside_thought = False\n",
313-
" messages[-1] = ChatMessage(\n",
314-
" role=\"assistant\",\n",
315-
" content=thought_buffer.strip(),\n",
316-
" metadata={\"title\": \"🧠 R1 Thinking...\", \"status\": \"done\"}\n",
317-
" )\n",
318-
" yield messages # Yield the thought message immediately\n",
319-
" thought_buffer = \"\"\n",
320-
" continue\n",
321-
" \n",
322-
" if inside_thought:\n",
323-
" thought_buffer += current_chunk\n",
324-
" messages[-1] = ChatMessage(\n",
325-
" role=\"assistant\",\n",
326-
" content=thought_buffer,\n",
327-
" metadata={\"title\": \"🧠 R1 Thinking...\", \"status\": \"pending\"}\n",
328-
" )\n",
329-
" yield messages # Yield the thought message as it updates\n",
330-
" else:\n",
331-
" response_buffer += current_chunk\n",
332-
" if messages and isinstance(messages[-1], ChatMessage) and messages[-1].role == \"assistant\" and (not messages[-1].metadata or \"title\" not in messages[-1].metadata):\n",
333-
" messages[-1] = ChatMessage(role=\"assistant\", content=response_buffer)\n",
334-
" else:\n",
335-
" messages.append(ChatMessage(role=\"assistant\", content=response_buffer))\n",
336-
" yield messages\n",
337-
"\n",
338-
"###############################################################################\n",
339-
"# 3) Gradio UI\n",
340-
"###############################################################################\n",
341-
"brand_theme = gr.themes.Default(\n",
342-
" primary_hue=\"blue\",\n",
343-
" secondary_hue=\"blue\",\n",
344-
" neutral_hue=\"gray\",\n",
345-
" font=[\"Segoe UI\", \"Arial\", \"sans-serif\"],\n",
346-
" font_mono=[\"Courier New\", \"monospace\"]\n",
347-
").set(\n",
348-
" button_primary_background_fill=\"#0f6cbd\",\n",
349-
" button_primary_background_fill_hover=\"#115ea3\",\n",
350-
" button_primary_background_fill_hover_dark=\"#4f52b2\",\n",
351-
" button_primary_background_fill_dark=\"#5b5fc7\",\n",
352-
" button_primary_text_color=\"#ffffff\",\n",
353-
" body_background_fill=\"#f5f5f5\",\n",
354-
" block_background_fill=\"#ffffff\",\n",
355-
" body_text_color=\"#242424\",\n",
356-
" body_text_color_subdued=\"#616161\",\n",
357-
" block_border_color=\"#d1d1d1\",\n",
358-
" input_background_fill=\"#ffffff\",\n",
359-
" input_border_color=\"#d1d1d1\",\n",
360-
" input_border_color_focus=\"#0f6cbd\",\n",
361-
")\n",
362-
"\n",
363-
"with gr.Blocks(title=\"DeepSeek R1 with Azure AI Foundry\", theme=brand_theme, css=\"footer {visibility: hidden}\", fill_height=True, fill_width=True) as demo:\n",
364-
" title = gr.Markdown(\"## DeepSeek R1 with Azure AI Foundry 🤭\")\n",
365-
" chatbot = gr.Chatbot(\n",
366-
" type=\"messages\",\n",
367-
" label=\"DeepSeek-R1\",\n",
368-
" render_markdown=True,\n",
369-
" show_label=False,\n",
370-
" scale=1,\n",
371-
" )\n",
372-
" \n",
373-
" input_box = gr.Textbox(\n",
374-
" lines=1,\n",
375-
" submit_btn=True,\n",
376-
" show_label=False,\n",
377-
" )\n",
378-
" \n",
379-
" with gr.Accordion(\"Model Parameters\", open=False):\n",
380-
" temperature = gr.Slider(0.0, 2.0, value=0.7, step=0.1, label=\"Temperature\", info=\"Controls randomness\", interactive=True)\n",
381-
" top_p = gr.Slider(0.0, 1.0, value=1.0, step=0.1, label=\"Top P\", info=\"Nucleus sampling\", interactive=True)\n",
382-
" max_tokens = gr.Slider(0, 4096, value=2048, step=128, label=\"Max Tokens\", info=\"Limits response length\", interactive=True)\n",
383-
" reset_button = gr.Button(\"Reset Defaults\")\n",
384-
" \n",
385-
" reset_button.click(lambda: (0.7, 1.0, 2048), outputs=[temperature, top_p, max_tokens])\n",
386-
" \n",
387-
" msg_store = gr.State(\"\")\n",
388-
" input_box.submit(lambda msg: (msg, msg, \"\"), inputs=[input_box], outputs=[msg_store, input_box, input_box], queue=False)\n",
389-
" input_box.submit(lambda msg, chat: (ChatMessage(role=\"user\", content=msg), chat + [ChatMessage(role=\"user\", content=msg)]), inputs=[msg_store, chatbot], outputs=[msg_store, chatbot], queue=False).then(\n",
390-
" stream_response, inputs=[msg_store, chatbot, temperature, top_p, max_tokens], outputs=chatbot\n",
391-
" )\n",
392-
" \n",
393-
" demo.launch()\n"
394-
]
395220
}
396221
],
397222
"metadata": {

‎src/content/blog/create-a-thinking-llm-chat-interface-with-azure-ai-foundry-deepseek-and-gradio.md‎

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -195,6 +195,6 @@ We covered:
195195
- Implementing real-time streaming responses.
196196
- Deploying the chatbot using Gradio.
197197

198-
Get started today by visiting **[Azure AI Foundry](https://azure.microsoft.com/en-us/products/ai-services/ai-foundry)** and **[DeepSeek on GitHub](https://github.com/DeepSeekAI/DeepSeek-R1)**.
198+
Get started today by visiting **[DeepSeek R1 Azure AI Foundry Model Catalog](https://ai.azure.com/explore/models/DeepSeek-R1/version/1/registry/azureml-deepseek)** or **[DeepSeek on GitHub Models](https://github.com/marketplace/models/azureml-deepseek/DeepSeek-R1)**.
199199

200200
Happy coding! 🚀

‎src/pages/index.astro‎

Lines changed: 0 additions & 4 deletions
Original file line numberDiff line numberDiff line change
@@ -19,10 +19,6 @@ import { SITE_TITLE, SITE_DESCRIPTION } from "../consts";
1919
plenty of recipes for you to follow and learn from. I hope you
2020
enjoy the content and find it useful.
2121
</p>
22-
<p>
23-
Pardon the dust, I'm just getting started. I'll be adding more
24-
content soon.
25-
</p>
2622
<p>
2723
In the meantime, check out my <a href="/cookbook/blog/">blog</a> page
2824
and please connect with me:

0 commit comments

Comments
 (0)