<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:dc="http://purl.org/dc/elements/1.1/">
  <channel>
    <title>DEV Community: Aditya Borgaonkar</title>
    <description>The latest articles on DEV Community by Aditya Borgaonkar (@aditya_borgaonkar).</description>
    <link>https://dev.to/aditya_borgaonkar</link>
    <image>
      <url>https://media2.dev.to/dynamic/image/width=90,height=90,fit=cover,gravity=auto,format=auto/https:%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F4055963%2F7b4eaa48-4359-46e0-8806-71ed29d8de7f.png</url>
      <title>DEV Community: Aditya Borgaonkar</title>
      <link>https://dev.to/aditya_borgaonkar</link>
    </image>
    <atom:link rel="self" type="application/rss+xml" href="https://dev.to/feed/aditya_borgaonkar"/>
    <language>en</language>
    <item>
      <title>I built an AI planetarium by adding a plugin to Stellarium</title>
      <dc:creator>Aditya Borgaonkar</dc:creator>
      <pubDate>Mon, 09 Nov 2026 21:02:33 +0000</pubDate>
      <link>https://dev.to/aditya_borgaonkar/i-built-an-ai-planetarium-by-adding-a-plugin-to-stellarium-3n3a</link>
      <guid>https://dev.to/aditya_borgaonkar/i-built-an-ai-planetarium-by-adding-a-plugin-to-stellarium-3n3a</guid>
      <description>&lt;p&gt;I wanted to say "show me Mars" to a planetarium and have it fly there. So I built &lt;strong&gt;Nebula&lt;/strong&gt;: an AI assistant plugin for &lt;a href="https://stellarium.org" rel="noopener noreferrer"&gt;Stellarium&lt;/a&gt;, the open-source planetarium.&lt;/p&gt;

&lt;h2&gt;
  
  
  What it does
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Ask the sky:&lt;/strong&gt; "show me Vega", "what does the sky look like at 9 pm?" and the view, time and zoom change.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Guided tours:&lt;/strong&gt; one click plans a five-stop tour of what's up now, with narration.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Photo identification:&lt;/strong&gt; upload a sky photo and get the objects named.&lt;/li&gt;
&lt;li&gt;Voice mode .&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;Demo, downloads (Windows and Linux) and source: &lt;a href="https://nebula.aitklabs.in" rel="noopener noreferrer"&gt;https://nebula.aitklabs.in&lt;/a&gt;&lt;/p&gt;

&lt;h2&gt;
  
  
  How it works
&lt;/h2&gt;

&lt;p&gt;Nebula is a Stellarium plugin written in C++/Qt, copied from the &lt;code&gt;HelloStelModule&lt;/code&gt; template. It adds a chat window and talks to any OpenAI-compatible endpoint(or Anthropic).&lt;/p&gt;

&lt;p&gt;The key design choice: &lt;strong&gt;the model never touches Stellarium directly.&lt;/strong&gt; I prompt it to reply only with JSON like this:&lt;/p&gt;



&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight plaintext"&gt;&lt;code&gt;
json
{"say": "Vega is about 25 light-years away.",
 "actions": [{"type": "goto", "name": "Vega"},
             {"type": "zoom", "fov": 20}]}
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;

</description>
      <category>showdev</category>
      <category>opensource</category>
      <category>ai</category>
      <category>cpp</category>
    </item>
    <item>
      <title>I got tired of managing AI API keys, so I built Yantra AI</title>
      <dc:creator>Aditya Borgaonkar</dc:creator>
      <pubDate>Fri, 31 Jul 2026 07:02:33 +0000</pubDate>
      <link>https://dev.to/aditya_borgaonkar/i-got-tired-of-managing-ai-api-keys-so-i-built-yantra-ai-40b0</link>
      <guid>https://dev.to/aditya_borgaonkar/i-got-tired-of-managing-ai-api-keys-so-i-built-yantra-ai-40b0</guid>
      <description>&lt;p&gt;&lt;a href="https://apikeys.in" rel="noopener noreferrer"&gt;&lt;/a&gt;&lt;br&gt;
Every time I wanted to experiment with a new LLM, the process looked like this:&lt;br&gt;
Create another account&lt;br&gt;
Add another billing method&lt;br&gt;
Generate another API key&lt;br&gt;
Learn another pricing model&lt;br&gt;
Change my code&lt;br&gt;
Hope the provider doesn't go down&lt;br&gt;
It wasn't hard—it was just repetitive.&lt;br&gt;
As an indie developer, I wanted to spend time building products, not managing AI providers.&lt;br&gt;
So I built Yantra AI.&lt;br&gt;
What is Yantra AI?&lt;br&gt;
Yantra AI is an AI routing gateway that lets you access multiple LLM providers through a single API key.&lt;br&gt;
Instead of managing separate integrations for OpenAI, Anthropic, Gemini, DeepSeek, Mistral, Grok, Qwen, and others, you integrate once.&lt;br&gt;
Your App&lt;br&gt;
      │&lt;br&gt;
      ▼&lt;br&gt;
 Yantra AI&lt;br&gt;
      │&lt;br&gt;
      ├── OpenAI&lt;br&gt;
      ├── Anthropic&lt;br&gt;
      ├── Gemini&lt;br&gt;
      ├── DeepSeek&lt;br&gt;
      ├── Mistral&lt;br&gt;
      └── ...&lt;br&gt;
No SDK rewrites.&lt;br&gt;
No juggling API keys.&lt;br&gt;
No separate billing dashboards.&lt;br&gt;
Why I built it&lt;br&gt;
Three things kept bothering me:&lt;/p&gt;

&lt;ol&gt;
&lt;li&gt;Too many API keys
Every provider has its own dashboard, billing system, and authentication.
Managing them becomes painful as your app grows.&lt;/li&gt;
&lt;li&gt;Vendor lock-in
Switching from GPT to Claude shouldn't require rewriting your backend.
It should be as simple as changing the model name.&lt;/li&gt;
&lt;li&gt;Reliability
If one provider has an outage or rate limits your requests, your application shouldn't stop working.
A routing layer makes failover much easier.
Features
One API key for multiple AI providers
OpenAI-compatible API
Smart model routing
Automatic failover
UPI billing (India)
Unified dashboard
Simple pricing
Example
If you're already using the OpenAI SDK, switching is minimal.
from openai import OpenAI&lt;/li&gt;
&lt;/ol&gt;

&lt;p&gt;client = OpenAI(&lt;br&gt;
    base_url="&lt;a href="https://api.apikeys.in/v1" rel="noopener noreferrer"&gt;https://api.apikeys.in/v1&lt;/a&gt;",&lt;br&gt;
    api_key="YOUR_API_KEY"&lt;br&gt;
)&lt;/p&gt;

&lt;p&gt;response = client.chat.completions.create(&lt;br&gt;
    model="claude-sonnet",&lt;br&gt;
    messages=[&lt;br&gt;
        {"role": "user", "content": "Hello!"}&lt;br&gt;
    ]&lt;br&gt;
)&lt;br&gt;
That's it.&lt;br&gt;
Lessons from building it&lt;br&gt;
Building the routing engine was only half the challenge.&lt;br&gt;
The harder part was creating a developer experience that feels invisible.&lt;br&gt;
The best infrastructure is the one developers don't have to think about.&lt;br&gt;
What's next?&lt;br&gt;
I'm working on:&lt;br&gt;
More model providers&lt;br&gt;
Better routing strategies&lt;br&gt;
Analytics&lt;br&gt;
Team workspaces&lt;br&gt;
SDKs&lt;br&gt;
Streaming improvements&lt;br&gt;
I'd love your feedback&lt;br&gt;
If you're building with AI APIs, I'd love to know:&lt;br&gt;
What's the biggest pain point today?&lt;br&gt;
Which providers do you use most?&lt;br&gt;
What would make a routing gateway genuinely useful for you?&lt;br&gt;
I'm always open to suggestions and feature requests.&lt;br&gt;
Happy building! 🚀&lt;/p&gt;

</description>
      <category>ai</category>
      <category>webdev</category>
      <category>programming</category>
    </item>
  </channel>
</rss>
