The first version of the Alfe AI Cloud Platform https://alfe.sh (beta-2.30). This initial cloud release includes the image design component of the Alfe AI Platform. It now defaults to OpenAI's gpt-image-1 model for image generation via the built-in API. You can change the model globally via the new image_gen_model setting which accepts gptimage1, dalle2, or dalle3. If the model returns a base64 string instead of a URL, the server automatically decodes and saves the image. The server also includes an optional color swatch detector that can trim any palette band from the bottom of generated images. This feature is disabled by default and can be enabled via the remove_color_swatches setting. The software development component is coming soon, and is available now as a Pre-release on GitHub.

Alfe AI beta-2.30+ (Image Design): https://github.com/alfe-ai/alfe-ai-2.0_beta
Alfe AI beta-0.4x+ (Software Development): https://github.com/alfe-ai/Sterling

wget https://raw.githubusercontent.com/alfe-ai/alfe-ai-Aurelix/refs/heads/Aurora/Aurelix/dev/main-rel2/deploy_aurelix.sh && chmod +x deploy_aurelix.sh && ./deploy_aurelix.sh

Set HTTPS_KEY_PATH and HTTPS_CERT_PATH to the SSL key and certificate files to enable HTTPS across the included servers. If the files are missing the services fall back to HTTP.

You can quickly obtain free certificates from Let's Encrypt by running the setup_certbot.sh script. It installs Certbot and generates the key and certificate files for the domain you specify.

After obtaining the certificates, run setup_ssl_permissions.sh <domain> [user] to grant the specified user (default: admin) read access to the key and certificate so Aurora can run without root privileges.

The Aurora server reads its port from the AURORA_PORT environment variable (default: 3000). Binding directly to port 443 typically requires root privileges. If you prefer to run the server as a regular user, you can forward incoming connections from port 443 to your configured AURORA_PORT.

Run the helper script with sudo to set up the forwarding rule:

sudo ./forward_port_443.sh 3000

Replace 3000 with your chosen AURORA_PORT. After adding the rule, start the server normally and clients can connect using https://your-domain/ on port 443 while the Node.js process continues to run on the higher port.

Set SQL_SERVER_PORT in .env (see Aurora/.env.example) to configure the port for a simple HTTP interface to the SQLite database. Start the server with:

npm run sqlserver --prefix Aurora

Send POST requests to /sql with a JSON body containing a sql string and optional params array. Select queries return rows while other statements return change information.

A small command-line script perplexity-cli.js lets you query the official Perplexity API and prints any citation URLs from the response. A newer interactive version pplx.js is also provided.

Supported Perplexity models:

sonar
sonar-pro
sonar-reasoning
sonar-reasoning-pro
sonar-deep-research
r1-1776

npm install axios commander

Set your API key with PERPLEXITY_API_KEY or pass --key when running.

chmod +x pplx.js
./pplx.js "What causes aurora borealis?"

The script outputs the answer and lists cited URLs if available.

The CLI follows Perplexity's official API which mirrors the OpenAI Chat Completions format. Set your API key in the PERPLEXITY_API_KEY environment variable and send requests to https://api.perplexity.ai/chat/completions.

{
  "model": "sonar-pro",
  "messages": [
    { "role": "system", "content": "You are a helpful assistant." },
    { "role": "user", "content": "Who discovered penicillin?" }
  ],
  "max_tokens": 512,
  "temperature": 0.7
}

Responses include optional citations and search_results arrays with the source URLs used. Common errors are a 400 Bad Request when the model name is wrong and 401/403 for missing or invalid API keys. Use plain model names like sonar, sonar-pro, sonar-reasoning, sonar-reasoning-pro, sonar-deep-research, or r1-1776 without any provider prefix.

The order of models shown in the reasoning tooltip can be customized. Edit Aurora/public/reasoning_tooltip_config.js and reorder the chatModels and reasoningModels arrays to suit your preferences.

The chat model list now includes Anthropic Claude Sonnet 4 as an ultimate tier option. Pricing is $3 per million input tokens, $15 per million output tokens, and $4.80 per thousand input images.

Anthropic Claude 3.7 Sonnet is available under the pro tier. Created Feb 24, 2025 with a 200,000 token context limit, pricing is $3 per million input tokens and $15 per million output tokens. Anthropic Claude Opus 4 is also available under the ultimate tier. Created May 22, 2025 with a 200,000 token context limit, pricing is $15 per million input tokens and $75 per million output tokens.

Anthropic Claude 3.7 Sonnet (thinking) has been added to the reasoning menu as an ultimate tier option. Created Feb 24, 2025 with a 200,000 token context limit, pricing is $3 per million input tokens and $15 per million output tokens.

perplexity/sonar-reasoning has been added to the reasoning menu under the pro tier. Created Jan 29, 2025 with a 127,000 token context limit, pricing starts at $1 per million input tokens and $5 per million output tokens.

perplexity/sonar-reasoning-pro has been added to the reasoning menu under the pro tier. Created Mar 7, 2025 with a 128,000 token context limit, pricing starts at $2 per million input tokens and $8 per million output tokens.

Anthropic Claude 3.5 Haiku has been added to the chat model menu under the pro tier. It offers enhanced speed, coding accuracy, and tool use, making it ideal for real-time applications like chat interactions and instant coding suggestions. Created Nov 4, 2024 with a 200,000 token context limit, pricing is $0.80 per million input tokens and $4 per million output tokens.

OpenAI GPT-5 Chat has been added under the pro tier for advanced, natural, multimodal, and context-aware conversations. It supports a 400,000 token context with pricing at $1.25 per million input tokens and $10 per million output tokens.

OpenAI GPT-5 is now available under the pro tier, offering major improvements in reasoning, code quality, and user experience. It also supports a 400,000 token context and is priced at $1.25 per million input tokens and $10 per million output tokens.

OpenAI GPT-5 Mini provides lighter-weight reasoning with reduced latency and cost. Available under the pro tier, it offers a 400,000 token context and costs $0.25 per million input tokens and $2 per million output tokens.

OpenAI GPT-5 Nano is optimized for ultra-low latency environments. Under the pro tier, it features a 400,000 token context and pricing of $0.05 per million input tokens and $0.40 per million output tokens.