Vane AI Answering Engine

When a user sends a message via the UI, the application invokes the POST /api/chat endpoint. The processing pipeline follows three main stages:

1. Classification: The system analyzes the question to determine if research is required, which widgets to display, and how to rewrite the question for better clarity.
2. Parallel Execution: The system runs Research (web lookups or searching uploaded files) and Widgets (structured helpers like weather, stocks, or calculations) simultaneously.
3. Answer Generation: Once context is gathered, a chat model generates the final response, incorporating citations from the gathered sources.

Vane uses two distinct mechanisms to augment answers:

Widgets

Widgets are small, structured helpers (e.g., weather, stocks, calculations) that run in parallel with research. They are displayed in the UI immediately while the main answer is being generated. Note that widgets provide context but are not intended to be cited by the model.

Research

Research is the process of gathering background information. Depending on your configuration, this includes performing web lookups and searching through user-uploaded files.

When stream: true is passed in the request, the API returns a text/event-stream using Server-Sent Events (SSE). Each line is a newline-delimited JSON object.

Message Types:

• init: Initial connection message (data contains "Stream connected").
• sources: An array of all sources used in the response.
• response: Chunks of the generated answer text.
• done: Indicates the stream is complete.

{"type":"init","data":"Stream connected"}
{"type":"sources","data":[...]}
{"type":"response","data":"Vane is an "}
{"type":"response","data":"innovative..."}
{"type":"done"}

If you are using pre-built Docker images, update Vane by pulling the latest image and recreating the container. Your settings are preserved automatically if you are using a named volume like vane-data.

For the standard image:

docker pull itzcrazykns1337/vane:latest
docker stop vane
docker rm vane
docker run -d -p 3000:3000 -v vane-data:/home/vane/data --name vane itzcrazykns1337/vane:latest

For the slim version (requires SEARXNG_API_URL environment variable):

docker pull itzcrazykns1337/vane:slim-latest
docker stop vane
docker rm vane
docker run -d -p 3000:3000 -e SEARXNG_API_URL=http://your-searxng-url:8080 -v vane-data:/home/vane/data --name vane itzcrazykns1337/vane:slim-latest

After updating, verify the changes at http://localhost:3000.

If you already have a SearxNG instance running, use the slim version of the Vane Docker image. You must provide the SEARXNG_API_URL environment variable.

Requirements for your SearxNG instance:

1. JSON format must be enabled in SearxNG settings.
2. Wolfram Alpha search engine must be enabled.

Run the following command:

docker run -d -p 3000:3000 -e SEARXNG_API_URL=http://your-searxng-url:8080 -v vane-data:/home/vane/data --name vane itzcrazykns1337/vane:slim-latest

Replace http://your-searxng-url:8080 with your actual SearxNG URL.

You can add Vane as a shortcut in your browser's search bar to perform searches directly from the address bar.

1. Open your browser's settings and navigate to Search Engines.
2. Add a new site search with the following URL format: http://localhost:3000/?q=%s

Note: If Vane is hosted on a different machine or domain, replace localhost with your server's IP address or domain name, and ensure the port (default 3000) is correct.

To install Vane without Docker, you must first install SearXNG manually and ensure JSON format and the Wolfram Alpha engine are enabled in its settings.

Follow these steps:

1. Clone the repository:

   git clone https://github.com/ItzCrazyKns/Vane.git
   cd Vane

2. Install dependencies:

   npm i

3. Build the application:

   npm run build

4. Start the application:

   npm run start

Access the setup screen at http://localhost:3000 to configure your environment.

If you manage Vane by building your own Docker images from the source code, follow these steps:

1. Pull the latest changes from the repository:

   cd Vane
   git pull origin master

2. Rebuild the local Docker image:

   docker build -t vane .

3. Replace the existing container with the new build:

   docker stop vane
   docker rm vane
   docker run -p 3000:3000 -p 8080:8080 --name vane vane

Verify the update at http://localhost:3000 once the process completes.

The easiest way to run Vane is via Docker. The standard image includes both Vane and a bundled SearxNG search engine, so no additional setup is required for the search backend.

Run the following command to pull and start the container:

docker run -d -p 3000:3000 -v vane-data:/home/vane/data --name vane itzcrazykns1337/vane:latest

• -p 3000:3000: Maps the web interface to port 3000.
• -v vane-data:/home/vane/data: Creates a persistent volume for your data and uploaded files.

Once running, access the interface at http://localhost:3000 to configure API keys and models.

To update a non-Docker installation of Vane, pull the latest source code, update dependencies, rebuild, and restart the application:

1. Pull the latest changes:

   cd Vane
   git pull origin master

2. Install new dependencies:

   npm i

3. Rebuild the application:

   npm run build

4. Restart the application:

   npm run start

Settings are preserved automatically. Verify the update at http://localhost:3000.

You can control the tradeoff between response speed and the quality of the generated answer using the optimizationMode setting.

Available modes:

• speed: Prioritizes fast response times.
• balanced: A middle ground between speed and quality.
• quality: Prioritizes the depth and accuracy of the generated response.