AI Chat

Nya AI offers first-class AI chatting functionality, designed to completely replace separate AI chatting applications.

Document Parsing

When adding files to a chat, the following file types are supported and will be parsed before being provided as input:

• docx: Parsed into Markdown format; any embedded images are automatically extracted
• pptx: Text content is extracted
• pdf: Provides two parsing methods:
  • Extract all text content: Suitable for mainly text-based content
  • Screenshot each page for input: Suitable for pages with complex content and fewer pages. The model must support image inputs.
• xslx: Converted into Markdown tables

Additionally, for text-type files (code, csv, json, etc.), the entire text is automatically identified and directly read.

Native File Input

By default, Nya AI only directly inputs text and images to the model. Other file types will attempt to use the built-in parsers mentioned above for conversion.

However, some service providers and models natively support certain file types (e.g., pdfs, audio) as input. To provide these directly to the model, you need to manually adjust the Multimodal Capabilities configuration for that model:

1. Find the model you want to adjust in the model list on the provider page.
2. Click the "tune" button on the right of the model.
3. Modify the "User Input Types" under "Multimodal Capabilities" to add the MIME types you want to be input directly.

Built-in Plugins

Nya AI provides several built-in plugins to extend AI capabilities. You can enable them by clicking the "extension" button in the chat page:

• Web Search & Crawling: This plugin gives the AI two tools to access the Internet:
  • search: Calls a search engine to search the web
  • crawl: Crawls the full page content of a specific URL
• Mermaid Diagrams: Allows the AI to draw diagrams using Mermaid syntax in its responses
• Workspace Search: Enables the AI to search workspace content and use it to answer questions, similar to RAG

MCP

Besides built-in plugins, Nya AI also supports adding MCP (Model Context Protocol) servers:

1. Click on "extension MCP Plugins" in the left shortcut sidebar.
2. Click on "add MCP Plugin" in the right sidebar.
3. Enter any name and the MCP server URL to add it.

NOTE

Since Nya AI currently only has a web version, it only supports Streamable HTTP connections, and the server needs to allow cross-origin requests (CORS). We will release a desktop client later for full support.

Once added, you can enable the plugin in chats. Supported MCP features include: Tools, Resources, and Prompts.

Custom Providers

In addition to the models provided by the platform, you can add any custom provider and models. We support multiple formats including OpenAI Compatible / Responses, Google, Anthropic, etc.

Note that all workspace members can see and use the providers added to the workspace. We recommend setting a usage limit for your API key at your service provider's end to avoid unexpected expenses.