Skip to content

Roadmap

Zuni is developing from a small command-line question-answering tool into a compact terminal research assistant.

This roadmap describes direction rather than a fixed release schedule. Items can change as the project evolves.

Available now

  • One-command questions with zuni ask
  • Interactive configuration with zuni config
  • OpenAI-compatible chat-completions support
  • Markdown rendering in the terminal
  • Asynchronous HTTP requests
  • Retry handling for temporary LLM API failures
  • Tool-calling agent
  • DuckDuckGo web search
  • Web-page extraction to Markdown
  • Numbered source tracking and citations
  • Fallback search flow for models without tool calling
  • Direct-answer mode with --no-search
  • First PyPI release (0.1.0)

In progress

  • Stronger validation of model-generated tool calls
  • More robust URL and redirect safety
  • Automated tests for configuration, tools, citations, and the agent loop
  • Expanded documentation and API reference

Planned

  • Streaming responses
  • Conversation mode
  • Reading input from stdin
  • Multiple saved configuration profiles
  • Continuous integration

Ideas

  • Shell completions
  • JSON and other structured output formats
  • Additional research tools
  • Local and offline-first workflows
  • More provider-specific configuration options

Note

The roadmap is intentionally flexible. Features may move, change, or be removed as Zuni's design develops.