OWLEN

Terminal-native assistant for running local language models with a comfortable TUI.

Pre-Alpha Status

• This project is currently pre-alpha and under active development.
• Expect breaking changes, missing features, and occasional rough edges.
• Feedback, bug reports, and ideas are very welcome while we shape the roadmap.

What Is OWLEN?

OWLEN is a Rust-powered, terminal-first interface for interacting with local large language models. It focuses on a responsive chat workflow that runs against Ollama and surfaces the tools needed to manage sessions, inspect project context, and iterate quickly without leaving your shell.

Current Highlights

• Chat-first terminal UI built with ratatui and crossterm.
• Out-of-the-box Ollama integration with streaming responses.
• Persistent configuration, model caching, and session statistics.
• Project-aware context loading (reads OWLEN.md when present).
• Experimental coding assistant mode (opt-in build feature).

Getting Started

Prerequisites

• Rust 1.75+ and Cargo (rustup recommended).
• A running Ollama instance with at least one model pulled (defaults to http://localhost:11434).
• A terminal that supports 256 colours.

Clone and Build

git clone https://github.com/Owlibou/owlen.git
cd owlen
cargo build -p owlen-cli

Run the Chat Client

Make sure Ollama is running, then launch:

cargo run -p owlen-cli --bin owlen

(Optional) Try the Coding Client

The coding-focused TUI is experimental and ships behind a feature flag:

cargo run -p owlen-cli --features code-client --bin owlen-code

Using the TUI

• i / Enter – focus the input box.
• Enter – send the current message.
• Shift+Enter / Ctrl+J – insert a newline while editing.
• m – open the model selector.
• n – start a fresh conversation.
• c – clear the current chat history.
• h – open inline help.
• q – quit.

The status line surfaces hints, error messages, and streaming progress.

Configuration

OWLEN stores configuration in ~/.config/owlen/config.toml. The file is created on first run and can be edited to customise behaviour:

[general]
default_model = "llama3.2:latest"
enable_streaming = true
project_context_file = "OWLEN.md"

[providers.ollama]
provider_type = "ollama"
base_url = "http://localhost:11434"

Additional sections cover UI preferences, file limits, and storage paths. Each client persists its latest selections back to this file on exit.

Repository Layout

• crates/owlen-core – shared types, configuration, and session orchestration.
• crates/owlen-ollama – provider implementation that speaks to the Ollama API.
• crates/owlen-tui – ratatui-based UI, helpers, and event handling.
• crates/owlen-cli – binaries (owlen, owlen-code) wiring everything together.
• tests – integration-style smoke tests.

Development Notes

• Standard Rust workflows apply (cargo fmt, cargo clippy, cargo test).
• The codebase uses async Rust (tokio) for event handling and streaming.
• Configuration and chat history are cached locally; wipe ~/.config/owlen to reset.

Roadmap

• Add autoscroll.
• Push user message before loading the LLM response.
• Add support for "thinking" models.
• Add theming options.
• Provide proper configuration UX.
• Add chat-management tooling.
• Reactivate and polish the coding client.
• Add support for streaming responses.
• Add support for streaming chat history.
• Add support for streaming model statistics.
• Add coding client
  • Add support for in project code navigation.
  • Add support for code completion.
  • Add support for code formatting.
  • Add support for code linting.
  • Add support for code refactoring.
  • Add support for code navigation.
  • Add support for code snippets.
  • Add support for in project config folder.
• Add support for more local LLM providers.
• Add support for cloud LLM providers.

Contributing

Contributions are encouraged, but expect a moving target while we stabilise the core experience. Opening an issue before a sizeable change helps coordinate the roadmap.

License

License terms are still being finalised for the pre-alpha release.