llmost (Binary Bundle)

Run local models from a terminal UI with a local OpenAI-compatible gateway.

This repository is a binary distribution bundle centered on ./bin/llmost.

Quick Start

./bin/llmost

On Home:

1. Press Enter to install runtime (if needed), pull the starter model, and start the gateway.
2. Press Enter again to jump to Chat.

What This Binary Offers

• Terminal UI tabs: Home, Models, Serve, Setup, Tuning, Advisor, Chat, Use, Logs
• Local OpenAI-compatible API gateway
• Runtime install/start/stop from TUI and CLI
• Model scan/import/pull/register lifecycle
• Scoped tuning via llmost tune ...
• Python host-first resolution with managed fallback support
• Diagnostics: doctor/status/ports/runtime checks/logs

Included Docs

• User guide: docs/user_guide.md link
• Troubleshooting: docs/troubleshooting.md link
• Security: docs/security.md link

Essential Commands

./bin/llmost
./bin/llmost --help
./bin/llmost doctor
./bin/llmost status
./bin/llmost ports
./bin/llmost stop
./bin/llmost cleanup-ghosts
./bin/llmost logs

Python Reliability Commands

llmost prefers host Python and only falls back to managed setup when needed.

./bin/llmost python status
./bin/llmost python check
./bin/llmost python use /absolute/path/to/python3
./bin/llmost python clear-preference

Tuning Commands (Scoped)

./bin/llmost tune show
./bin/llmost tune set serve.context_length 8192
./bin/llmost tune set chat.temperature 0.2
./bin/llmost tune set chat.thinking_mode off
./bin/llmost tune reset chat.temperature

Security Defaults

• default bind host is loopback (127.0.0.1)
• local-first workflow requires no cloud account
• when exposing outside loopback, use bearer auth

Example:

./bin/llmost serve --model-id <id> --host 0.0.0.0 --port 8787 --bearer-token '<strong-token>'

Bundle Notes

• Binary path: ./bin/llmost
• Runtime/model state lives under ./config and ./var
• Bundled vendor components are under ./vendor

License

Apache-2.0. See LICENSE.