v0.2.17 - Desktop agent tools

This release combines a few useful tools for developing the LLMs using computers almost like we do.

Webtop

Courtesy of LinuxServer.io, a service providing a dockerized GUI sandbox to aid with development and tests of desktop LLM agents (👋🏻 computer use). Additionally, provides a mean of accessing Harbor App and Harbor CLI remotely from the instance where they are running (via VNC)

# Will take a while
harbor build webtop
# Start and open
harbor up webtop --open

webtop service will mount Harbor CLI from your host as well as install latest release of Harbor App.

OmniParser

A release from Microsoft, breakthrough tool for annotating GUI for using with autonomous desktop agents.

⚠️ Requires Nvidia+CUDA to run.

# Will take a long while
harbor build omniparser
# Start and open
harbor up omniparser --open

Promptfoo

promptfoo is a CLI tool that allows to develop and evaluate prompts and LLM APIs. It's neatly built and comes with all imaginable ways to configure and evaluate prompts. Harbor's version comes with a couple of examples and is preconfigured to work with built-in ollama

Misc

• Small tweaks to harbor CLI
• Unsuccessful perplexideez integration (blocked by support for running on localhost)
• docs: harbor app README.md updated by @ic4l4s9c in #79

Full Changelog: v0.2.16...v0.2.17

v0.2.16 - 4 new services, bugfixes, Discord

Overview

This release comes a few new services, as well as some additional helper features.

Open WebUI Pipelines

Integration authored by @ic4l4s9c 🎉

pipelines bring modular, customizable workflows to any UI client supporting OpenAI API specs.

harbor up pipelines

• Official documentation
• Examples on GitHub

Qdrant

Also authored by @ic4l4s9c 🎉

First shared vector store for Harbor - will be used by supported satellites in the future.

harbor up qdrant

REST API: http://localhost:34221
Web UI: http://localhost:34221/dashboard
GRPC API: http://localhost:34222

Chat Nio

New promising WebUI for LLMs. Harbor pre-configures it for ollama and searxng out of the box.

# Start the service
harbor up chatnio
# [Optional] Open the UI
harbor open chatnio

K6

Load testing toolkit. Harbor's version comes with a lot of things pre-configured testing of OpenAI-compatible LLM APIs, including a custom grafana dashboard (with token stats), api client helpers and sample scripts.

Discord

We have a Discord now. Come say "Hi 👋🏻" if you want to chat to someone about Harbor.

Misc

• webui - fixing missing override.env after v0.2.15
• harbor open - now supports overrides via <service>.open_url config

New Contributors

• @ic4l4s9c made their first contribution in #75

Full Changelog: v0.2.15...v0.2.16

v0.2.15 - CLI fixes and improvements

harbor env

You can now set service-specific env vars directly via Harbor CLI

# Show the current environment variables for the "n8n" service
harbor env n8n

# Get a specific environment variable
# for the dify service (LOG_LEVEL under the hood)
harbor env dify log.level

# Set a brand new environment variable for the service
# All three are equivalent
harbor env cmdh NODE_ENV development

Misc

• cmdh - fixing service build and init
• n8n - using N8N_SECURE_COOKIE by default, since Harbor is expected to run on localhost
• comfyui - disable web auth by default to provide seamless integration with webui out of the box
• documentation fixes and improvements

v0.2.14

• Relaxing Python version requirement for PyPI installs
• Attempt to explicitly resolve home folder permissions for Harbor App

Full Changelog: v0.2.13...v0.2.14

v0.2.13 - n8n

v0.2.13 - n8n integration

n8n is a low-code workflow automation tool that has good support for LLMs.

# Start the service
harbor up n8n
# [Optional] open in the browser
harbor open n8n

Harbor's integration is mostly around allowing to run and configure n8n as the other Harbor services, as pre-configuring the connections/workflows dynamically is not something that is currently supported.

Please find detailed instructions in the service docs

Misc

• Broken (for now) integration with the Bolt.new, awaiting for the fixes in the upstream

Full Changelog: v0.2.12...v0.2.13

v0.2.12 - CLI improvements and bugfixes

The most notable feature in this release are Aliases - allowing you to configure shortcuts for the harbor run to execute.

# Configure an alias - can be arbitrary shell command
harbor alias set echo 'echo "I like $APP!"'

# Run with "harbor run"
APP=Harbor harbor run echo
"I like Harbor!"

# Reference
alias|aliases|a [ls|get|set|rm] - Manage Harbor aliases"
  echo "    alias ls|list               - List all aliases"
  echo "    alias get <name>            - Get an alias"
  echo "    alias set <name> <command>  - Set an alias"
  echo "    alias rm|remove <name>      - Remove an alias"

See more detailed documentation in the CLI reference.

Misc

• harbor up - Now supports additional modifiers (logs, open, no defaults)
• harbor stats - shows stats stream for Harbor services
• harbor down - now correctly matches sub-services
• harbor shell - allows to choose shell out of the box: harbor shell <service> ash (note that the service must bundle given shell)
• harbor config - correctly handle escape for dict values
• minor refactoring of internal command structure
• harbor doctor - output specific location of Harbor home for the debug
• Harbor App - handle case when a selected profile is deleted outside of the app
• User Guide improvements
• CLI Reference expansion

Full Changelog: v0.2.11...v0.2.12

v0.2.11

v0.2.11 - NPM, PyPI, Windows app

very experimental features:

• Installing Harbor via npm, pipx
  • See more in the revised install instructions
• Enabling Windows Harbor App builds (likely won't be functional)

Misc

• Switching to Tauri v2.0.0 release version
• Fix being unable to run multiple Ollama CLIs at the same time
• fixfs now accounts for plandex

Full Changelog: v0.2.7...v0.2.11

v0.2.7

v0.2.7

• boost - support per-request parameterization
  • small improvements to llm and chat APIs
  • a series of experimental custom modules
• bench
  • short judge prompt type
  • fix tasks report generation to correctly display used prompt
  • allow specifying judge max_tokens via config and CLI

Full Changelog: v0.2.6...v0.2.7

v0.2.6 - Search, Repopack

v0.2.6

Harbor App now allows searching for services as well as configuration items.

We now also have a friend.

simplescreenrecorder-2024-10-05_00.30.29.mp4

Repopack integration

Repopack is a tool that packs your entire repository into a single, AI-friendly file.
Perfect for when you need to feed your codebase to Large Language Models (LLMs) or other AI tools like Claude, ChatGPT, and Gemini (feeding Harbor to Gemini 1.5 Pro costs ~$1).

cd ./my/repo
harbor repopack -o my.repo --style xml

See fabric and other CLI satellites for some interesting use-cases.

Full Changelog: v0.2.5...v0.2.6

v0.2.5

v0.2.5 - Theme freedom

Harbor is all about customization. I decided that the the app theme should be fully representative of this, plus my designer colleagues told me it's a very bad idea, so enjoy - full customization of the Harbor App theme.

Demo on YouTube

Misc

• Integrated Nexa - unfortunately there's a big bug in the Nexa Server that makes streaming unusable, otherwise this integration is something I'm quite proud of, as Harbor delivers two key improvements upon official Nexa releases:
  • Compatibility with Open WebUI (with streaming disabled, for now)
  • CUDA support in Docker
  • That said, only NLP models were tested, so Audio/Visual verification is warmly welcomed!
  • Nexa Service docs
• Small improvements to the App UI: Loaders, gradient service cards based on the tags (very faint)

Full Changelog: v0.2.4...v0.2.5