DEV Community

Ashwin Belbase
Ashwin Belbase

Posted on AI-assisted

Improvements across content creation, AI providers, and terminal experience in Relay

NOTE: This article is drafted by AI and published via Relay for testing.

I spent time implementing several features that improve how users create, manage, and connect their written work within Relay. I also overhauled core system areas like the terminal and security settings to make the tool more robust.

Content Management and Organization

I added better tools for structuring content ideas. Now there is a Kanban board with drag functionality, filters, and search capabilities available. When organizing module lanes, I included a dedicated view for that purpose, along with an item drawer that supports comments and activity tracking for individual items.

For quick capture, users can utilize the palette activated by ⌘K. The Markdown editor was also updated to include a preview mode and image placeholders. Furthermore, sending articles directly to Dev.to is now supported as an unpublished draft, alongside general article syncing.

Integrations and Content Flow

I focused heavily on improving external service connections, particularly with Buffer. I implemented full OAuth sign-in using the Authorization Code + PKCE flow, along with a hosted HTTPS callback page for secure setup. The Buffer tab can now display channels, posts filtered by status, and allows users to edit or delete content directly.

The integration process is smoother:

  • When Plane connects through the app-installation bot token, I updated the settings screen to include an explicit guide on creating a Plane application.
  • I addressed several bugs related to integration status, ensuring that if a service disconnects, it clears cleanly from the UI.
  • For better control over posts, users can now view scheduled and sent Buffer posts within the queue panel. I also added functionality to summarize Dev.to posts into draft social media updates for various channels.

AI Functionality and Data Privacy

The artificial intelligence capabilities were significantly expanded and secured. I built a dedicated AI provider layer in Go that supports multiple providers, including OpenAI, Anthropic, and OpenRouter, alongside local options like Ollama and LM Studio with auto-detection. Critically, when handling private data, the system is blocked from cloud AI providers if the data comes from internal sources.

In the settings, users can now test an AI model before saving it as a default provider, enhancing the setup process.

Developer Tools and System Infrastructure

I improved several core operational areas:

  • Terminal: I updated the terminal to offer advanced features including xterm panes, tabs, splits, zoom functionality, and support for tmux keybindings. A quick terminal view (⌘/) is now available over other main views, and the system can detect installed shells to ensure proper operation on different operating systems.
  • Security & Settings: To improve security, I upgraded data storage to use AES-256-GCM with a keychain key for local vaults, and projects are lockable behind Touch ID on macOS. Secrets previously stored in plaintext config files are now moved into the OS keychain. The settings screen itself was overhauled, including sections for system guides, service lists, and update checks.
  • Performance & Build: I upgraded the application to Wails v3.1.0 and improved CI builds to support macOS, Windows, and Linux consistently, along with adding full build status publishing capabilities.

What's next
I plan to continue refining the existing features, including improving how links are managed across different systems and integrating more granular controls within the settings panel.

Top comments (0)