Skip to content

Releases: joelst/flint

Release list

Flint v0.5.0

Flint v0.5.0 Latest
Latest

Choose a tag to compare

View all tags
@joelst joelst released this 01 Sep 13:56
v0.5.0
93e7c1f
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

What's Changed

  • Harden release packaging, fix the updater endpoint, validate audio input by @joelst in #27
  • Run your own models, and serve them on demand by @joelst in #29
  • Warn before memory runs out, and let the model pool bound itself by @joelst in #31

Full Changelog: v0.4.4...v0.5.0

Models
• SDK-backed catalog (broader than Foundry CLI), search + filter
• Hardware-aware recommendations; CPU/GPU/NPU acceleration variants; EP detection
• Download w/ progress, load, unload, delete, cache removal
• Multi-model pool — concurrent loads, variant-ID HTTP routing
• Per-acceleration-track update notifications with direct update download
• Recommended starter models for first run

Chat
• Streaming responses, cancel/stop
• Multi-conversation sidebar, persisted to localStorage
• 9 personas (Default, Coder, Code Reviewer, Tutor, Analyst, Ultra Concise, Creative, Vision Analyst, Technical Writer) + custom system prompts
• Vision: up to 4 images via attach, paste, or drag-and-drop
• Host-aware Flint/Foundry context injection
• Guarded URL → context fetch (SSRF + size limits, article extraction)
• Push-to-talk dictation into the chat box
• Markdown rendering w/ code blocks

Audio
• STT via mic or file upload; STT-capable model filtering

Compare
• Side-by-side bake-off across 2–3 models/variants, one prompt
• Ratings + markdown export

Monitor
• Live pool table, resource gauges
• Access log (ring buffer + ~/.flint/logs/ rotation) and audit trail for destructive/config commands
• Export as JSON or CSV

Integrations
• Copy-paste snippets, OS-aware, for 13 targets: OpenAI SDK (JS/Python), Continue.dev, OpenCode, Codex CLI, Cline, Claude Code, GitHub Copilot (VS Code), Copilot CLI, Codex App, Hermes Agent, OpenClaw, Droid, Pi

Service & network
• OpenAI-compatible endpoint at http://127.0.0.1/:/v1
• Start/stop from Diagnostics; configurable bind address + port with Apply & restart
• Loopback default; non-loopback requires explicit confirmation

Settings / Diagnostics / Help
• Light & dark theme, OS autostart login item, auto-start service, default chat/audio models
• Keyboard shortcuts ( 1 – 6 views, new chat, send, push-to-talk, ? reference)
• First-run coach + Help tab with troubleshooting
• Diagnostics with sidecar/service state; About shows bundled vs PATH Node

Platform
• Tauri 2 (Rust + WebView), Svelte 5 + TypeScript SPA
• Bundled Node 22 sidecar (externalBin) + bundled Foundry native cores
• Auto-updater (signed latest.json ); Windows MSI + NSIS signed via Azure Trusted Signing; macOS Apple silicon DMG/app

Contributors

joelst
Assets 7
Loading

Flint v0.4.4

Flint v0.4.4

Choose a tag to compare

View all tags
@github-actions github-actions released this 30 Aug 20:43
v0.4.4
6225799
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

Models

• SDK-backed catalog (broader than Foundry CLI), search + filter
• Hardware-aware recommendations; CPU/GPU/NPU acceleration variants; EP detection
• Download w/ progress, load, unload, delete, cache removal
• Multi-model pool — concurrent loads, variant-ID HTTP routing
• Per-acceleration-track update notifications with direct update download
• Recommended starter models for first run

Chat

• Streaming responses, cancel/stop
• Multi-conversation sidebar, persisted to  localStorage 
• 9 personas (Default, Coder, Code Reviewer, Tutor, Analyst, Ultra Concise, Creative, Vision Analyst, Technical Writer) + custom system prompts
• Vision: up to 4 images via attach, paste, or drag-and-drop
• Host-aware Flint/Foundry context injection
• Guarded URL → context fetch (SSRF + size limits, article extraction)
• Push-to-talk dictation into the chat box
• Markdown rendering w/ code blocks

Audio

• STT via mic or file upload; STT-capable model filtering

Compare

• Side-by-side bake-off across 2–3 models/variants, one prompt
• Ratings + markdown export

Monitor

• Live pool table, resource gauges
• Access log (ring buffer +  ~/.flint/logs/  rotation) and audit trail for destructive/config commands
• Export as JSON or CSV

Integrations

• Copy-paste snippets, OS-aware, for 13 targets: OpenAI SDK (JS/Python), Continue.dev, OpenCode, Codex CLI, Cline, Claude Code, GitHub Copilot (VS Code), Copilot CLI, Codex App, Hermes Agent, OpenClaw, Droid, Pi

Service & network

• OpenAI-compatible endpoint at  http://127.0.0.1:/v1 
• Start/stop from Diagnostics; configurable bind address + port with Apply & restart
• Loopback default; non-loopback requires explicit confirmation

Settings / Diagnostics / Help

• Light & dark theme, OS autostart login item, auto-start service, default chat/audio models
• Keyboard shortcuts ( 1 – 6  views, new chat, send, push-to-talk,  ?  reference)
• First-run coach + Help tab with troubleshooting
• Diagnostics with sidecar/service state; About shows  bundled  vs  PATH  Node

Platform

• Tauri 2 (Rust + WebView), Svelte 5 + TypeScript SPA
• Bundled Node 22 sidecar (externalBin) + bundled Foundry native cores
• Auto-updater (signed  latest.json ); Windows MSI + NSIS signed via Azure Trusted Signing; macOS Apple silicon DMG/app

Loading