Skip to main content

About Us

Product Feedback​

If you have any feedback (bugs or suggestions), please send it to hello@gptaiflow.com or let us know through our user surveys:

2026-05-30-v0.1.102-Smoother Obsidian Writing · Faster Search · More Reliable Results​

This release is not about adding a pile of new buttons. It is about making the main Obsidian writing flow smoother, more stable, and less distracting: from finding material, to calling tools, to continuing your draft, there are fewer interruptions, shorter waits, and more reliable outcomes.

Smoother Obsidian Writing

  • Long-running tasks in writing flows now provide clearer execution feedback, so they are less likely to look stuck while still running.
  • Agent timeout and interruption behavior is clearer. Extra-long tasks are terminated correctly and return an explicit status instead of leaving you waiting with no response.
  • The new execution pipeline has been further refined, making writing steps feel more continuous and making switching and resuming feel more natural.

Faster Content Search, Better Efficiency

  • Added a search_file alias, so common writing-related search commands work directly without making you remember the exact command name.
  • grep no longer fails in directory-search scenarios using flags: "-l", making file discovery more stable.
  • If rg is not available locally, the system now falls back to grep automatically, so search does not break your writing flow.

More Reliable Results, Better Accuracy

  • The tool-calling pipeline has been further hardened, reducing failures caused by tool_calls pairing mismatches.
  • Observability for long-running workflows has been improved, making issues easier to trace and reducing the chances of lost results or unclear status.
  • The overall goal is simple: spend your time on what to write, not on fixing tool calls.

What You Will Notice Right Away

  • Fewer writing sessions get interrupted by technical problems.
  • The path from idea to finished draft feels more continuous.
  • The same task usually takes fewer back-and-forth actions.

2026-05-10-v0.1.101-Clearer SKILLS · More Reliable Continue · Smoother Interaction​

Simpler SKILLS New Skills now appear in the selector by default, with clearer selection logic and a cleaner panel overall.

More reliable Continue Fixed cases where tool results returned but were not saved. Continue now keeps more complete records, making long tasks and resumed runs more stable.

Experience improvements Interaction is smoother in long-message and multi-message scenarios, and Skill management copy is now more consistent.

2026-05-05-v0.1.100-Multi-Channel Concurrency · Audio Transcription · Smoother Experience​

Multi-channel concurrent execution AI can now work across multiple channels at once — no more waiting for one to finish before switching. The channel menu shows live status for each channel, with notifications when background channels complete.

Audio transcription tool Supports mp3, wav, m4a, aac, flac, ogg, mp4, and more. Large files are split automatically; output in SRT/VTT subtitles or plain text.

Skill enable management & mode/tool permission control A new Skill management tab in Settings lets you enable or disable built-in, remote, and local Skills individually. AI tools now strictly follow the active conversation mode — a notice appears at the top when there's a permission mismatch.

Experience improvements: The input box is now independent for smoother typing; switching back to the same channel within 2 hours restores your reading position.

2026-04-26-0.1.99-More Stable Backend · Longer Context · Smarter SKILL Ecosystem​

Background processing now actually stops Previously, clearing messages and switching away didn't stop AI from running in the background. Now it does — you stop, it stops.

Long conversations no longer "forget" As conversations grow longer, important context no longer silently drops away. When context gets too long, a summary is automatically generated so you can keep going without interruption.

Chinese input no longer stutters Occasional lag and delays with macOS Chinese input methods have been fixed. Typing is much smoother.

SKILL cloud sync SKILLs from built-in, custom, and cloud sources are now automatically merged and silently synced on startup — no manual action needed.

Search bar now displays correctly The search bar was previously blocked by the top button bar. This has been fixed.

2026-04-19-0.1.98-Let AI Actually Work Through Your Obsidian Vault — LLM-Wiki & Wiki Mode​

One question drove this release: you've been collecting notes for years — can AI finally help you use them?

With v0.1.98, the answer is yes.

✨ Highlights

📚 Wiki Mode + LLM-Wiki SKILL — AI that understands your knowledge base

Switch to Wiki Mode and AI can search your notes, follow [[wikilink]] connections, and answer "which notes link to this one?" Pair it with the new LLM-Wiki SKILL to have AI organize notes around a topic, generate an index, or surface the thread connecting your ideas — no manual digging required.

Agent won't quit halfway through

When a long task is interrupted, a "Continue" button appears at the end of the message.

10+ new Obsidian SKILLs

Thought evolution tracker, article structure optimizer, morning planner, reading companion, sentence recycler… Each is a writing workflow designed for real situations, triggered directly from the file tree right-click menu.

🔧 Other Improvements

  • Note organization outputs versioned files (-v1, -v2) — originals are never overwritten
  • AI adjusts its tone based on time of day (morning, afternoon, evening, etc.)
  • Writing Material Panel: open from the toolbar to pull up related materials while writing
  • Fixed occasional Claude reply errors; faster and more stable AI responses