Ever tried to find a command in your terminal history by pressing ↑ the up arrow key repeatedly?
If you use the terminal a lot, you’ve probably faced this:
You know you’ve run a command before…
But now you’re stuck doing:
Ctrl + R → type → scroll → scroll → wrong command → repeat
And somehow, it still pulls results from a completely different project.
That’s when it hit me:
Shell history isn’t broken because it’s slow.
It’s broken because it has zero context.
So I built something to fix that.
Meet Termim
Termim = your terminal remembers per project.
Instead of one global history mess, Termim gives you:
- Project-specific history
- Behavior-based suggestions
- Fast navigation without friction
No noise. Just relevant commands.
The Real Problem
Your terminal history today:
- mixes all projects together
- has no idea where you are
- treats everything as equal
So when you search, you get:
- commands from last week
- commands from another repo
- commands that don’t even apply anymore
It’s not helpful—it’s overwhelming.
What Termim Changes
1. Project-Aware History
When you’re inside a project:
- you only see commands from that project
- no cross-project pollution
- no accidental mistakes
Your ↑ (arrow up) key finally makes sense again.
2. It Learns How You Work
Termim doesn’t just store commands—it tracks patterns.
Example:
git add .
→ git commit
→ git push
Over time, it learns these flows using weighted transitions.
So instead of searching, you just… move forward.
3. Navigation That Feels Natural
↑ → project history → global history
↓ → back into project → predictions
No new mental model.
Just better behavior.
4. Built for Speed
No DB. No daemon. No bloat.
- ~15ms latency
- zero dependency core
- works across Bash, Zsh, Fish, PowerShell
It feels like part of your shell, not an add-on.
Why I Didn’t Just Use Existing Tools
I tried tools like Atuin, McFly, etc.
They’re good—but:
- they still feel global
- context is “soft”, not enforced
- some rely on background services
I wanted something stricter and simpler:
When I’m in a project, my terminal should behave like it knows it.
Small Details That Matter
A few things I cared a lot about:
- Hard context isolation (no leaking commands)
- Success-only learning (ignore failed commands)
- Privacy-first (everything local, secrets redacted)
- Deterministic behavior (no randomness in navigation)
What It’s Not
Termim is NOT:
- an AI copilot
- a cloud product
- a heavy productivity suite
It’s intentionally minimal.
It solves one problem well:
making terminal history actually useful.
Who This Is For
If you:
- jump between multiple repos daily
- live inside your terminal
- rely on command recall a lot
You’ll feel the difference immediately.
Final Thought
We’ve accepted bad terminal history for years.
Termim is just a small attempt to fix that.
Not by adding more features... but by adding context.
Termim on GitHub: Termim
If you want to try it or give feedback, I’d genuinely appreciate it 🙌
Top comments (1)
Interesting stuff! As someone who has started spending more of my time in the terminal this is genuinely something that I could benefit from using! Thanks for sharing!