I shipped a hotfix. It came with a dashboard.

Flash 0.5.6 was supposed to be a hotfix.

Sub-agents were broken in the web UI. One bug, one fix, done.

Then I kept going.

Bring your memory with you

ChatGPT knows you. Claude knows you. Your local model starts from zero every time.

Not anymore.

Open Settings in Flash Web, click Import memory, and copy one prompt. Paste it into whichever AI already knows you. It answers with everything it remembers, in one code block. Paste that back.

Flash saves it line by line, exactly as written. No model in between to summarize or invent anything, and you see every fact before it lands.

Your model runs on your machine, and now it knows how you like to work.

Your usage, at a glance

Settings also has a usage dashboard:

• Messages, tokens, tool calls, and time working
• Your current streak and your longest
• A year of activity as a heatmap
• Which models you actually use

All counted from chats saved on your machine. Nothing leaves it.

The small stuff that adds up

• Edit, add, or forget memories right in the page
• Light, Dark, or System, with a circle that sweeps the new theme across the screen
• Pick a project before you send the first message
• Rename chats from the sidebar
• A real logo in your tab

Oh, and the actual hotfix

• Sub-agents now wake their chat when they finish and report back, same as the terminal
• Starting a sub-agent no longer crashes the turn with a TypeError
• Plans show up as a live checklist in the browser instead of in the server's terminal

Try it

Flash is a local AI coding agent for your terminal and browser, running on Ollama. Free, open source, and it runs on your own hardware unless you point it somewhere else.

curl -fsSL https://raw.githubusercontent.com/Natuworkguy/Flash/main/install.sh | bash
flash --web

Already have it? Run /update.

Release notes: https://github.com/Natuworkguy/Flash/releases/tag/v0.5.6

What would you want your local model to remember about you?