pi-autotalk
Advanced tools
Timed AutoTalk for Pi: periodically send your editor thoughts to the agent for brainstorming.
Pi AutoTalk turns Pi's interactive editor into a lightweight thinking buffer. Turn it on, type rough notes or mutterings into the editor, and Pi periodically sends them back to the agent as an AutoTalk thought memo.
Status: MVP implemented and locally dogfooded. AutoTalk is always off by default and only runs after /autotalk:on.
Sometimes you do not want to craft a polished prompt. You want to think out loud, let the agent react, and keep moving. AutoTalk is for that mode:
When enabled, AutoTalk:
[AutoTalk] thought-memo prefix;It is deliberately conservative. Slash commands are protected, enabled state is not persisted, and AutoTalk messages tell the agent not to edit files, run commands, or send external requests unless explicitly asked.
| Command | Purpose |
|---|---|
/autotalk:on | Start timed sending. Does not immediately send existing editor text. |
/autotalk:off | Stop timed sending and clear AutoTalk footer status. |
/autotalk:mode | Select busy-agent delivery mode: followUp or steer. |
/autotalk:settings | Configure interval seconds and delivery mode. |
10 seconds.5 to 120 seconds.AutoTalk skips a tick if the editor changed within the last 2 seconds. This avoids sending half-typed fragments at the exact moment you are still writing.
If the editor starts with /, AutoTalk pauses for that tick and does not clear the editor. This prevents commands like /autotalk:off, /reload, or /model from being auto-sent as thought memos.
After sending a non-empty buffer, AutoTalk clears the editor only if the editor still exactly matches the text that was sent. If you type more while a send is happening, the new text is preserved.
If the editor is empty, AutoTalk sends a one-shot continuation prompt:
[AutoTalk]
The editor is empty. From the conversation so far, ask one question to think about next.
That empty prompt is sent only once. It resets after a real non-empty thought memo is sent.
When the agent is already responding:
followUp queues the thought memo after the current work finishes.steer sends it as steering input for the current stream.Use followUp for orderly logs. Use steer for live brainstorming and course correction.
Non-empty editor text is sent like this:
[AutoTalk]
This is an automatically sent user thought memo.
Expand the ideas, organize the key points, and ask one follow-up question.
Unless explicitly requested, do not edit files, run commands, or send external messages.
--- Thought memo ---
{editor text}
AutoTalk persists interval and delivery mode globally at:
~/.pi/agent/extensions/autotalk/settings.json
Example:
{
"intervalSec": 10,
"deliveryMode": "followUp"
}
The enabled/disabled state is never persisted. Every Pi session starts with AutoTalk off.
pi install npm:pi-autotalk
pi install git:github.com/eiei114/pi-autotalk
Clone or keep the repo under:
C:/Users/Keisu/Projects/OSS/pi-autotalk
Then add it to a project .pi/settings.json:
{
"packages": [
"..\\..\\..\\OSS\\pi-autotalk"
]
}
After editing settings, reload Pi:
/reload
/autotalk:on
Type a thought into the editor and stop typing. After the configured interval, Pi will receive it as an AutoTalk thought memo.
Useful commands:
/autotalk:settings
/autotalk:mode
/autotalk:off
| Path | Purpose |
|---|---|
extensions/index.ts | Pi extension entrypoint and command/timer wiring. |
lib/autotalk.ts | Settings validation, persistence, and message formatting helpers. |
tests/ | Node test suite for package metadata, formatting, settings, and registration. |
docs/ | Release and publishing docs. |
npm install
npm run ci
npm run ci runs:
tsc --noEmitnode --test tests/*.test.mjsnpm pack --dry-runAutoTalk intentionally sends editor contents to the active Pi agent. Treat it like an auto-submit feature.
Safety choices in the MVP:
/autotalk:on;/;Pi packages can execute code with your local permissions. Review extensions before installing third-party packages.
For vulnerability reporting, see SECURITY.md.
This package uses npm Trusted Publishing (OIDC). No NPM_TOKEN is required.
On merge to main with a package.json version bump, .github/workflows/auto-release.yml creates the v* tag and GitHub release, then dispatches .github/workflows/publish.yml to publish to npm.
Maintainer flow:
npm version patch # or minor/major — updates package.json, package-lock.json (if present), creates a git commit and tag by default
git push origin main
You can also bump version in package.json in a PR; after merge, auto-release handles tagging and publish.
After merge, run the post-release verification checklist in docs/release.md to confirm the GitHub tag/release, Publish Package workflow, and npm latest all match the new version. A green auto-release run alone is not proof npm received the package.
See docs/release.md for setup, recovery, and verification details.
MIT
FAQs
A Pi extension for timed AutoTalk brainstorming from the editor input.
We found that pi-autotalk demonstrated a healthy version release cadence and project activity because the last version was released less than a year ago. It has 1 open source maintainer collaborating on the project.
Did you know?
Socket for GitHub automatically highlights issues in each pull request and monitors the health of all your open source dependencies. Discover the contents of your packages and block harmful activity before you install or update your dependencies.
Research
/Security News
11 malicious NuGet tools pose as game cheats to deploy Windows payloads, track hosts, and use Google Sheets for telemetry and control.
Research
/Security News
4 compromised asyncapi packages deliver miasma botnet loader on macOS, Linux and Windows.
Research
/Security News
A compromised jscrambler npm release added a malicious preinstall hook that runs hidden native binaries on Linux, macOS, and Windows.