louis030195

pi-coding-agent --mode rpc — streaming events lack originating-command id

Reproduction for a race condition in @mariozechner/pi-coding-agent's RPC mode. When a consumer sends a new prompt while a previous prompt is still streaming, events on stdout cannot be reliably attributed to the correct turn because streaming events (agent_start, turn_start, message_start, message_end, message_update, turn_end, agent_end) carry no originating-command id. Only command-ack response events are tagged.

How to run

bash repro.sh

louis030195

screenpipe chat prompts — just paste & go

works best if you've connected some integrations in Settings → Connections (Gmail, Google Sheets/Docs, QuickBooks, Fireflies, Otter, Notion, Obsidian, Linear, Limitless, etc.). screenpipe already knows what you've seen/said/heard — these prompts turn that raw context into something useful.

Download or update: https://screenpi.pe

📝 daily rituals

1. "look at my connections, write a morning brief for today based on yesterday's activity, my calendar, and unread emails. drop it into a Google Doc called 'morning-brief' (append a dated section). run every morning at 7am."
2. "summarize what i actually worked on today — apps used, time per project, meetings attended, people i talked to. write it as a journal entry into my obsidian daily note. run every evening at 6pm."

louis030195

schedule	*/30 8-22 * * *
enabled	false
timeout	180
icon	🎙
title	Plaud Sync
description	Pull Plaud recordings from api.plaud.ai (via the desktop app's session) and insert them as screenpipe memories

Sync Plaud voice-recorder data into screenpipe memory. Reads the bearer token from the Plaud desktop app's local storage (no credentials stored by this pipe) and calls the single endpoint the desktop client uses to enumerate the user's recordings (GET /file/). Each recording becomes a tagged memory.

louis030195

WorkTrace / Angela Jiang — Coffee Meeting Prep

1. OPEN: Build rapport (5 min)

• She's ex-OpenAI (PM #3, launched ChatGPT & GPT-4), PhD CMU, now building WorkTrace
• Genuine curiosity: "What made you leave OpenAI policy to start WorkTrace?"
• Mirror her energy — she's a builder, not a suit
• DO NOT pitch screenpipe yet. Listen first.

2. DISCOVER: Understand their world (15 min)

louis030195

Hey @everyone

v2.3.32 is here

Please update your app ⚠️

🧠 Memories — now front and center - moved from hidden settings tab to main sidebar. Create memories manually, edit tags inline, render markdown. Your AI builds a knowledge base about you, the people you meet, and every project you touch - automatically, over time.

📧 Gmail integration - draft emails from your memories and screenpipe data, get meeting notes sent to your inbox, task reminders at end of day, auto-update your CRM. Just go to connections, turn it on, and ask in the chat.

louis030195

screenpipe #09 | "What did I talk about with Sarah 3 months ago?" — your AI finally remembers everything

Hey all,

You had a conversation 3 months ago. You mentioned a project, a name, an idea. You forgot. Your AI didn't.

screenpipe now builds a living memory of every conversation, every meeting, every screen — across weeks and months. Ask it anything about your past and it answers instantly, with context.

This is the update that makes screenpipe feel like a second brain that actually works.

louis030195

Hey @everyone

v2.3 is here

Massive update. Please update your app ⚠️

🔗 Connections — 10 new OAuth integrations: Notion, Monday.com, Limitless, Brex, Stripe, Sentry, Vercel, Pipedrive, Intercom, and more. New pipes: Apple Photo sync, WisprFlow, To-Do list assistant, Notion auto CRM.

🏢 Fleet Management — screenpipe discovers instances on your network automatically. Merged fleet tab into pipes view as a device dropdown. Deploy to devices in one click.

louis030195

---

schedule: every 30m enabled: false preset:

• Oai connections: [] history: false timeout: 120 title: Team Data Sync description: Centralize screenpipe accessibility trees + audio from team members into Supabase Storage

louis030195

Screenpipe Audio Pipeline: Transcription Quality Analysis

Problem

Transcriptions are cut in the middle, words get mangled at boundaries (e.g. "screenpipe" → "screen type"), and running the same Whisper model directly on the saved audio files gives significantly better results.

GitHub issue: screenpipe/screenpipe#2345

Root Cause

louis030195

Hey all,

Louis here.

161 commits this week. New connections system, redesigned onboarding, work hours schedule, fleet management, enterprise features, and a ton of stability fixes.

---

🔗 Connections — plug screenpipe into your tools