DevMind Pro

Real-time AI meeting assistant — transcription, contradiction detection, RAG-powered replies, Teams integration.

---

Quick Install

pip install devmind-pro
devmind setup      # interactive wizard — configures everything
devmind start      # launch the assistant

---

What it does

Feature	Description
🎙️ Live transcription	Whisper-powered, works offline
🔍 Contradiction detection	Flags when meeting discussion contradicts your docs
💬 AI reply suggestions	Suggests answers to questions directed at you
📱 Teams integration	Catches incoming Teams messages, suggests replies
📋 Commitments tracker	Tracks "I will..." promises made in the meeting
🎯 Pre-meeting briefing	Generates briefing from past summaries + docs
📝 Auto summary	Saves meeting summary as markdown + JSON

---

Requirements

Component	Version
Python	3.11.x (not 3.12+)
macOS	Ventura 13+ (Apple Silicon M1/M2/M3)
Ollama	Latest
RAM	16 GB recommended

Windows / Linux: Core functionality works. Replace brew commands with your package manager.

---

Full Setup Guide

Step 1 — System dependencies

macOS:

brew install portaudio python@3.11
# Install Ollama from https://ollama.com/download

Windows:

# Python 3.11: https://www.python.org/downloads/release/python-3119/
# Ollama: https://ollama.com/download
pip install pipwin && pipwin install pyaudio

Linux:

sudo apt install python3.11 python3.11-venv portaudio19-dev
curl -fsSL https://ollama.com/install.sh | sh

Step 2 — Pull AI model

ollama serve                    # start Ollama (separate terminal)
ollama pull llama3.1:8b         # 4.9GB — do this once

Step 3 — Install DevMind Pro

python3.11 -m venv venv
source venv/bin/activate        # Windows: venv\Scripts\activate

# Install torch first (prevents version conflicts)
pip install torch==2.2.2 torchaudio==2.2.2
pip install sentence-transformers==3.0.1 transformers==4.41.2 \
            tokenizers==0.19.1 huggingface-hub==0.23.4
pip install devmind-pro
pip uninstall torchvision -y    # remove if pulled in

Step 4 — Setup and run

devmind setup      # interactive wizard — creates .env, checks everything
devmind check      # verify all dependencies
devmind start      # launch!

---

CLI Commands

devmind setup              # first-time interactive setup wizard
devmind start              # launch the meeting assistant
devmind start --env /path/to/.env   # use specific config file
devmind check              # pre-flight dependency check
devmind config             # show current configuration
devmind --version          # show version

---

AI Provider Options

Change AI_PROVIDER in .env to switch:

Provider	Value	Notes
Ollama (default)	ollama	Free, local, private
OpenAI	openai	Fast, needs API key
Gemini	gemini	Free tier available
Groq	groq	Very fast, free tier

# Groq (fastest free option)
AI_PROVIDER=groq
GROQ_API_KEY=your_key
GROQ_MODEL=llama3-70b-8192

# OpenAI
AI_PROVIDER=openai
OPENAI_API_KEY=your_key
OPENAI_MODEL=gpt-4o-mini

# Any Ollama model
OLLAMA_MODEL=llama3.2:1b    # smallest/coolest
OLLAMA_MODEL=phi3:mini       # Microsoft efficient model
OLLAMA_MODEL=mistral:7b      # fast

---

Teams Integration

DevMind reads incoming Teams notifications automatically (macOS).

To enable sending replies to teammates, add their emails to .env:

TEAMS_CONTACTS=Rajiv:rajiv@company.com,Arif:arif@company.com

Clicking "Send to Rajiv" opens Teams in their chat with the message copied to clipboard — just press Cmd+V + Enter.

---

Whisper Models

Model	Size	Speed	Use for
tiny	75MB	Fastest	Testing
small	250MB	Fast	Daily use (default)
medium	750MB	Medium	Accented speech
large	1.5GB	Slow	Critical meetings

WHISPER_MODEL=small
WHISPER_LANG=en

---

Project Docs (RAG)

Put your project documents in ./docs/. Supported: .md, .txt, .pdf, .docx

DevMind uses these for:

• Contradiction detection (flags when meeting discussion conflicts with docs)
• Context-aware reply suggestions
• Pre-meeting briefings

---

SDK Usage

import devmind_pro as dm

dm.configure(
    provider="ollama",
    model="llama3.1:8b",
    docs_folder="./my_docs",
    my_name="Jai",
)
dm.start()

---

Troubleshooting

404 Not Found on Ollama

ollama serve          # start Ollama
ollama list           # check model name matches OLLAMA_MODEL in .env

Cannot copy out of meta tensor

pip install sentence-transformers==3.0.1 transformers==4.41.2 \
            tokenizers==0.19.1 huggingface-hub==0.23.4 --force-reinstall
pip uninstall torchvision -y

Garbage transcription — add to .env:

WHISPER_LANG=en
TOKENIZERS_PARALLELISM=false

Teams messages not appearing — ensure macOS permission: System Settings → Privacy & Security → Automation → Terminal → System Events ✓ systen Settings - privacy & security _ all discs - terminal or vs code

---

License

MIT