⚡ v0.2.0 · 5 providers · cloud + local

Your meetings,
without the noise.

Drop in a transcript. Get back decisions, tasks, and blockers — with owners, confidence scores, and drift tracking. Works with the AI you already use, or runs fully offline.

works with Gemini Claude GPT-4 Mistral Ollama

→ Get started View source

●Runs locally

●No cloud needed

●Node.js 18+

teampulse — zsh

AI Reasoning

analyze

--local

batch

what got decided this week?

thinking…

scanning 4 transcripts Mon–Fri

filtering decisions, confidence ≥ 0.82

matching owners and deadlines

tool analyze_transcripts(since="monday", type="decisions")

Found

3 decisions across 4 meetings:
› Mon — Migrate auth to Clerk (→ @sara)
› Wed — Freeze scope for v1.2 (→ @team)
› Fri — Deploy to staging EOD (→ @carlos)

❯ teampulse analyze standup-2026-05-23.txt --provider claude

analyzing with claude-3-5-sonnet…

✓ 3 decisions · 5 tasks · 2 risks

DECISIONS

───────────────────────────────────

› Migrate auth to Clerk [sara] [high]

› Freeze scope for v1.2 [team] [confirmed]

TASKS

───────────────────────────────────

› Update onboarding copy [carlos] [due fri]

› Write migration script [sara] [due mon]

RISKS

───────────────────────────────────

› API rate limits on free tier [unresolved]

❯ teampulse analyze meeting.txt --provider ollama --model mistral

running locally · mistral:latest · no API key

● offline · nothing leaves your machine

✓ 2 decisions · 4 tasks · 1 risk

AVAILABLE LOCAL MODELS

───────────────────────────────────

› ollama --model mistral / llama3 / phi3 / gemma2

› mistral API · mistral-large / mixtral-8x7b

tip: ollama pull mistral to get started

❯ teampulse batch ./meetings/ --since 2026-05-01 --filter risk

processing 7 files…

✓ sprint-planning.txt

✓ design-review.txt

✓ backend-sync.txt

···

RISKS · 7 meetings

───────────────────────────────────

› API rate limits [3×]

› Auth migration timeline [2×]

› Staging env unstable [2×]

Features

Less noise, more signal

The stuff that usually falls through the cracks — pulled out automatically, every time.

🎯

Decisions

Every decision tagged with owner and confidence. No more "wait, what did we agree on?"

✅

Tasks

Action items extracted from context, not just explicit assignments. Deadlines flagged when missing.

⚠️

Risks & blockers

Surfaces problems early — even when nobody in the meeting called them out directly.

📈

Drift tracking

Compares meetings over time. Spots the same topic coming back, unresolved.

🏠

Fully local

Run everything offline with Ollama. No API key, no usage limits, no data leaving your machine.

📦

Batch mode

Process a whole month of meetings at once with --since, --filter, and --title.

AI Providers

Pick a provider. Keep moving.

One flag to switch. No config changes, no rewrite. Cloud when you want speed, local when you need privacy.

✨ Google Gemini Default

Fast and reliable for long transcripts. Gemini 2.0 Flash handles a full hour of meeting in one shot.

gemini-2.0-flash gemini-1.5-pro gemini-1.5-flash

🤖 Anthropic Claude

Better reasoning on complex meetings with lots of speakers, back-and-forth, and implicit context.

claude-3-5-sonnet claude-3-haiku claude-3-opus

⚡ OpenAI GPT-4

Consistent, well-structured output. GPT-4o is solid if you're already in the OpenAI ecosystem.

gpt-4o gpt-4o-mini gpt-4-turbo

🌪️ Mistral AI Local-first

Open-weight and fast. Run via the Mistral API, or use any Mistral model through Ollama without a key.

mistral-large mixtral-8x7b mistral-7b

🦙 Ollama 100% Offline

Runs entirely on your hardware. No API key, no internet, no data leaving the machine. Ideal for anything sensitive.

mistral llama3 phi3 gemma2 qwen2

CLI Reference

Built for real workflows

Pipe it, script it, automate it. Everything you need, nothing you don't.

teampulse analyzesingle file

Analyze one transcript. Prints decisions, tasks, and risks.

❯ teampulse analyze meeting.txt

--providerAI model

Switch provider on the fly — no config edits. gemini · claude · openai · mistral · ollama.

❯ teampulse analyze m.txt --provider mistral

--provider ollamaoffline

Fully local. No key needed. Use --model to pick which Ollama model to use.

❯ teampulse analyze m.txt --provider ollama --model llama3

--watchlive

Monitor a folder. Analyzes each new transcript as it appears.

❯ teampulse analyze --watch ./meetings/

batch + --sincedate filter

Process everything since a date. Useful for weekly or monthly reviews.

❯ teampulse batch ./ --since 2026-05-01

--filtertype

Narrow down to one insight type: risk · task · decision.

❯ teampulse batch ./ --filter risk

App Demo

See what comes out

Preview of the upcoming Chrome Extension — the same engine, now with a visual layer on top of your meetings.

TeamPulse — Meeting Dashboard

Insights

Tasks

Drift

Sprint Planning — May 22

47 min · 5 people · ollama/mistral

84Score

Migrate auth to Clerk before v1.2 launch

@sara · high confidence

Update onboarding copy and illustrations

@carlos · due Friday

API rate limits may block the testing pipeline

unresolved · flagged

Staging is down — QA blocked until Thursday

@devops · critical

Write the auth migration script

@sara · due Monday

QA staging once it's back up

@team · due Thursday

Deploy to production Friday EOD

@carlos · due Friday

API rate limits — came up in 3 meetings in a row

drift detected · still open

Auth migration re-discussed — owner changed since last sprint

2 meetings · possible confusion

Multi-provider

Switch without touching config

Just pass --provider and go. Works the same whether you're using Claude or running Ollama locally.

❯ teampulse analyze m.txt --provider ollama --model mistral

🏠 Offline mode

Nothing leaves your machine

With Ollama, all processing happens on your hardware. Good for sensitive discussions, air-gapped environments, or just not wanting to pay for API calls.

❯ ollama pull mistral

Privacy

Your transcripts stay yours

No telemetry, no cloud sync, no data stored anywhere. You run the model, you own the output.

Installation

Up in two minutes

Node.js 18+ is all you need to start. Ollama is optional, but worth it.

🍎macOS & Linux

Check Node.js

You need v18 or higher.

node --version

Clone & install

Get the repo and install dependencies.

git clone https://github.com/mcontrerasmalpar-pixel/teampulse.git

cd teampulse && npm install

Add your API key

Copy the env file and fill in whichever key you want to use.

cp .env.example .env

Set GEMINI_API_KEY, ANTHROPIC_API_KEY, OPENAI_API_KEY, or MISTRAL_API_KEY

Optional: Ollama

If you want to run fully offline, no API key needed.

brew install ollama

ollama pull mistral

Once pulled, works completely offline

Link and run

Make it available globally and run your first analysis.

npm link

teampulse analyze data/sample.txt

🪟Windows (PowerShell)

Check Node.js

Open PowerShell and verify your version.

node --version

Clone & install

Get the repo and install dependencies.

git clone https://github.com/mcontrerasmalpar-pixel/teampulse.git

cd teampulse; npm install

Add your API key

Copy the env file and fill in your key.

Copy-Item .env.example .env

Set GEMINI_API_KEY, ANTHROPIC_API_KEY, OPENAI_API_KEY, or MISTRAL_API_KEY

Optional: Ollama

Download from ollama.com and pull a model.

ollama pull mistral

Once pulled, works completely offline

Link and run

Make it available globally and run your first analysis.

npm link

teampulse analyze data/sample.txt

Your meetings,without the noise.

Less noise, more signal

Decisions

Tasks

Risks & blockers

Drift tracking

Fully local

Batch mode

Pick a provider. Keep moving.

Built for real workflows

See what comes out

Up in two minutes

Less re-discussing.More shipping.

Your meetings,
without the noise.

Less re-discussing.
More shipping.