freyWylfred/minutes

C#

0

3 commits

updated Aug 5, 2026

See the code

README

minutes

A Windows desktop app that turns Microsoft Teams meeting transcripts into clean, structured meeting minutes — fully offline, powered by a local LLM.

Features

  • Teams transcript import — Open transcript files exported from Teams (WebVTT .vtt or plain .txt), or simply paste raw transcript text.
  • Automatic transcript cleanup — Removes WebVTT headers, cue identifiers, timestamps, and HTML tags; merges consecutive utterances by the same speaker into a readable Speaker: text format. Handles both half-width (:) and full-width (:) colons used in Japanese Teams exports.
  • AI summarization & organization — A local LLM (via LLamaSharp) analyzes the transcript and produces structured minutes in Japanese with the following sections:
    • Meeting overview
    • Participants and key remarks
    • Decisions
    • Action items (with owners and due dates when available)
    • Open issues / carry-overs to the next meeting
  • Fully local & private — No transcript data ever leaves your machine. The model is downloaded once on first launch, then everything runs offline.
  • Streaming output — Summary tokens are streamed to the UI as they are generated.

Requirements

  • Windows 10 version 1809 (build 17763) or later
  • .NET 8 SDK
  • Visual Studio 2022 or later with the Windows App SDK / WinUI 3 workload

Getting Started

  1. Clone the repository:

    git clone https://github.com/freyWylfred/minutes.git
    
  2. Open minutes.slnx in Visual Studio.

  3. Set the packaging project (minutes (Package)) as the startup project and run.

  4. On first launch the app downloads the LLM model automatically — wait for the progress bar to complete.

  5. Click トランスクリプトを開く (Open transcript) to load a .vtt / .txt file exported from Teams, or paste the transcript text directly.

  6. Click 要約・整理を生成 (Generate summary) and the structured minutes appear in the output pane.

How It Works

Teams transcript (.vtt / .txt)
		│
		▼
TranscriptParserService   ── strips timestamps & noise, normalizes to "Speaker: text"
		│
		▼
LlmService (LLamaSharp)   ── local LLM inference with a Teams-minutes-specialized prompt
		│
		▼
Structured minutes (overview / decisions / action items / open issues)

Project Structure

PathDescription
minutes/minutes/MainWindow.xaml(.cs)WinUI 3 main window: file picking, input/output UI
minutes/minutes/Services/TranscriptParserService.csWebVTT / Teams text transcript parser & cleaner
minutes/minutes/Services/LlmService.csLocal LLM summarization via LLamaSharp
minutes/minutes/Services/ModelDownloadService.csFirst-run model download
minutes/minutes (Package)MSIX packaging project

License

This project is provided as-is. See the repository for details.

freyWylfred/minutes

C#

0

3 commits

updated Aug 5, 2026

See the code

README

minutes

A Windows desktop app that turns Microsoft Teams meeting transcripts into clean, structured meeting minutes — fully offline, powered by a local LLM.

Features

  • Teams transcript import — Open transcript files exported from Teams (WebVTT .vtt or plain .txt), or simply paste raw transcript text.
  • Automatic transcript cleanup — Removes WebVTT headers, cue identifiers, timestamps, and HTML tags; merges consecutive utterances by the same speaker into a readable Speaker: text format. Handles both half-width (:) and full-width (:) colons used in Japanese Teams exports.
  • AI summarization & organization — A local LLM (via LLamaSharp) analyzes the transcript and produces structured minutes in Japanese with the following sections:
    • Meeting overview
    • Participants and key remarks
    • Decisions
    • Action items (with owners and due dates when available)
    • Open issues / carry-overs to the next meeting
  • Fully local & private — No transcript data ever leaves your machine. The model is downloaded once on first launch, then everything runs offline.
  • Streaming output — Summary tokens are streamed to the UI as they are generated.

Requirements

  • Windows 10 version 1809 (build 17763) or later
  • .NET 8 SDK
  • Visual Studio 2022 or later with the Windows App SDK / WinUI 3 workload

Getting Started

  1. Clone the repository:

    git clone https://github.com/freyWylfred/minutes.git
    
  2. Open minutes.slnx in Visual Studio.

  3. Set the packaging project (minutes (Package)) as the startup project and run.

  4. On first launch the app downloads the LLM model automatically — wait for the progress bar to complete.

  5. Click トランスクリプトを開く (Open transcript) to load a .vtt / .txt file exported from Teams, or paste the transcript text directly.

  6. Click 要約・整理を生成 (Generate summary) and the structured minutes appear in the output pane.

How It Works

Teams transcript (.vtt / .txt)
		│
		▼
TranscriptParserService   ── strips timestamps & noise, normalizes to "Speaker: text"
		│
		▼
LlmService (LLamaSharp)   ── local LLM inference with a Teams-minutes-specialized prompt
		│
		▼
Structured minutes (overview / decisions / action items / open issues)

Project Structure

PathDescription
minutes/minutes/MainWindow.xaml(.cs)WinUI 3 main window: file picking, input/output UI
minutes/minutes/Services/TranscriptParserService.csWebVTT / Teams text transcript parser & cleaner
minutes/minutes/Services/LlmService.csLocal LLM summarization via LLamaSharp
minutes/minutes/Services/ModelDownloadService.csFirst-run model download
minutes/minutes (Package)MSIX packaging project

License

This project is provided as-is. See the repository for details.

Languages

C#

100.0%