jameshtwose/ledex-demo
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

adding the first version of the presentation and the initial attempt at the .meetings script

Browse Source
This commit is contained in:
itsamejms
2026-06-07 16:13:10 +01:00
parent 41f99557cc
commit 05c7d6aef3
34 changed files with 7623 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files
meetings-cli/2026-06-07_1302_Note/action_items.md
+33
View File
@@ -0,0 +1,33 @@
Here are the extracted action items, tasks, and commitments:
1. **Who**: Unassigned
**What**: Review and revise meeting notes to ensure accuracy and completeness.
**When**: No deadline specified
**Priority**: Low
2. **Who**: Alaz
**What**: Follow up with Snorri regarding his dimensional teapot and its contents.
**When**: No deadline specified
**Priority**: Medium
3. **Who**: Tuxi
**What**: Provide Alaz with more information about the dimensional teapot's prize.
**When**: No deadline specified
**Priority**: Low
4. **Who**: Unassigned
**What**: Research and provide answers to Snorri's questions regarding the family portraits in his lounge.
**When**: No deadline specified
**Priority**: Medium
5. **Who**: Alaz
**What**: Investigate the origins of the jar image found in Snorri's lounge.
**When**: No deadline specified
**Priority**: Low
6. **Who**: Unassigned
**What**: Clarify and resolve any outstanding questions or issues from the meeting.
**When**: No deadline specified
**Priority**: High
Note: The conversation at the end of the transcript appears to be a discussion about whether anything was missed, but no specific action items were assigned.
meetings-cli/2026-06-07_1302_Note/report.md
+75
View File
File diff suppressed because one or more lines are too long
meetings-cli/2026-06-07_1302_Note/summary.md
+20
View File
@@ -0,0 +1,20 @@
**Topic**: **Campaign Session Recap**
**Key Points**:
* The group participated in a combat encounter with the Beatles, using creative methods to defeat them.
* A "beat-o" was defeated by Snorri using an apple, granting him a ghostly speed ability for a week.
* The group prepared and presented three dishes: Tuxi's colorful boom-boom decoration, Snorri's sashimi, and Alaz's flamed dish.
* For dessert, the group went on a foraging task, with Tuxy preparing lychee surprise and Snorri making a lingon berry jelly.
* Alaz visited a village of awakened apes and obtained ingredients to make a rice pudding.
**Decisions**:
* The competition was decided, with Alaz winning a dimensional teapot and Snorri coming second with an arcane Rubik's gift.
* Snorri exchanged prizes with Alaz, obtaining the teapot and rubbing it three times to teleport into a different dimension.
* Snorri obtained an image of a jar from the new dimension.
**Open Questions**:
* What is the significance of the family portraits in the new dimension?
* How will the group's experience in this new dimension affect their future gameplay?
meetings-cli/2026-06-07_1302_Note/transcript.txt
+2
View File
File diff suppressed because one or more lines are too long
meetings-cli/README.md
+178
View File
@@ -0,0 +1,178 @@
# 🎤 meetings — local audio → transcript → summary + action items
A single-file shell script CLI that transcribes meeting recordings (using GGUF Whisper or Parakeet models), then generates a summary and extracts action items using Ollama. Everything runs 100% locally.
```
audio file ──▶ ffmpeg ──▶ whisper.cpp / parakeet.cpp ──▶ Ollama ──▶ report.md
│ │ │
16kHz mono WAV GGUF transcription summary + actions
```
## Quick start
```bash
# 1. Clone/download
git clone <this-repo> meetings-cli && cd meetings-cli
# 2. One-time setup (installs whisper.cpp, downloads model, pulls Ollama LLM)
./meetings setup
# 3. Process a meeting recording
./meetings recording.mp3
# 4. Check everything is healthy
./meetings doctor
```
## What it does
| Step | Tool | What happens |
|------|------|--------------|
| 1. Convert | ffmpeg | Any audio → 16kHz mono WAV |
| 2. Transcribe | whisper.cpp or parakeet.cpp | GGUF/GGML model → text transcript |
| 3. Summarize | Ollama | Transcript → structured summary (topic, key points, decisions, open questions) |
| 4. Extract | Ollama | Transcript → numbered action items (who, what, when, priority) |
## Output
For each audio file, a directory is created containing:
```
2026-06-07_1402_team_standup/
├── report.md # Combined: summary + actions + transcript
├── transcript.txt # Raw transcription
├── summary.md # LLM-generated summary
└── action_items.md # Extracted action items
```
## Requirements
| Dependency | Install | Purpose |
|------------|---------|---------|
| **ffmpeg** | `brew install ffmpeg` | Audio format conversion |
| **whisper.cpp** | `brew install whisper-cpp` | Speech-to-text (GGML models) |
| **Ollama** | [ollama.com](https://ollama.com) | LLM for summarization |
| **jq** | `brew install jq` | JSON parsing for Ollama API |
> `./meetings setup` handles all of this automatically.
## STT engines
### whisper.cpp (default, recommended)
- Battle-tested, many languages, large model ecosystem
- Models from [ggerganov/whisper.cpp](https://huggingface.co/ggerganov/whisper.cpp)
- Install: `brew install whisper-cpp`
| Model | Size | Best for |
|-------|------|----------|
| tiny.en | 75 MB | Quick tests, English |
| base.en | 142 MB | Good balance, English |
| small.en | 466 MB | **Recommended for English** |
| medium.en | 1.5 GB | High accuracy, English |
| large-v3-turbo | 809 MB | Best multilingual, fast |
| large-v3 | 2.9 GB | Best accuracy, any language |
### parakeet.cpp (alternative, faster)
- NVIDIA Parakeet models, excellent English, smaller footprint
- Models from [mudler/parakeet-cpp-gguf](https://huggingface.co/mudler/parakeet-cpp-gguf)
- Install: Build from [source](https://github.com/mudler/parakeet.cpp) or use Docker
| Model | Size | Best for |
|-------|------|----------|
| tdt_ctc-110m-q8_0 | 178 MB | Fast, good English |
| tdt_ctc-110m-f16 | 268 MB | Fast, lossless English |
| tdt-0.6b-v3-f16 | 1.4 GB | Multilingual |
## Configuration
### Environment variables
```bash
MEETINGS_DIR # Config & models directory (default: ~/.meetings)
MEETINGS_STT # STT engine: whisper | parakeet
MEETINGS_STT_MODEL # Path to GGUF/GGML model file
MEETINGS_LLM # Ollama model for summarization (default: llama3.1:8b)
MEETINGS_THREADS # Thread count for STT (default: 4)
MEETINGS_LANG # Language code (default: en; use "auto" for multilingual)
MEETINGS_OUTPUT # Output directory (default: .)
```
### CLI flags
```bash
./meetings recording.mp3 --stt whisper --llm llama3.1:8b --lang en --output ./reports
```
### Config file
Saved at `~/.meetings/config` after running `./meetings setup`:
```
STT_ENGINE=whisper
OLLAMA_MODEL=llama3.1:8b
THREADS=4
STT_MODEL=/home/user/.meetings/models/ggml-small.en.bin
```
## Commands
```bash
./meetings <audio_file> # Run the full pipeline
./meetings setup # Install deps + download model (interactive)
./meetings doctor # Check all dependencies
./meetings config # Show current configuration
./meetings help # Show help
```
## How Ollama fits in
**Ollama does NOT run the whisper/parakeet models** — those use their own inference engines (whisper.cpp / parakeet.cpp). Ollama is only used for the LLM steps:
1. **Summary generation** — sends the transcript to an Ollama model with a structured summarization prompt
2. **Action item extraction** — sends the transcript to an Ollama model with an action-item extraction prompt
You can use any Ollama model. Smaller models (llama3.2:1b, gemma3:1b) are faster; larger models (llama3.1:8b, qwen2.5-coder:7b) produce better summaries.
## Example
```bash
$ ./meetings team_standup.m4a
┌─────────────────────────────────────────────┐
│ 🎤 M E E T I N G S │
│ audio → transcript → summary + actions │
│ whisper.cpp · parakeet.cpp · ollama │
└─────────────────────────────────────────────┘
Input: team_standup.m4a
STT engine: whisper
STT model: ggml-small.en.bin
LLM model: llama3.1:8b
Language: en
Output: ./2026-06-07_1402_team_standup/
── Step 1/4 — Converting audio ──
▸ Converting audio to 16kHz mono WAV...
✓ Audio converted: 1.2M
── Step 2/4 — Transcribing with whisper ──
▸ Transcribing with whisper.cpp...
✓ Transcript: 847 words
── Step 3/4 — Summarizing (llama3.1:8b) ──
✓ Summary saved
── Step 4/4 — Extracting action items (llama3.1:8b) ──
✓ Action items saved
✓ All done! Files saved to: ./2026-06-07_1402_team_standup/
📄 Report: ./2026-06-07_1402_team_standup/report.md
📝 Transcript: ./2026-06-07_1402_team_standup/transcript.txt
📋 Summary: ./2026-06-07_1402_team_standup/summary.md
✅ Action Items: ./2026-06-07_1402_item_standup/action_items.md
```
## License
MIT
meetings-cli/meetings
Executable
+659
View File
@@ -0,0 +1,659 @@
#!/usr/bin/env bash
#
# meetings — audio → transcript → summary + action items, all local
#
# Transcribes audio using whisper.cpp or parakeet.cpp (GGUF/GGML models),
# then uses Ollama to summarize and extract action items.
#
# Usage:
# ./meetings <audio_file> # full pipeline
# ./meetings setup # install deps + download model
# ./meetings doctor # check dependencies
# ./meetings config # show current config
#
set -euo pipefail
# ─── colours ────────────────────────────────────────────────────────
RED='\033[0;31m'; GRN='\033[0;32m'; YEL='\033[1;33m'
BLU='\033[0;34m'; CYN='\033[0;36m'; RST='\033[0m'
BOLD='\033[1m'
# ─── paths ──────────────────────────────────────────────────────────
MEETINGS_DIR="${MEETINGS_DIR:-$HOME/.meetings}"
CONFIG_FILE="$MEETINGS_DIR/config"
MODELS_DIR="$MEETINGS_DIR/models"
# ─── helpers ───────────────────────────────────────────────────────
die() { printf "${RED}error: %s${RST}\n" "$*" >&2; exit 1; }
info() { printf "${BLU}▸ %s${RST}\n" "$*" >&2; }
ok() { printf "${GRN}✓ %s${RST}\n" "$*" >&2; }
warn() { printf "${YEL}⚠ %s${RST}\n" "$*" >&2; }
step() { printf "\n${BOLD}${CYN}── %s ──${RST}\n" "$*" >&2; }
banner() {
cat <<'BAN'
┌─────────────────────────────────────────────┐
│ 🎤 M E E T I N G S │
│ audio → transcript → summary + actions │
│ whisper.cpp · parakeet.cpp · ollama │
└─────────────────────────────────────────────┘
BAN
}
# ─── load config: env vars > config file > defaults ────────────────
load_config() {
# defaults
STT_ENGINE=""
STT_MODEL=""
OLLAMA_MODEL="llama3.1:8b"
OLLAMA_HOST="http://localhost:11434"
THREADS="4"
LANGUAGE="en"
OUTPUT_DIR="."
# config file overrides defaults
[[ -f "$CONFIG_FILE" ]] && source "$CONFIG_FILE"
# env vars override everything (|| true prevents set -e exit on empty vars)
[[ -n "${MEETINGS_STT:-}" ]] && STT_ENGINE="$MEETINGS_STT" || true
[[ -n "${MEETINGS_STT_MODEL:-}" ]] && STT_MODEL="$MEETINGS_STT_MODEL" || true
[[ -n "${MEETINGS_LLM:-}" ]] && OLLAMA_MODEL="$MEETINGS_LLM" || true
[[ -n "${MEETINGS_THREADS:-}" ]] && THREADS="$MEETINGS_THREADS" || true
[[ -n "${MEETINGS_LANG:-}" ]] && LANGUAGE="$MEETINGS_LANG" || true
[[ -n "${MEETINGS_OUTPUT:-}" ]] && OUTPUT_DIR="$MEETINGS_OUTPUT" || true
}
# ─── detect STT engine ──────────────────────────────────────────────
detect_stt() {
if command -v parakeet-cli &>/dev/null; then
echo "parakeet"
elif command -v whisper-cli &>/dev/null; then
echo "whisper"
else
echo ""
fi
}
get_stt_binary() {
local engine="${1:-$STT_ENGINE}"
case "$engine" in
whisper) command -v whisper-cli 2>/dev/null || echo "" ;;
parakeet) command -v parakeet-cli 2>/dev/null || echo "" ;;
*) echo "" ;;
esac
}
find_stt_model() {
local engine="${1:-whisper}"
if [[ -n "${STT_MODEL:-}" && -f "${STT_MODEL}" ]]; then
echo "$STT_MODEL"; return 0
fi
case "$engine" in
whisper)
for f in "$MODELS_DIR"/ggml-*.bin "$MODELS_DIR"/ggml-*.gguf; do
[[ -f "$f" ]] && echo "$f" && return 0
done
;;
parakeet)
for f in "$MODELS_DIR"/*parakeet*.gguf "$MODELS_DIR"/*tdt*.gguf; do
[[ -f "$f" ]] && echo "$f" && return 0
done
;;
esac
return 1
}
# ─── doctor ─────────────────────────────────────────────────────────
cmd_doctor() {
banner >&2
local ok_count=0 total=0
total=$((total+1))
if command -v ffmpeg &>/dev/null; then
ok "ffmpeg: $(command -v ffmpeg)"; ok_count=$((ok_count+1))
else
warn "ffmpeg: not found"
fi
total=$((total+1))
if command -v ollama &>/dev/null; then
ok "ollama: $(command -v ollama)"; ok_count=$((ok_count+1))
if curl -sf "${OLLAMA_HOST:-http://localhost:11434}/api/tags" &>/dev/null; then
ok " server: running at ${OLLAMA_HOST:-http://localhost:11434}"
else
warn " server: not responding (run: ollama serve)"
fi
else
warn "ollama: not found"
fi
total=$((total+1))
local engine="${STT_ENGINE:-$(detect_stt)}"
local bin
bin="$(get_stt_binary "$engine")"
if [[ -n "$bin" ]]; then
ok "STT engine: $engine ($bin)"; ok_count=$((ok_count+1))
else
warn "STT engine: not found (run: ./meetings setup)"
fi
total=$((total+1))
local model
model="$(find_stt_model "$engine")" || true
if [[ -n "$model" ]]; then
ok "STT model: $model"; ok_count=$((ok_count+1))
else
warn "STT model: not found (run: ./meetings setup)"
fi
total=$((total+1))
local ollama_model="${OLLAMA_MODEL:-llama3.1:8b}"
local model_name
model_name="$(echo "$ollama_model" | cut -d: -f1)"
if ollama list 2>/dev/null | awk '{print $1}' | grep -qF "$model_name"; then
ok "LLM model: $ollama_model (pulled)"; ok_count=$((ok_count+1))
else
warn "LLM model: $ollama_model (not pulled — run: ollama pull $ollama_model)"
fi
echo "" >&2
if [[ $ok_count -eq $total ]]; then
ok "All good ($ok_count/$total)"
else
warn "Ready ($ok_count/$total). Run './meetings setup' to install missing pieces."
fi
}
# ─── setup ──────────────────────────────────────────────────────────
cmd_setup() {
load_config
banner >&2
mkdir -p "$MEETINGS_DIR" "$MODELS_DIR"
# ── ffmpeg ──
step "Checking ffmpeg"
if command -v ffmpeg &>/dev/null; then
ok "ffmpeg already installed"
else
info "Installing ffmpeg..."
if [[ "$(uname)" == "Darwin" ]]; then
brew install ffmpeg || die "Could not install ffmpeg"
elif command -v apt-get &>/dev/null; then
sudo apt-get update && sudo apt-get install -y ffmpeg
elif command -v dnf &>/dev/null; then
sudo dnf install -y ffmpeg
else
die "Please install ffmpeg manually: https://ffmpeg.org/download.html"
fi
fi
# ── ollama ──
step "Checking Ollama"
if command -v ollama &>/dev/null; then
ok "ollama already installed"
else
info "Installing Ollama..."
curl -fsSL https://ollama.com/install.sh | sh || die "Could not install Ollama"
fi
if ! curl -sf "${OLLAMA_HOST:-http://localhost:11434}/api/tags" &>/dev/null; then
info "Starting Ollama server..."
ollama serve &>/dev/null &
sleep 3
fi
# ── choose STT engine ──
step "Choosing STT engine"
echo "" >&2
echo " 1) whisper.cpp — battle-tested, many languages, brew install" >&2
echo " 2) parakeet.cpp — faster, great English, smaller footprint" >&2
echo "" >&2
local choice
if [[ -n "${STT_ENGINE:-}" ]]; then
choice="$STT_ENGINE"
info "Using pre-configured engine: $choice"
else
read -rp " Choose [1/2, default=1]: " choice
case "${choice:-1}" in
2|parakeet) choice="parakeet" ;;
*) choice="whisper" ;;
esac
fi
case "$choice" in
whisper) setup_whisper ;;
parakeet) setup_parakeet ;;
esac
STT_ENGINE="$choice"
grep -q "^STT_ENGINE=" "$CONFIG_FILE" 2>/dev/null \
&& sed -i '' "s/^STT_ENGINE=.*/STT_ENGINE=$choice/" "$CONFIG_FILE" 2>/dev/null \
|| echo "STT_ENGINE=$choice" >> "$CONFIG_FILE"
# ── pull LLM ──
step "Pulling LLM: ${OLLAMA_MODEL:-llama3.1:8b}"
local ollama_model="${OLLAMA_MODEL:-llama3.1:8b}"
if ollama list 2>/dev/null | awk '{print $1}' | grep -qF "$(echo "$ollama_model" | cut -d: -f1)"; then
ok "$ollama_model already pulled"
else
ollama pull "$ollama_model" || warn "Could not pull $ollama_model. Run: ollama pull $ollama_model"
fi
echo "" >&2
ok "Setup complete! Run './meetings doctor' to verify, then './meetings <audio_file>' to process."
}
setup_whisper() {
step "Installing whisper.cpp"
if command -v whisper-cli &>/dev/null; then
ok "whisper-cli already available at $(command -v whisper-cli)"
else
info "Installing via package manager..."
if [[ "$(uname)" == "Darwin" ]]; then
brew install whisper-cpp || die "Could not install whisper-cpp via brew"
else
die "Please install whisper-cli manually: https://github.com/ggml-org/whisper.cpp
On macOS: brew install whisper-cpp
Or build from source: git clone https://github.com/ggml-org/whisper.cpp && cd whisper.cpp && cmake -B build && cmake --build build -j"
fi
fi
step "Downloading Whisper model"
local model_choice
echo "" >&2
echo " Available models (from HuggingFace ggerganov/whisper.cpp):" >&2
echo " tiny.en — 75 MB (fastest, English only)" >&2
echo " base.en — 142 MB (good for quick tests)" >&2
echo " small.en — 466 MB (recommended for English) ★" >&2
echo " medium.en — 1.5 GB (high accuracy, English only)" >&2
echo " large-v3-turbo — 809 MB (best multilingual, fast)" >&2
echo " large-v3 — 2.9 GB (best accuracy, any language)" >&2
echo "" >&2
if [[ -n "${STT_MODEL:-}" && -f "${STT_MODEL}" ]]; then
info "Using existing model: $STT_MODEL"
else
read -rp " Choose model [default=small.en]: " model_choice
model_choice="${model_choice:-small.en}"
local model_file="$MODELS_DIR/ggml-${model_choice}.bin"
if [[ -f "$model_file" ]]; then
ok "Model already downloaded: $model_file"
else
info "Downloading ggml-${model_choice}.bin (this may take a moment)..."
curl -L --progress-bar \
-o "$model_file" \
"https://huggingface.co/ggerganov/whisper.cpp/resolve/main/ggml-${model_choice}.bin" \
|| die "Failed to download model"
ok "Downloaded: $model_file"
fi
STT_MODEL="$model_file"
echo "STT_MODEL=$model_file" >> "$CONFIG_FILE"
fi
}
setup_parakeet() {
step "Installing parakeet.cpp"
if command -v parakeet-cli &>/dev/null; then
ok "parakeet-cli already available at $(command -v parakeet-cli)"
else
info "Checking for pre-built binary or Docker..."
if command -v docker &>/dev/null; then
ok "Docker available — parakeet.cpp will run via container"
else
die "parakeet-cli not found and Docker not installed. Please either:
- Build parakeet.cpp: https://github.com/mudler/parakeet.cpp
- Or install Docker for the container approach"
fi
fi
step "Downloading Parakeet model"
local model_choice
echo "" >&2
echo " Available models (from HuggingFace mudler/parakeet-cpp-gguf):" >&2
echo " tdt_ctc-110m-q8_0 — 178 MB (fast, good English, smallest) ★" >&2
echo " tdt_ctc-110m-f16 — 268 MB (fast, lossless English)" >&2
echo " tdt-0.6b-v3-f16 — 1.4 GB (multilingual, recommended)" >&2
echo "" >&2
if [[ -n "${STT_MODEL:-}" && -f "${STT_MODEL}" ]]; then
info "Using existing model: $STT_MODEL"
else
read -rp " Choose model [default=tdt_ctc-110m-q8_0]: " model_choice
model_choice="${model_choice:-tdt_ctc-110m-q8_0}"
local model_file="$MODELS_DIR/${model_choice}.gguf"
if [[ -f "$model_file" ]]; then
ok "Model already downloaded: $model_file"
else
info "Downloading ${model_choice}.gguf..."
curl -L --progress-bar \
-o "$model_file" \
"https://huggingface.co/mudler/parakeet-cpp-gguf/resolve/main/${model_choice}.gguf" \
|| die "Failed to download model"
ok "Downloaded: $model_file"
fi
STT_MODEL="$model_file"
echo "STT_MODEL=$model_file" >> "$CONFIG_FILE"
fi
}
# ─── config display ─────────────────────────────────────────────────
cmd_config() {
load_config
banner >&2
echo "" >&2
if [[ -f "$CONFIG_FILE" ]]; then
echo " Saved config ($CONFIG_FILE):" >&2
echo "" >&2
while IFS='=' read -r key val; do
printf " %-16s %s\n" "$key" "$val" >&2
done < "$CONFIG_FILE"
else
echo " (no saved config — run: ./meetings setup)" >&2
fi
echo "" >&2
echo " Effective settings:" >&2
local stt_display="${STT_ENGINE:-$(detect_stt)}"
stt_display="${stt_display:-(not set)}"
echo " STT engine: $stt_display" >&2
echo " STT model: ${STT_MODEL:-(auto-detect)}" >&2
echo " LLM model: $OLLAMA_MODEL" >&2
echo " Ollama host: $OLLAMA_HOST" >&2
echo " Threads: $THREADS" >&2
echo " Language: $LANGUAGE" >&2
echo " Output dir: $OUTPUT_DIR" >&2
echo "" >&2
}
# ─── convert audio ──────────────────────────────────────────────────
convert_audio() {
local input="$1" output="$2"
info "Converting audio to 16kHz mono WAV..."
ffmpeg -y -i "$input" -ar 16000 -ac 1 -c:a pcm_s16le "$output" \
-loglevel warning 2>/dev/null \
|| die "ffmpeg conversion failed for $input"
ok "Audio converted: $(du -h "$output" | cut -f1)"
}
# ─── transcribe with whisper ───────────────────────────────────────
transcribe_whisper() {
local wav_file="$1" model="$2" binary="$3"
info "Transcribing with whisper.cpp..."
local lang_flag=""
[[ "$LANGUAGE" != "auto" ]] && lang_flag="-l $LANGUAGE"
local transcript
transcript=$("$binary" \
-m "$model" \
-f "$wav_file" \
-t "$THREADS" \
$lang_flag \
-nt \
--no-prints \
2>/dev/null) || die "Whisper transcription failed"
echo "$transcript"
}
# ─── transcribe with parakeet ──────────────────────────────────────
transcribe_parakeet() {
local wav_file="$1" model="$2" binary="$3"
info "Transcribing with parakeet.cpp..."
local decoder_flag=""
if echo "$model" | grep -qi "ctc"; then
decoder_flag="--decoder ctc"
elif echo "$model" | grep -qi "tdt"; then
decoder_flag="--decoder tdt"
fi
if [[ -z "$binary" ]] && command -v docker &>/dev/null; then
info "Using Docker for parakeet.cpp..."
docker run --rm \
-v "$model:/model.gguf:ro" \
-v "$wav_file:/audio.wav:ro" \
ghcr.io/mudler/parakeet.cpp-cli:latest \
transcribe --model /model.gguf --input /audio.wav $decoder_flag 2>/dev/null \
|| die "Docker parakeet transcription failed"
else
"$binary" transcribe --model "$model" --input "$wav_file" $decoder_flag 2>/dev/null \
|| die "Parakeet transcription failed"
fi
}
# ─── call Ollama ───────────────────────────────────────────────────
ollama_generate() {
local prompt="$1" system="${2:-}"
local json_payload
if [[ -n "$system" ]]; then
json_payload=$(jq -n \
--arg model "$OLLAMA_MODEL" \
--arg system "$system" \
--arg prompt "$prompt" \
'{model: $model, system: $system, prompt: $prompt, stream: false}')
else
json_payload=$(jq -n \
--arg model "$OLLAMA_MODEL" \
--arg prompt "$prompt" \
'{model: $model, prompt: $prompt, stream: false}')
fi
curl -sf "$OLLAMA_HOST/api/generate" \
-d "$json_payload" 2>/dev/null \
| jq -r '.response // empty' \
|| die "Ollama request failed. Is the server running? (ollama serve)"
}
# ─── main pipeline ──────────────────────────────────────────────────
cmd_process() {
local input_file="$1"
shift || true
# Load config BEFORE flag parsing, so flags can override
load_config
while [[ $# -gt 0 ]]; do
case "$1" in
--stt) STT_ENGINE="$2"; shift 2 ;;
--model) STT_MODEL="$2"; shift 2 ;;
--llm) OLLAMA_MODEL="$2"; shift 2 ;;
--lang) LANGUAGE="$2"; shift 2 ;;
--threads) THREADS="$2"; shift 2 ;;
--output) OUTPUT_DIR="$2"; shift 2 ;;
*) die "Unknown option: $1. Run: ./meetings help" ;;
esac
done
[[ -z "$input_file" ]] && die "Usage: ./meetings <audio_file>"
[[ ! -f "$input_file" ]] && die "File not found: $input_file"
STT_ENGINE="${STT_ENGINE:-$(detect_stt)}"
[[ -z "$STT_ENGINE" ]] && die "No STT engine found. Run: ./meetings setup"
local binary
binary="$(get_stt_binary "$STT_ENGINE")"
[[ -z "$binary" ]] && die "Could not find $STT_ENGINE binary. Run: ./meetings setup"
local model
if [[ -n "${STT_MODEL:-}" ]] && [[ -f "$STT_MODEL" ]]; then
model="$STT_MODEL"
else
model="$(find_stt_model "$STT_ENGINE")" \
|| die "No STT model found. Run: ./meetings setup"
fi
[[ ! -f "$model" ]] && die "Model file not found: $model"
command -v ffmpeg &>/dev/null || die "ffmpeg not found. Install it first."
curl -sf "$OLLAMA_HOST/api/tags" &>/dev/null \
|| die "Ollama server not responding at $OLLAMA_HOST. Run: ollama serve"
# ── create output directory ──
local basename
basename="$(date +%Y-%m-%d_%H%M)_$(basename "${input_file%.*}" | tr ' ' '_')"
local outdir="$OUTPUT_DIR/$basename"
mkdir -p "$outdir"
banner >&2
echo " Input: $input_file" >&2
echo " STT engine: $STT_ENGINE" >&2
echo " STT model: $(basename "$model")" >&2
echo " LLM model: $OLLAMA_MODEL" >&2
echo " Language: $LANGUAGE" >&2
echo " Output: $outdir/" >&2
echo "" >&2
# ── step 1: convert ──
step "Step 1/4 — Converting audio"
local wav_file="$outdir/audio_16k.wav"
convert_audio "$input_file" "$wav_file"
# ── step 2: transcribe ──
step "Step 2/4 — Transcribing with $STT_ENGINE"
local transcript
case "$STT_ENGINE" in
whisper) transcript="$(transcribe_whisper "$wav_file" "$model" "$binary")" ;;
parakeet) transcript="$(transcribe_parakeet "$wav_file" "$model" "$binary")" ;;
*) die "Unknown engine: $STT_ENGINE" ;;
esac
if [[ -z "$transcript" || -z "$(echo "$transcript" | tr -d '[:space:]')" ]]; then
die "Transcription produced empty output"
fi
echo "$transcript" > "$outdir/transcript.txt"
local word_count
word_count=$(echo "$transcript" | wc -w | tr -d ' ')
ok "Transcript: $word_count words"
# ── step 3: summarize ──
step "Step 3/4 — Summarizing ($OLLAMA_MODEL)"
local summarize_prompt
summarize_prompt="Please provide a clear, concise summary of the following meeting transcript.
Structure the summary with:
- **Topic**: What the meeting was about
- **Key Points**: Main topics discussed (3-5 bullet points)
- **Decisions**: Any decisions that were made
- **Open Questions**: Things left unresolved
TRANSCRIPT:
$transcript"
local summary
summary="$(ollama_generate "$summarize_prompt" "You are an expert meeting summarizer. Be concise but thorough. Always use markdown formatting.")"
echo "$summary" > "$outdir/summary.md"
ok "Summary saved"
# ── step 4: action items ──
step "Step 4/4 — Extracting action items ($OLLAMA_MODEL)"
local actions_prompt
actions_prompt="Extract all action items, tasks, and commitments from the following meeting transcript.
For each action item, provide:
- **Who**: The person responsible (or 'Unassigned' if unclear)
- **What**: The specific task or action
- **When**: Any deadline mentioned (or 'No deadline specified')
- **Priority**: High / Medium / Low (based on urgency and context)
Be thorough — capture every commitment, follow-up, and task mentioned or implied.
Format as a numbered list.
TRANSCRIPT:
$transcript"
local actions
actions="$(ollama_generate "$actions_prompt" "You are an expert at extracting action items from meeting notes. Be thorough and specific. Always use markdown formatting.")"
echo "$actions" > "$outdir/action_items.md"
ok "Action items saved"
# ── combine report ──
cat > "$outdir/report.md" <<REPORT
# Meeting Report
**Date:** $(date '+%B %d, %Y at %H:%M')
**Source:** $(basename "$input_file")
**STT:** $STT_ENGINE | **LLM:** $OLLAMA_MODEL
---
## Summary
$summary
---
## Action Items
$actions
---
## Full Transcript
$transcript
REPORT
echo "" >&2
ok "All done! Files saved to: $outdir/" >&2
echo "" >&2
echo " 📄 Report: $outdir/report.md" >&2
echo " 📝 Transcript: $outdir/transcript.txt" >&2
echo " 📋 Summary: $outdir/summary.md" >&2
echo " ✅ Action Items: $outdir/action_items.md" >&2
echo "" >&2
echo "━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━" >&2
echo "" >&2
head -50 "$outdir/report.md" | tail -35 >&2
echo "" >&2
echo "(showing preview — full report at $outdir/report.md)" >&2
}
# ─── entry point ────────────────────────────────────────────────────
case "${1:-help}" in
setup) shift; cmd_setup ;;
doctor) load_config; cmd_doctor ;;
config) cmd_config ;;
help|--help|-h)
banner >&2
cat <<'HELP'
Usage:
./meetings <audio_file> Run the full pipeline
./meetings setup Install deps + download model
./meetings doctor Check all dependencies
./meetings config Show current configuration
Options (override config):
--stt <whisper|parakeet> STT engine to use
--model <path> Path to GGUF/GGML model file
--llm <ollama_model> Ollama model for summarization
--lang <code> Language code (default: en, auto for whisper)
--threads <N> Number of threads for STT
--output <dir> Output directory (default: .)
Environment variables:
MEETINGS_DIR Config & models directory (default: ~/.meetings)
MEETINGS_STT STT engine (whisper/parakeet)
MEETINGS_STT_MODEL Path to model file
MEETINGS_LLM Ollama model name (default: llama3.1:8b)
MEETINGS_THREADS Thread count (default: 4)
MEETINGS_LANG Language code (default: en)
MEETINGS_OUTPUT Output directory (default: .)
Examples:
# First time — install everything
./meetings setup
# Process a meeting recording
./meetings recording.mp3
# Use a different LLM model
./meetings meeting.wav --llm llama3.1:8b
# Use parakeet with multilingual model
./meetings call.wav --stt parakeet --lang auto
# Specify output directory
./meetings interview.m4a --output ./reports
HELP
;;
*)
cmd_process "$@"
;;
esac