healthyselfjournal 0.2.8

Voice journaling CLI with Whisper transcription and Claude follow-ups

Healthyself Journal

Speak your thoughts. Get good questions. Build healthier patterns.

A voice-based journaling and self-reflection tool. Just ramble for a while out loud about what's on your mind, and receive a warm, thoughtful, evidence-based question each time in response to help you think things through.

Note: this beta version runs in the command-line and relies on OpenAI & Anthropic models.

What makes this different

• 🎙️ Voice-first: Start journaling instantly by speaking – no typing, no friction.
• 🧠 Wise, helpful questions: Evidence-based prompts adapted from cognitive behavioral therapy, psychology research, mindfulness practice, and famous coaches.
• 🔄 Keeps you moving: Gentle redirection when you're spiraling; deeper exploration when you're onto something.
• 📊 Builds on your history: Each session connects to previous ones for continuity and growth.
• 🔒 Privacy choice: Use Anthropic + OpenAI OR private/local LLM+transcription (in beta), as you prefer. Healthyself doesn't have any backend, storage, or analytics of its own outside your machine at all).
• 🛡️ Aims to be safe: Anti-sycophancy, rumination pivots, and clear boundaries; see docs/reference/SAFEGUARDING.md.
• Open source. So you can see & modify the prompts and research for yourself.
• This is assisted reflection. It's not trying to be or replace human conversation or therapy.

Quick start

# Recommended: run without installing (uvx)
uvx healthyselfjournal -- init
uvx healthyselfjournal -- journal cli

That's it. Recording starts immediately. Press ENTER to stop and get your next question.

Tip for first‑time users: if uvx isn’t found, install uv (which provides uvx) and try again:

# macOS/Linux
curl -LsSf https://astral.sh/uv/install.sh | sh
# Windows (PowerShell)
powershell -ExecutionPolicy ByPass -c "irm https://astral.sh/uv/install.ps1 | iex"

How it works

1. Start speaking – Recording begins immediately when you launch
2. Press any key to stop – Your audio is transcribed automatically
3. Get a thoughtful question (out loud if you want) – Based on what you shared and your patterns
4. Keep going or wrap up – Continue as long as helpful, or press Q to end with a summary

Example session

🧑 You: "I'm torn between applying for the new role and doubling down on my current project. I'm worried I'll disappoint people either way."

🤖 AI: "I hear the pull between gratitude for people who believed in you and the desire for the growth you've been wanting. When you say these are 'real relationships, not just work' — what would honoring those relationships look like as you make this decision?"

[conversation continues...]

You can see the full questioning prompt, and the research on which it's based below.

For example, the question above draws on the following approaches: [Values exploration, Clean Language (preserving "real relationships"), Socratic questioning, autonomy-supportive framing].

Installation & Setup

Requirements

• Python 3.10+
• Microphone access (grant permission to your terminal/app)
• Optional cloud keys:
  • OPENAI_API_KEY (speech-to-text when using --stt-backend cloud-openai)
  • ANTHROPIC_API_KEY (LLM questions/summaries; default cloud provider)
• Optional local/offline (in beta):
  • Ollama running (for local LLM via --llm-model ollama:<model>)
  • One STT backend installed: mlx-whisper (Apple Silicon), faster-whisper, or whispercpp + .gguf model
• Optional: ffmpeg on PATH for background MP3 conversion
• Linux only: install audio libs (e.g., sudo apt install portaudio19-dev libsndfile1)

See docs/reference/AUDIO_VOICE_RECOGNITION_WHISPER.md and docs/reference/PRIVACY.md for details.

Install

# Option 1: Run without installing (recommended)
uvx healthyselfjournal -- init

# Option 2: Install with pip
pip install healthyselfjournal

First-time setup

The setup wizard will help you:

• Add your API keys securely
• Choose between Cloud mode (recommended) or Privacy mode (fully offline, but still in beta)
• Pick where to save your journal sessions

# Recommended
uvx healthyselfjournal -- init

Daily use

# Start a new session
uvx healthyselfjournal -- journal cli

# Continue your last session
uvx healthyselfjournal -- journal cli --resume

Use --sessions-dir to decide where to store them - otherwise it defaults to sessions/.

Insights (beta)

Generate high-level reflective insights every so often spanning recent conversations, saved under sessions/insights/:

# List existing insights
uvx healthyselfjournal -- insight list --sessions-dir ./sessions

# Generate multiple insights in a single file
uvx healthyselfjournal -- insight generate --sessions-dir ./sessions --count 3

Controls

• Any key: Stop recording and get your next question
• ESC: Cancel the current recording (discard it)
• Q: Save and quit after this response

Privacy options

Cloud mode (default): Uses OpenAI for transcription and Anthropic Claude for questions. Best accuracy and response quality.

Privacy mode: Everything stays on your device. Requires Ollama for local AI and choosing a local transcription option. See docs/reference/PRIVACY.md for details.

Where your journal lives

Your sessions are saved as markdown files with audio recordings:

sessions/
├── 250919_143022.md          # Today's afternoon session
├── 250919_143022/
│   ├── 250919_143022_01.wav  # Your voice recordings
│   └── 250919_143022_02.wav
└── events.log                 # Activity log

You own all your data. Export it, back it up, or delete it anytime.

The research behind it

Healthyself journal is built primarily on evidence-based psychological research (and a few other things that we have found helpful), integrating over 30 documented therapeutic and coaching frameworks:

Core Therapeutic Foundations

• Cognitive Behavioral Therapy (CBT): Socratic questioning to identify and reframe thought patterns (meta-analyses show d=0.73 effect size)
• Motivational Interviewing: Amplifying "change talk" and intrinsic motivation (70+ RCTs supporting effectiveness)
• Clean Language (David Grove): Using your exact words and metaphors to maintain authenticity and avoid therapist contamination
• Explanatory Style (Seligman's 3 P's): Challenging permanence, pervasiveness, and personalization in negative thinking

Anti-Rumination & Safety Features

• Structured vs. Destructive Rumination: Evidence-based detection of maladaptive thought loops
• Self-Distancing Techniques: Third-person perspective and temporal distancing (strong neurological evidence)
• Concrete vs. Abstract Processing: Redirecting to specific, actionable thoughts when stuck
• Session Timing Optimization: 15-20 minute sweet spot to prevent rumination (based on expressive writing research)

Narrative & Meaning-Making

• Redemptive Narrative Construction (McAdams): Guiding from contamination to growth narratives
• Implementation Intentions: "When-then" planning for 2-3x better habit formation
• Cognitive-Emotional Integration: Balanced processing outperforms emotion-only expression

Mindfulness & Contemplative Practices

• Plum Village Tradition: Mindful reflection and present-moment awareness
• Beginning Anew Practice: Four-part framework for relationship and self-compassion
• Body Awareness Integration: Somatic grounding when caught in mental loops

Coaching Methodologies

• GROW Model: Goal-Reality-Options-Will framework with strong evidence base
• Solution-Focused Brief Therapy: Future-oriented questions emphasizing strengths
• Values Clarification (ACT): Connecting actions to core personal values

Expert Practitioner Wisdom

Questions inspired by renowned coaches and researchers:

• Tim Ferriss' fear-setting and simplification frameworks
• Jerry Colonna's radical self-inquiry
• Martha Beck's body compass methodology
• Tony Robbins' reframing techniques
• Arthur Brooks' failure integration

Cultural & Individual Adaptation

• Cultural Sensitivity: Avoiding Western-centric assumptions about gratitude and individual achievement
• Personalization: Adapting to user patterns, chronotype, and emotional states
• Developmental Considerations: Age-appropriate approaches based on psychological development

The system continuously analyzes your responses for emotional intensity, thought patterns, topic persistence, exhaustion signals, and readiness for change, adapting its questioning strategy based on session phase and your current needs.

For an overview of all 30+ research areas and methodologies, see docs/research/RESEARCH_TOPICS.md and docs/reference/SCIENTIFIC_RESEARCH_EVIDENCE_PRINCIPLES.md.

Advanced options

(These are in development, and probably won't work yet)

Desktop app (in alpha)

healthyselfjournal journal desktop --voice-mode

Different AI models

# Use a local model (requires Ollama)
healthyselfjournal journal cli --llm-model ollama:gemma3:27b-instruct-q4_K_M

Custom session location

healthyselfjournal journal cli --sessions-dir ~/Documents/journal

Support & Documentation

• Issues or questions: GitHub Issues
• Full documentation: See the docs/ folder
• Contributing: Contributions welcome! See CONTRIBUTING.md

Technical details

For developers and technical users:

• Built with Python, using Typer for the CLI.
• OPTIONAL/ALPHA FastHTML for the web interface and PyWebView for desktop
• Transcription via OpenAI Whisper API (or local alternatives)
• Questions generated by Anthropic Claude (or local Ollama models)
• Either everything runs locally (your data never leaves your device), or choose cloud services
• See docs/reference/ARCHITECTURE.md for system design
• See AGENTS.md for development setup