GuideYourMuscle Review
A Chinese Streamlit AI fitness coach with RAG, meal photo recognition, diet tracking, and training plans.
Each review covers deployability, value versus commercial alternatives, and privacy model. Tools that can run locally were started and exercised; mobile or backend-dependent tools were assessed from published builds, source code, and deploy guides. Ratings reflect what we were able to verify.
GuideYourMuscle — Open Source AI Fitness Coach
A Chinese Streamlit AI fitness coach with RAG, meal photo recognition, diet tracking, and training plans. Last updated: 2026-06-21.
One-sentence verdict: A feature-rich fitness-coach prototype for Chinese-speaking users; the partial English translation and DashScope dependency make it less accessible for English audiences.
What the System Is
GuideYourMuscle is a multi-page Streamlit app that acts as an AI fitness coach. The deploy guide reports the service was started on port 5018; page titles and subtitles were translated to English, but the application body remains Chinese.
The documented modules are:
- AI coach chat — ReAct agent with seven tools and four skills.
- AI meal photo recognition — uploads a food photo for nutrition logging.
- Diet management — nutrition targets and meal records.
- Training plans — workout logging and AI feedback.
- Data analytics — weight and training trends with Plotly charts.
- Personal settings — model selection, including optional local fine-tuned Qwen2.5-7B.
The RAG layer combines BM25, FAISS, and a reranker over a local fitness knowledge base (training_science.md, nutrition_guide.md, injury_prevention.md, exercise_library.md).
| Key data | |
|---|---|
| Category | Fitness & Nutrition |
| Language | Python / Streamlit |
| License | MIT |
| Self-hosted | Yes |
| AI provider | DashScope (embedding, reranker, vision); optional local Qwen2.5-7B |
| Database | SQLite |
| Default login | admin / 123456 |
How to Install and Deploy
cd /data2/docker/going-global/repos/GuideYourMuscle
python -m venv venv
source venv/bin/activate
pip install -r requirements.txt
cp config.example.py config.py
# Edit config.py and add your DashScope API key
streamlit run App.py --server.port 5018 --server.headless true
Then open http://localhost:5018.
The optional local model path uses deploy/api_server.py to serve a fine-tuned Qwen2.5-7B and then selects it in the settings page.
How to Test
The documented test flow is:
- Open
http://localhost:5018and log in withadmin/123456. - Navigate to the AI coach page and ask a fitness or nutrition question.
- Upload a food photo on the meal-recognition page and verify a macro estimate is returned.
- Log a meal and a workout, then check the diet and training pages.
- Open the analytics page and confirm trend charts render.
The deploy guide notes that vision and embedding features require a valid DashScope key.
Privacy & Compliance
GuideYourMuscle is not HIPAA compliant. Food photos, chat text, and embeddings are sent to DashScope unless you switch to the optional local Qwen model. User data is stored in a local SQLite database with multi-user isolation.
GuideYourMuscle vs Commercial Fitness Coaches
| Dimension | GuideYourMuscle | Commercial Apps (e.g., Freeletics, Future, Caliber) |
|---|---|---|
| Cost | Free / self-hosted + DashScope usage | Subscription |
| Language | Primarily Chinese UI | English and local languages |
| Data model | Local SQLite; cloud vision/embeddings | Cloud-hosted |
| Personalization | RAG + agent over local knowledge base | Proprietary coaching algorithms |
| Setup effort | Python venv + DashScope key | Mobile app signup |
| Offline use | Partial (local model path only) | No |
Who Should Use It
- Chinese-speaking users wanting an open-source AI fitness coach.
- Developers interested in RAG (BM25 + FAISS + reranker) and ReAct agents in Streamlit.
- Teams with access to DashScope who want to customize a fitness knowledge base.
Who Shouldn't Use It
- English-speaking users who need a fully localized interface.
- Users who want a polished native mobile experience.
- Organizations needing HIPAA-compliant health data handling.
FAQ
Is the interface in English?
The deploy guide reports that page titles and subtitles were translated to English, but the main application body remains Chinese.
Does it require a cloud API?
Yes, for embeddings, reranker, and vision recognition it uses DashScope. A locally fine-tuned Qwen2.5-7B can be used as an alternative chat model if you have GPU resources.
What data is stored locally?
User profiles, meal logs, training records, and analytics are stored in a local SQLite database. Photos may be sent to DashScope for vision analysis.
Verdict
GuideYourMuscle is the most feature-complete open-source fitness coach in this batch. Its RAG architecture, skill system, and optional local model support are well thought out. The Chinese UI and DashScope dependency are the main barriers for English-speaking users.
Ratings: Deployability 3/5 · Value vs Commercial 3/5 · Privacy Compliance 3/5