Write Once, Teach Personally

English | 简体中文

Developed by the AI-Shifu Team and the Research Center of Intelligent Software Engineering at Harbin Institute of Technology.

Core Capabilities

• Personalized explanation engine — Generates learning paths and tone based on learner background, goals, and level.
• Interactive Q&A & probing — Decomposes questions, asks clarifiers, and suggests next actions during sessions.
• Rapid course assembly — Author with high-level frameworks and intent; AI-Shifu elaborates into lessons, activities, and assessments.
• Reduced production & delivery overhead — Minimizes repetitive prep and support; every learner gets a dedicated “AI tutor.”
• Multi-channel integration — Embeddable in websites, course platforms, and enterprise training portals.

Use Cases

• Course creators — Hand a single lesson framework to AI-Shifu; learners receive personalized explanations and real-time interaction.
• Enterprise training — Input training content once; employees get role- and background-specific learning paths.
• Educators — Provide a syllabus to generate personalized coaching content plus a Q&A assistant.

Roadmap

• [ ] Writing AI agent for rapid script generation and maintenance
• [ ] Knowledge base
• [ ] Speech input and output

Using AI-Shifu

Platform

AI-Shifu.com is an education platform powered by AI-Shifu. You can try it and learn the AI-guided courses developed by human experts.

Self-hosting

For source code installation, please refer to the Installation Manual

Make sure your machine has installed Docker and Docker Compose.

Quick Start (Docker)

git clone https://github.com/ai-shifu/ai-shifu.git
cd ai-shifu/docker
Use Docker-ready defaults (matches bundled MySQL service; Redis is optional)
cp .env.example.full .env
Edit .env: configure an LLM provider API key and set LLM_MODEL_1_ID
to a text model ID served by that provider (required; no default).
LLM_MODEL_1_NAME is optional; an empty name displays the model ID.
Start all services
docker compose -f docker-compose.latest.yml up -d

Notes

• LLM_MODEL_1_ID is required and has no default. Configure optional model IDs for numbers 2-9; gaps are supported. All LLM_MODEL__NAME values are optional: omitted or blank names display the configured model ID. A name without a model ID does not enable an option. Historical selections automatically use model 1 when they do not reference a configured number; no data cleanup is needed.
• For an existing installation, follow Upgrading to numbered models before starting the new images. This applies to latest, pinned-release and development Compose modes.
• First verified user is automatically promoted to Admin and Creator; the bundled demo course is assigned to this user.
• Default universal verification code for demos is 1024 (change via UNIVERSAL_VERIFICATION_CODE).
• docker-compose.latest.yml pulls the freshest :latest images (or your own locally built latest tags). Use docker-compose.yml when you need pinned release tags for reproducible environments.

Using Docker Hub image (customize)

git clone https://github.com/ai-shifu/ai-shifu.git
cd ai-shifu/docker
Copy the full template (contains defaults for Docker usage)
cp .env.example.full .env
Edit .env: a provider key and the model 1 binding are required:
- OPENAI_API_KEY / ERNIE_API_KEY / BIGMODEL_API_KEY / ...
- LLM_MODEL_1_NAME: optional display name; blank uses the configured model ID
- LLM_MODEL_1_ID: a configured text model ID (no default)
- SQLALCHEMY_DATABASE_URI: Defaults to docker MySQL service
- REDIS_HOST: Optional; set to enable Redis caching/locks (leave empty to disable)
- SECRET_KEY: Defaults to a demo value; change for production (generate with: python -c "import secrets; print(secrets.token_urlsafe(32))")
- UNIVERSAL_VERIFICATION_CODE: Test verification code (remove/empty in production)
- Any other optional integrations
docker compose -f docker-compose.latest.yml up -d  # Use -f docker-compose.yml for pinned versions

Development mode (dev_in_docker.sh)

git clone https://github.com/ai-shifu/ai-shifu.git
cd ai-shifu/docker
cp .env.example.full .env
Edit .env: set your LLM API key(s) and map LLM_MODEL_1_ID
to a text model ID available through the configured provider.
Optionally set LLM_MODEL_1_NAME; blank uses the configured model ID.

./dev_in_docker.sh

dev_in_docker.sh builds the backend and frontend images from your local source tree and then launches docker-compose.dev.yml (hot reload + bind mounts). Use it whenever you need to iterate on code without managing Python/Node runtimes locally.

The script records the checkout's current commit before building the Web image, so the user menu can show its source revision. Source archives without Git metadata show only the package version.

To start the dev stack directly with Compose, run these commands from docker/:

git rev-parse HEAD > ../src/web/.app-build-sha 2>/dev/null || rm -f ../src/web/.app-build-sha
docker compose -f docker-compose.dev.yml up --build -d

Prepare the marker and rebuild each time you start from a different commit so an existing dev image does not keep an older revision.

Compose files

• docker-compose.latest.yml: tracks the :latest tags for aishifu/ai-shifu-api and aishifu/ai-shifu-web. Use this when you want the freshest container build (either from Docker Hub or after running your own docker build ... -t aishifu/...:latest).
• docker-compose.yml: pins each image to a specific release tag for reproducible deployments (recommended for staging/prod mirrors or CI).

Access

After Docker starts:

1. Open http://localhost:8080 in your browser to access Cook Web (learner interface and authoring console)
2. Use any phone number for login; the default universal verification code is 1024 (for demo/testing only — change or disable in production)
3. The first verified user becomes Admin and Creator

Internationalization (i18n)

• Shared translations live in src/i18n//*/.json and are consumed by both Backend and Cook Web.
• See the consolidated i18n guide for the product-locale

• Supported frontend languages are declared in src/i18n/locales.json.

Text-to-Speech (TTS)

AI-Shifu supports multiple TTS providers. To enable Volcengine HTTP v1/tts, set:

• VOLCENGINE_TTS_APP_KEY (AppID)
• VOLCENGINE_TTS_ACCESS_KEY (Token used by Authorization: Bearer;{token})
• VOLCENGINE_TTS_CLUSTER_ID (Cluster, default: volcano_tts)