| name | setup |
| description | Configure Tongji Look credentials, check system dependencies (Python, Node.js, ffmpeg, vision-support, XeLaTeX), and set up the persistent course-wiki workspace. |
| license | MIT |
| metadata | {"author":"WALKERKILLER","version":"1.1"} |
Setup
Configure credentials, dependencies, and workspace for Look Tongji Notes.
When to Use
- User says
/setup or asks to configure the skill.
- First-time setup before transcribing lectures.
- User wants to change or migrate the course knowledge-base path.
Workflow
- Run setup:
python "<SKILL_DIR>/../../scripts/look_tongji.py" setup
- Non-interactive workspace options:
python "<SKILL_DIR>/../../scripts/look_tongji.py" setup \
--workspace-root "<COURSE_WIKI_ROOT>" \
--owner-name "<OWNER_NAME>" \
--site-name "<OWNER_NAME>的课程知识库"
-
The CLI checks: Python deps (requests, playwright, python-dotenv, markdown, markitdown), ffmpeg, Node.js, vision-support config, and optionally XeLaTeX.
-
Vision Support setup: If vision-support/config.json does not exist, the CLI prints the init command. The embedded vision-support is at <SKILL_DIR>/../../vision-support/. Help the user configure a vision provider (OpenAI, Google, Anthropic, deepseek, dashscope, zhipuai, ollama, or custom).
-
Verify vision-support: After configuration, test with:
node "<SKILL_DIR>/../../vision-support/scripts/vision.mjs" "<SKILL_DIR>/../../komari.jpg"
The agent should correctly identify image content (should mention "red-haired girl" or equivalent).
-
Never ask for passwords in chat. Use interactive terminal input or environment variables.
Where <SKILL_DIR> Points
<SKILL_DIR> is the directory containing this SKILL.md. Shared scripts (look_tongji.py, timeline_tools.py, tongji_backend/) and references live two levels up in the repository root (<SKILL_DIR>/../../scripts/ and <SKILL_DIR>/../../references/).
