by2kb CLI
Turns a Bilibili or YouTube link or a local file into a transcript, a short abstract and study notes in Markdown
- 158 ★ on the repository
- Last commit
Extended access
Why this level
- The CLI downloads media from external platforms and writes files into your knowledge base
- The transcript text goes to the agent model or to a provider through your API key
Install
In your terminal, with SkillFoxx CLI
npx skillfoxx add cli/by2kbDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- The CLI downloads media from external platforms and writes files into your knowledge base
- The transcript text goes to the agent model or to a provider through your API key
This is third-party code. Review the repository files before installing.
What it does
by2kb is a Python CLI that files knowledge from videos into your Markdown folder. The `by2kb ingest` command takes a Bilibili or YouTube link or a local audio or video file, downloads the track, transcribes it with local Whisper or a cloud service and saves three files: a timestamped transcript, a short "worth reading?" abstract and detailed study notes. The main scenario targets someone who talks to an agent, for example Hermes in Telegram. You send the link to the agent, `by2kb` handles media retrieval and transcription, and the agent writes the summaries with its own model. Standalone mode works with your own key for any OpenAI-compatible API. The `by2kb agent install hermes` command installs a plugin for Hermes. Version 0.7 added topic discovery: a request like "I want to understand TiDB" returns video lists from Bilibili and YouTube, and a reply with a number starts transcribing the chosen video. The `by2kb doctor`, `by2kb status` and `by2kb wait` commands show the state of the environment and jobs.
Who it is for. People who build a personal notes base from lectures, talks and meeting recordings and want to forward links to an agent straight from a messenger.
Good fit when
- A Bilibili or YouTube lecture runs an hour and you need notes tied to timestamps
- A meeting recording sits in `meeting.mp3` and has to become text and a short abstract
- You want to send a link to an agent in Telegram and get notes in your Markdown folder
- You want a list of videos on a topic to pick one and transcribe it right away
Not a fit when
- Videos on other platforms: built-in sources are currently only Bilibili and YouTube, with podcasts and articles listed as future extensions
- You need a resident service or native Telegram and Lark bots: those remain planned
- You only need a summary of existing text with no media download
Example request
Transcribe https://www.bilibili.com/video/<bvid>/ and put the transcript, abstract and notes in my knowledge baseLimitations
Python 3.12 or newer, `pipx` and `ffmpeg` with `ffprobe` on PATH are required. Local Whisper transcription downloads models and loads the CPU, a cloud service needs its own key, for example Doubao. YouTube may need extra download settings.
The agent or your API key writes the summaries, so the transcript text goes to the provider of the chosen model. Media download from platforms depends on their terms and network reach from your region. The author lists a resident service, native bots and remote knowledge-base sinks as planned: they do not exist yet.
Jobs have operation locking and status commands, but test batch processing on your own data.
How to disable. Remove the package with `pipx uninstall by2kb`. If you installed the Hermes plugin, remove it from the Hermes plugin directory and restart the gateway. Notes in your Markdown folder stay.
FAQ
How do I install by2kb?
Run `pipx install "by2kb[asr-whisper,youtube]"`, then `by2kb init --preset agent-local` and `by2kb models install`. For cloud transcription there is an `asr-doubao` extra instead of Whisper.
Which files does by2kb create for one video?
Three Markdown files: a transcript that preserves the evidence, a short abstract on whether it is worth reading and detailed study notes.
Do I need a separate API key?
Not in agent mode: the agent model writes the summaries using its own authorization. Standalone mode needs a key for an OpenAI-compatible API.
Can I transcribe a local file?
Yes, `by2kb ingest ./lecture.mp4` and `by2kb ingest ./meeting.mp3` accept local video and audio.
Which agents does the plugin work with?
The author describes Hermes installation with `by2kb agent install hermes`. Other agents can call the CLI directly and use the request recipes in `docs/`.
Discussion
No questions or reviews yet.
Related
CLI toolsA skill that fetches YouTube captions and builds an HTML report with takeaways, timestamps and an embedded player
A CLI and Codex skill turning YouTube and Bilibili videos into transcripts and timestamped summaries, with Notion export.
Skill that syncs favorite Douyin videos into a local Markdown or Obsidian base with speech transcripts and a daily digest.
A CLI and skill for Yandex Tracker and Wiki: short output readable by people and agents alike, instead of raw MCP payloads costing thousands of tokens
A skill that teaches an agent to search, browse and read local documents through Linkly AI: PDF, Markdown, Word, Excel and more.
The Linkly AI client: an agent searches local documents, reads and saves notes, with an MCP bridge for Claude Desktop
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Repository stars | Main difference |
|---|---|---|---|
| by2kb (this card) | no | 158 | |
| video-to-notes | no | 113 | takes notes on videos without Bilibili support or a Markdown knowledge base |
| bilibili-cli | no | 1,152 | works with Bilibili itself from the terminal with no transcripts or notes |
README in short
The repository describes a pipeline from a video link to three knowledge files. According to the docs it implements agent-guided installation, local Whisper by default, Bilibili and YouTube ingestion, Agent and API enrichment, task control and a Hermes plugin. Version 0.7.0 added topic discovery.
README badge
Original SKILL.md
--- name: default-video-digest description: Turn a transcript into grounded, structured notes for deep study. version: 1.0.0 --- # Deep study notes ## Use when Creating the long-form updated Markdown artifact from a normalized video transcript. ## Rules 1. Preserve the raw transcript separately; never overwrite it. 2. Do not invent claims, speakers, chapters, or timestamps. 3. Write in the transcript's primary language unless metadata strongly indicates another language. 4. Every chapter and important claim must link to the nearest source timestamp. If useful timing is unavailable, omit links rather than inventing them. 5. Distinguish what the speaker said from your own synthesis or questions. 6. Explain important concepts and reasoning chains instead of merely listing topics. 7. If the transcript is incomplete or low quality, state that prominently. ## Output Produce these sections: 1. `Core thesis` — the purpose, conclusion, and overall argument. 2. `Knowledge map` — how the major concepts relate. 3. `Guided walkthrough` — timestamped sections with detailed explanations. 4. `Claims and evidence` — important claims paired with their support. 5. `Concepts and terminology` — definitions in context. 6. `Questions and checks` — assumptions, counterarguments, and points to verify. 7. `Study and action list` — review prompts, exercises, or next actions.