# AI use cases for By2 KB (Charlesmpc/by2kb)

Repo: [Charlesmpc/by2kb](https://github.com/Charlesmpc/by2kb) | Use cases: 5 | Page: https://dedupedai.com/use-cases/charlesmpc-by2kb/

by2kb is a Python CLI and Hermes plugin that turns Bilibili and YouTube video links, or local audio and video files, into transcript, short abstract, and study-note Markdown files in a local knowledge base.

This file holds every use case of this repo. Each one starts with a "## " heading and ends with its URL.

- Save a Bilibili or YouTube video as Markdown notes: https://dedupedai.com/use-cases/charlesmpc-by2kb/save-video-link-to-knowledge-base/
- Send video links to a Hermes bot and get notes back: https://dedupedai.com/use-cases/charlesmpc-by2kb/forward-video-links-from-hermes-chat/
- Browse video picks for a topic and save the ones you choose: https://dedupedai.com/use-cases/charlesmpc-by2kb/explore-topic-and-save-chosen-videos/
- Turn local meeting audio and lecture video into notes: https://dedupedai.com/use-cases/charlesmpc-by2kb/transcribe-local-meetings-and-lectures/
- Track and control video ingest jobs from an agent: https://dedupedai.com/use-cases/charlesmpc-by2kb/control-long-ingest-jobs-from-agent/

## How to save a Bilibili or YouTube video as Markdown notes with By2 KB

Repo: [Charlesmpc/by2kb](https://github.com/Charlesmpc/by2kb) | Difficulty: Intermediate | Time: 45 minutes | Basis: From the README

Turn one video link into a raw transcript, short abstract, and long study notes on disk.

**Who it is for:** A researcher or student who finds a useful lecture or talk and wants its content kept in a personal Markdown knowledge base.

**The problem:** Videos are hard to search and revisit, and the transcript usually stays locked inside a browser or app.

**The outcome:** A content folder with source.json, transcript.json, raw, short, and long Markdown files for the video.

### Steps

1. Run `pipx install "by2kb[asr-whisper,youtube]"` to install By2 KB with local Whisper and YouTube support.
2. Run `by2kb init --preset agent-local` and follow the prompts for the knowledge base folder and summary mode.
3. Run `by2kb models install` to download the local transcription model.
4. Run `by2kb doctor` and fix any failed check it reports.
5. Run `by2kb ingest "<video URL>"` with the Bilibili or YouTube link in place of the placeholder. (inferred, check the README)

### Prompt to paste into your agent

```text
Use the by2kb repository (github.com/Charlesmpc/by2kb) to ingest the video at <video URL> into my knowledge base. Expected output: a folder under library/ containing raw, short, and long Markdown files plus source.json and transcript.json. Verify by running by2kb doctor and confirming the three Markdown files exist.
```

### Caveat

Setup needs ffmpeg on PATH and a terminal, and YouTube player challenges also need Deno or Node.js.

Dual-use notice: only use this on software, sites and systems you own or are authorized to test.

Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/

URL: https://dedupedai.com/use-cases/charlesmpc-by2kb/save-video-link-to-knowledge-base/

## How to send video links to a Hermes bot and get notes back with By2 KB

Repo: [Charlesmpc/by2kb](https://github.com/Charlesmpc/by2kb) | Difficulty: Intermediate | Time: 60 minutes | Basis: From the README

Forward a Bilibili or YouTube link in chat and receive the abstract and knowledge-base paths.

**Who it is for:** A person who discovers videos in a mobile app and wants to save them without leaving the chat app.

**The problem:** Sharing a video to a notes tool usually means switching apps and waiting without feedback.

**The outcome:** The bot acknowledges the link, processes it in the background, writes all three artifacts, and replies with the abstract and file paths.

### Steps

1. Run `pipx install "by2kb[asr-whisper,youtube]"` and then `by2kb init --preset agent-local`.
2. Run `by2kb models install` and `by2kb doctor` to confirm the deployment.
3. Run `by2kb agent install hermes` after selecting the agent option during initialization.
4. Run `hermes gateway restart`.
5. Send a Bilibili, b23.tv, YouTube, or youtu.be link to the Telegram Hermes bot.

### Prompt to paste into your agent

```text
Set up the by2kb repository (github.com/Charlesmpc/by2kb) so an authorized Hermes Telegram bot processes video links. Send the link <video URL> to the bot. Expected output: an immediate acknowledgement, then a reply with the short abstract and knowledge-base paths. Verify that the three Markdown files appear in the library folder.
```

### Caveat

The README notes that the legacy copy-based installer should not be used if Hermes already manages By2 KB through Git or the catalog, and that a native Telegram bot is still planned.

Dual-use notice: only use this on software, sites and systems you own or are authorized to test.

Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/

URL: https://dedupedai.com/use-cases/charlesmpc-by2kb/forward-video-links-from-hermes-chat/

## How to browse video picks for a topic and save the ones you choose with By2 KB

Repo: [Charlesmpc/by2kb](https://github.com/Charlesmpc/by2kb) | Difficulty: Intermediate | Time: 60 minutes | Basis: From the README

Ask Hermes about a learning topic, review recommendations, then reply with a number to transcribe one.

**Who it is for:** A self-learner starting a new subject who wants a short list of videos to study before committing time to one.

**The problem:** Finding a good starting video on an unfamiliar topic takes time, and full transcription of every candidate is wasteful.

**The outcome:** Separate Bilibili and YouTube recommendation lists, and the chosen video moves through the normal transcription and knowledge-base workflow.

### Steps

1. Complete the Hermes setup with `by2kb init --preset agent-local`, `by2kb models install`, and `by2kb agent install hermes`.
2. Restart the gateway with `hermes gateway restart`.
3. Send a message such as `by2kb 我想了解一下 TiDB 向量检索` to the Hermes bot.
4. Review the recommendation lists returned for each enabled source.
5. Reply with the number of the video you want to transcribe and save.

### Prompt to paste into your agent

```text
Use the by2kb repository (github.com/Charlesmpc/by2kb) through Hermes to search for videos about <topic>. Expected output: separate Bilibili and YouTube lists with unique numbers. Reply with the number of the chosen video and verify that three Markdown artifacts are written to the knowledge base.
```

### Caveat

Lists are limited to 3 to 5 videos per enabled source, and caption-only previews never download audio or run ASR, so previews only show what captions provide.

Dual-use notice: only use this on software, sites and systems you own or are authorized to test.

Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/

URL: https://dedupedai.com/use-cases/charlesmpc-by2kb/explore-topic-and-save-chosen-videos/

## How to turn local meeting audio and lecture video into notes with By2 KB

Repo: [Charlesmpc/by2kb](https://github.com/Charlesmpc/by2kb) | Difficulty: Intermediate | Time: 45 minutes | Basis: From the README

Ingest an mp3 or mp4 file from disk and produce the same three Markdown artifacts.

**Who it is for:** A team member or student with recorded meetings or lectures that are not hosted on Bilibili or YouTube.

**The problem:** Recordings sit unsearched on disk, and writing study notes by hand takes hours.

**The outcome:** Transcript, short abstract, and long study notes for the local file, stored in the knowledge base.

### Steps

1. Run `pipx install "by2kb[asr-whisper]"` for local Whisper support, or add the youtube extra if needed. (inferred, check the README)
2. Run `by2kb init --preset agent-local` and `by2kb models install`.
3. Confirm ffmpeg and ffprobe are on PATH, then run `by2kb doctor`.
4. Run `by2kb ingest ./meeting.mp3` for an audio file.
5. Run `by2kb ingest ./lecture.mp4` for a video file.

### Prompt to paste into your agent

```text
Use the by2kb repository (github.com/Charlesmpc/by2kb) to ingest the local file <file path> into my knowledge base. Expected output: a content folder with raw, short, and long Markdown files. Verify by running by2kb doctor --json and checking that the media step succeeded.
```

### Caveat

Local files depend on ffmpeg and ffprobe being installed, and the README does not list every supported format beyond the examples given.

Dual-use notice: only use this on software, sites and systems you own or are authorized to test.

Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/

URL: https://dedupedai.com/use-cases/charlesmpc-by2kb/transcribe-local-meetings-and-lectures/

## How to track and control video ingest jobs from an agent with By2 KB

Repo: [Charlesmpc/by2kb](https://github.com/Charlesmpc/by2kb) | Difficulty: Advanced | Time: 60 minutes | Basis: From the README

Check status, wait, cancel, or retry a running By2 KB job using JSON output.

**Who it is for:** An operator or developer running By2 KB through an agent host who needs reliable job control instead of parsing logs.

**The problem:** Long transcriptions run in the background, and an agent needs a clear way to know whether a job finished, failed, or should be retried.

**The outcome:** JSON status for each job, a bounded wait, and clean cancel or retry actions with checkpoint-aware retries.

### Steps

1. Start an ingest with `by2kb ingest "<video URL>"`. (inferred, check the README)
2. Run `by2kb status --json` to see the current job state. (inferred, check the README)
3. Run `by2kb wait --timeout 30 --json` to wait up to the timeout for completion. (inferred, check the README)
4. Run `by2kb cancel --json` if the job should stop. (inferred, check the README)
5. Run `by2kb retry --json` to resume a failed job from its checkpoint. (inferred, check the README)

### Prompt to paste into your agent

```text
Use the by2kb repository (github.com/Charlesmpc/by2kb) to start ingesting <video URL> and then manage the job with status, wait, cancel, and retry in JSON mode. Expected output: a JSON status for each step and the final artifact paths. Verify that the job reaches a completed state or that the retry resumes it.
```

### Caveat

The README describes the control protocol but does not give the full JSON schema, so an agent should read docs/task-control.md before parsing fields.

Dual-use notice: only use this on software, sites and systems you own or are authorized to test.

Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/

URL: https://dedupedai.com/use-cases/charlesmpc-by2kb/control-long-ingest-jobs-from-agent/