YouTube Channel Strategist

README excerpt

# YouTubeStrategist


## Installation

```bash
uv venv ~/env313
. ~/env313/bin/activate
uv pip install jupyterlab jupytext ipywidgets omegaconf numpy 
```


```bash
REMOTE='yishun:/home/andrewsm/Videos'
rsync -avzm --include="*/" --include="YouTube_transcription.srt" --exclude="*" ${REMOTE}/ ./
rsync -avzm --include="*/" --include="AudienceRetention.zip"     --exclude="*" ${REMOTE}/ ./

# Add link to folder with videos in it
ln -s ../Videos .
```

## Running the Demo

```bash
cd retention-app
npm run dev
```


## Prompts

### Get the 'Channel Script Strategy Profile'

```md
You are an expert YouTube content analyst. Your goal is to map my video transcripts to my audience retention signals. 
For each youtube video, there is a transcript (with retention data per line) and YouTube info structure.

There are a number of these pairs, please identify the patterns across them, specifically, identify the 'Retention DNA' to find:
* the top 5 tactical patterns that correlate with high retention; and 
* the top 5 patterns that correlate with immediate drop-offs. 

Output this as a structured 'Channel Script Strategy Profile' (JSON).
```

### Get the 'Channel CTR Strategy Profile'

```md
From the data above, can you make concrete suggestions about how to increase the click-through-rate for the videos?
* If click-through-rate is not explicitly available, just use the number of views as a proxy

Output this as a structured 'Channel CTR Strategy Profile' (JSON).
```

### Get the 'Retention