Blog-writer-crewai

# Blog-writer-crewai

Multi-agent blog generation project built with CrewAI. It researches a topic from a target YouTube channel, extracts transcript-based insights, and writes a blog post in Markdown.

## Overview

This project uses two agents in sequence:

- `blog_researcher`: finds relevant YouTube content and transcript context
- `blog_writer`: turns research into a readable technical blog post

The default flow is defined in `exwc.py` and runs with CrewAI `Process.sequential`.

## Features

- Agent-based research and writing workflow
- YouTube channel search via `YoutubeChannelSearchTool`
- Transcript retrieval through `youtube_transcript_api`
- Topic-driven generation (`crew.kickoff(inputs={"topic": ...})`)
- Markdown output file generation (`new-blog-post.md`)

## Tech Stack

- Python
- CrewAI
- crewai-tools
- YouTube Transcript API
- python-dotenv
- OpenAI API

## Project Structure

```text
Blog-writer-crewai/
	agents.py            # Agent definitions (researcher, writer)
	tasks.py             # Task definitions and output config
	tools.py             # YouTube search + transcript tools
	exwc.py              # Main runnable crew pipeline
	try.py               # Alternative single-file experimental pipeline
	requirements.txt     # Dependencies
	.gitignore
```

## Prerequisites

- Python 3.10+
- OpenAI API key
- Internet access (YouTube + API calls)

## Setup

1. Clone the repository

```bash
git clone <your-repo-url>
cd Blog-writer-crewai
```

2. Create and activate a virtual environment

Windows PowerShell:

```powershell
python -m venv .venv
.\.venv\Scripts\Activate.ps1
```

macOS/Linux:

```bash
python -m venv .venv
source .venv/bin/activate
```

3. Install dependencies

```bash
pip install -r requirements.txt
```

4. Create a `.env` file in the project root

```env
OPENAI_API_KEY=your_openai_api_key_here
```

## Run

Primary pipeline:

```bash
python exwc.py
```

Alternative experimental script:

```bash
python try.py
```

## How It Works

1. `blog_researcher` searches the configured YouTube channel (`@krishnaik06`) for topic-relevant content.
2. Transcript text is fetched using the custom `YoutubeTranscriptTool`.
3. `research_task` summarizes key ideas.
4. `write_task` creates a blog article and saves it to `new-blog-post.md`.

## Configuration Notes

- Topic is currently hardcoded in `exwc.py`:

```python
result = crew.kickoff(inputs={'topic': 'AI VS ML VS DL vs Data Science'})
```

- To generate for other topics, update that input string.
- Model settings are configured in `agents.py` via environment variables:
	- `OPENAI_API_KEY`
	- `OPENAI_MODEL_NAME` (currently set to `gpt-4-0125-preview` in code)

## Output

- Main generated file:
	- `new-blog-post.md`

- Optional files from experimental script (`try.py`):
	- `research.txt`
	- `blog_post.md`

## Known Limitations

- Depends on transcript availability for the selected video.
- Channel is fixed in `tools.py` unless you modify the handle.
- API/model compatibility depends on your installed CrewAI version and active OpenAI models.

## Suggested Improvements

- Add CLI arguments for topic and channel.
- Add retries/fallback when transcripts are unavailable.
- Add logging and structured error handling.
- Add tests for tools and task orchestration.
- Make output folder configurable.

## Disclaimer

Generated content may require fact-checking and editorial review before publication.