BlogWritingCrew

A multi-agent AI system, built with crewAI, that turns a single topic into a polished blog post. Three agents — a researcher, a writer, and an editor — work one after another, each handing its output to the next.

How it was built

The project started from the standard crewai create crew template and was shaped into a 3-agent pipeline:

• config/agents.yaml — defines each agent's role, goal, and backstory (Researcher, Writer, Editor)
• config/tasks.yaml — defines what each agent has to do and what "done" looks like
• crew.py — wires agents to tasks and runs them as a Process.sequential crew
• main.py — the entry point crewai run calls, with a hardcoded topic input
• LLM — Google Gemini, configured entirely through .env (MODEL=gemini/... + GEMINI_API_KEY), no code changes needed

Architecture

flowchart LR
    U["crewai run"] --> M["main.py: run()"]
    M --> C["BlogWritingCrew.crew()"]

    subgraph Crew["Sequential Process"]
        direction LR
        R["Researcher Agent"] -->|research brief| W["Writer Agent"]
        W -->|draft post| E["Editor Agent"]
    end

    C --> R
    E --> O["output/blog_post.md"]

Loading

Task pipeline

sequenceDiagram
    participant Main as main.py
    participant Crew as BlogWritingCrew
    participant LLM as Gemini

    Main->>Crew: kickoff(inputs={topic, current_year})
    Crew->>LLM: research_task (Researcher)
    LLM-->>Crew: research brief
    Crew->>LLM: writing_task (Writer)
    LLM-->>Crew: draft blog post
    Crew->>LLM: editing_task (Editor)
    LLM-->>Crew: polished post
    Crew-->>Main: output/blog_post.md

Loading

Project structure

flowchart TD
    root["blog_writing_crew/"] --> src["src/blog_writing_crew/"]
    src --> crewpy["crew.py — agents + tasks + crew definition"]
    src --> mainpy["main.py — run / train / replay / test entry points"]
    src --> config["config/"]
    config --> agentsyaml["agents.yaml"]
    config --> tasksyaml["tasks.yaml"]
    src --> tools["tools/custom_tool.py — example tool (unused)"]
    root --> knowledge["knowledge/user_preference.txt — example knowledge source (unused)"]
    root --> output["output/blog_post.md — final result"]

Loading

Example output

Running crewai run executes the researcher → writer → editor pipeline:

...and writes the final result to output/blog_post.md:

Installation

Requires Python >=3.10 <3.14. Dependencies are managed with uv.

pip install uv
crewai install

Configuration

Add to .env:

MODEL=gemini/gemini-flash-lite-latest
GEMINI_API_KEY=your-key-here

• Edit src/blog_writing_crew/config/agents.yaml to change agent roles/goals
• Edit src/blog_writing_crew/config/tasks.yaml to change what each task asks for
• Edit src/blog_writing_crew/main.py to change the topic input

Running the project

crewai run

This runs the researcher → writer → editor pipeline and writes the final post to output/blog_post.md.

Support

• crewAI documentation
• crewAI GitHub