RepoPersona Agent

A multi-agent AI system built with CrewAI and Google Gemini that analyzes GitHub profiles and delivers HR-ready positioning strategies to help developers stand out to recruiters and hiring managers.

---

Overview

This project implements a two-agent CrewAI crew that:

1. Scrapes and analyzes a target GitHub profile — evaluating code quality, documentation, project diversity, contribution patterns, and overall professional presentation from an HR lens.
2. Generates a prioritized improvement plan — actionable recommendations tailored to how non-technical HR professionals and technical hiring managers evaluate candidates.

The crew runs sequentially: the GitHub Profile Analyst feeds its findings directly into the HR Positioning Strategist, producing a comprehensive, recruiter-ready report.

---

Architecture Flow Diagram

flowchart TD

    A[User Inputs GitHub Profile URL]

    B[CrewAI Orchestrator]

    C[GitHub Profile Analyst Agent]
    D[ScrapeWebsiteTool]
    E[GitHub Profile Analysis]

    F[HR Positioning Strategist Agent]
    G[Recruiter Optimization Strategy]

    H[Final HR-Ready Report]

    A --> B
    B --> C
    C --> D
    D --> E
    E --> F
    F --> G
    G --> H

Workflow

flowchart LR

    A[GitHub Profile URL]
    --> B[Profile Analysis]

    B --> C[Repository Review]

    C --> D[Skills Assessment]

    D --> E[HR Positioning Strategy]

    E --> F[Improvement Recommendations]

    F --> G[Final Report]

Technology Architecture

graph TD

    A[GitHub Profile]

    B[ScrapeWebsiteTool]

    C[GitHub Profile Analyst Agent]

    D[Google Gemini 2.5 Flash]

    E[HR Positioning Strategist Agent]

    F[CrewAI Sequential Process]

    G[Final HR Report]

    A --> B
    B --> C
    C --> D
    D --> E
    E --> F
    F --> G

Agents

| Agent | Role | Tools | |---|---|---| | github_profile_analyst | Technical recruiter persona — reviews profile, repos, docs, and contribution graph | ScrapeWebsiteTool | | hr_positioning_strategist | Career strategist — translates technical signals into recruiter-friendly recommendations | ScrapeWebsiteTool |

Both agents are powered by gemini/gemini-2.5-flash via the crewai.LLM wrapper and the Google GenAI integration.

---

Prerequisites

• Python >=3.10, <3.14
• uv — fast Python package manager
• CrewAI CLI — project scaffolding and execution
• A Google Gemini API key (free tier available at aistudio.google.com)

---

Installation

1. Install uv

pip install uv

2. Install the CrewAI CLI

uv tool install crewai-cli

3. Add the CLI to your PATH (Windows PowerShell)

$env:PATH = "C:\Users\<your-username>\.local\bin;$env:PATH"

On macOS/Linux, add ~/.local/bin to your shell profile (.bashrc, .zshrc, etc.).

4. Verify the installation

crewai --version

5. Clone the repo and install dependencies

git clone https://github.com/<your-username>/github-profile-analyst-agent.git
cd github-profile-analyst-agent

crewai install

6. Add Google Gemini support

uv add "crewai[google-genai]"

---

Configuration

Environment Variables

Create a .env file in the project root (or inside .venv/ as scaffolded):

GEMINI_API_KEY=your_google_gemini_api_key_here

The github_profile_analyst agent explicitly reads GEMINI_API_KEY via os.getenv. The hr_positioning_strategist inherits it from the environment automatically through the LiteLLM backend.

Set the Target GitHub Profile

Edit src/github_profile_hr_optimizer/main.py and update the inputs dictionary:

def run():
    inputs = {
        'github_url': 'https://github.com/<target-username>'
    }
    GithubProfileHrOptimizerCrew().crew().kickoff(inputs=inputs)

---

Running the Project

# Standard run
crewai run

# Alternate run via uv (recommended to ensure the correct virtual env)
uv run crewai run

The crew will scrape the target profile, run both agents sequentially, and print the analysis + recommendations to the console.

---

CLI Reference

# Run the crew
crewai run

# Train the crew over N iterations
crewai train -n 5 -f training.json

# Replay from a specific task ID
crewai replay -t <task_id>

# Test the crew
crewai test -n 3 -m gemini/gemini-2.5-flash

# View last task outputs
crewai log-tasks-outputs

# Reset agent memory
crewai reset-memories -a       # All memory
crewai reset-memories -s       # Short-term only
crewai reset-memories -l       # Long-term only

---

Customization

| File | What to modify | |---|---| | config/agents.yaml | Agent roles, goals, backstories | | config/tasks.yaml | Task descriptions, expected output format | | crew.py | LLM model, tools, process type (sequential / hierarchical) | | main.py | Input values, custom entry points | | tools/custom_tool.py | Add your own tools (GitHub API, LinkedIn scraper, etc.) | | knowledge/user_preference.txt | Persistent user context passed to the crew |

Switching Models

To use a different Gemini variant or another provider, update crew.py:

llm=LLM(
    model="gemini/gemini-2.0-flash",   # or "openai/gpt-4o", "anthropic/claude-3-5-sonnet" etc.
    api_key=os.getenv("GEMINI_API_KEY")
)

CrewAI uses LiteLLM under the hood, so any LiteLLM-supported provider works with the same pattern.

---

Dependencies

| Package | Version | Purpose | |---|---|---| | crewai | 1.14.4 | Core multi-agent framework | | crewai[google-genai] | — | Google Gemini API integration | | crewai[tools] | — | ScrapeWebsiteTool and other built-in tools | | crewai[litellm] | — | Unified LLM provider routing | | litellm | <=1.82.6 | Pinned for compatibility |

Dependency resolution and locking are managed by uv. Do not edit uv.lock manually.

---

Troubleshooting

crewai not found after install Add the uv tools bin directory to your PATH and restart the terminal.

# Windows PowerShell
$env:PATH = "C:\Users\<username>\.local\bin;$env:PATH"

GEMINI_API_KEY not set error Ensure the .env file exists at the project root and contains a valid key. The .venv/.env file is also loaded automatically by the CrewAI scaffold.

Rate limit or quota errors from Gemini The free tier of Gemini API has per-minute request limits. Add delays between runs or upgrade to a paid tier if hitting limits consistently.

Scraping blocked by GitHub GitHub rate-limits unauthenticated scrapers. For reliable access, consider adding a GithubSearchTool from crewai_tools with a personal access token.

---

Roadmap

• [ ] Accept GitHub URL as a CLI argument at runtime
• [ ] Export analysis report to a Markdown or PDF file
• [ ] Add GitHub API tool for authenticated, structured data access
• [ ] Support batch analysis of multiple profiles
• [ ] Add a web UI (Streamlit or Gradio) for non-technical users

---

License

This project is licensed under the MIT License. See LICENSE for details.

---

Acknowledgements

• CrewAI — multi-agent orchestration framework
• Google Gemini — LLM powering both agents
• Astral uv — fast, reliable Python package management