DEV Community: Aech

Build Smarter: AI-Powered Project Planning with Streamlit, CAMEL-AI & Mistral

Aech — Sun, 08 Jun 2025 07:58:15 +0000

In today’s fast-paced software development landscape, planning projects efficiently and effectively is paramount. Manual planning is time-consuming and error-prone, especially for complex projects. To tackle this, I built the AI Project Planning Team, an intelligent project planning tool that leverages the power of Mistral AI and CAMEL-AI frameworks.

This Streamlit-based application helps teams generate detailed technical architecture designs and project timelines interactively. I’ll walk you through the design, implementation, and key features of this project, and show how I used the latest AI tools to make it smarter and more practical.

Why Build an AI-Powered Project Planner?

Traditional project planning requires a lot of domain knowledge and experience. Automating parts of this with AI can help:

Quickly generate scalable and maintainable system architectures
Recommend modern technology stacks and frameworks
Create realistic project timelines based on project type and team size
Allow easy editing and exporting of project plans as PDFs

By combining state-of-the-art language models with a friendly UI, this project aims to democratize expert-level planning assistance for teams of all sizes.

Tech Stack and Prerequisites

The project uses:

Python 3.10+ as the programming language
Streamlit for the web-based user interface
CAMEL-AI Framework to orchestrate AI agents (Technical Architect and Timeline Planner)
Mistral AI large language model as the core engine
xhtml2pdf and Jinja2 for PDF report generation
Markdown rendering for clean display of AI-generated content

You’ll need a Mistral AI API key to run the app, which you can get from the Mistral AI Console.

How It Works

1. User Inputs Project Details

In the sidebar, users enter:

Project name and description
Project type (Web App, Mobile, Desktop, API Service, Data Pipeline, Other)
Preferred technology stack (optional)
Team size
Start date

2. AI Agents Generate Outputs

Upon clicking Generate Project Plan, two specialized AI agents are triggered:

Technical Architect: Designs the system architecture, recommends technologies, and considers scalability and security.
Project Timeline Planner: Breaks down the project into phases, milestones, dependencies, and realistic deadlines.

These agents leverage Mistral AI’s LLM capabilities orchestrated by CAMEL-AI for contextual and iterative planning.

3. Interactive Review and Editing

Generated architecture and timeline are displayed in the app with options to edit text directly. This makes it easy to refine AI outputs before finalizing.

4. PDF Export

Users can download a professional PDF report containing all project details, architecture, and timeline. The PDF supports rich formatting with tables and markdown-rendered content.

Building the System: Key Components & Code

Here are some notable parts of the implementation:

Streamlit app structure with sidebar inputs and state management for user inputs and AI-generated content.
Use of CAMEL-AI’s ChatAgent class to create specialized agents with system messages defining their roles.
Rendering AI responses as Markdown in the UI and converting them to HTML for PDF generation using Jinja2 templates.
PDF creation with xhtml2pdf, handling complex layout like tables and styled text.

Here’s a high-level snippet showing how agents are set up and invoked:

from camel.agents import ChatAgent
from camel.models import ModelFactory
from camel.configs import MistralConfig
from camel.types import ModelPlatformType, ModelType

mistral_model = ModelFactory.create(
    model_platform=ModelPlatformType.MISTRAL,
    model_type=ModelType.MISTRAL_LARGE,
    model_config_dict=MistralConfig(temperature=0.0).as_dict(),
)

architect = ChatAgent(
    system_message=(
        "You are a Technical Architect. Your role is to:\n"
        "1. Design scalable and maintainable system architecture\n"
        "2. Recommend appropriate technologies and frameworks\n"
        "3. Define key components and their interactions\n"
        "4. Consider security, scalability, and performance\n"
        "Be specific and practical."
    ),
    message_window_size=10,
    model=mistral_model,
)

planner = ChatAgent(
    system_message=(
        "You are a Project Timeline Planner. Your role is to:\n"
        "1. Break down the project into phases and milestones\n"
        "2. Create a realistic timeline with deadlines\n"
        "3. Consider team size and complexity\n"
        "4. Identify critical path and dependencies\n"
        "Format timeline clearly."
    ),
    message_window_size=10,
    model=mistral_model,
)

project_context = f"""
Project Name: {project_name}
Project Type: {project_type}
Team Size: {team_size}
Start Date: {start_date}
Preferred Technologies: {', '.join(tech_stack) if tech_stack else 'Not specified'}

Project Description:
{project_description}
"""

arch_resp = architect.step(f"Design the technical architecture for this project:\n\n{project_context}")
timeline_resp = planner.step(f"Create a detailed project timeline for this project:\n\n{project_context}")

architecture_markdown = arch_resp.msgs[0].content
timeline_markdown = timeline_resp.msgs[0].content

Demo: What Does the Output Look Like?

Project Timeline Sample

| Phase          | Duration  | Milestones                        |
|----------------|-----------|----------------------------------|
| Requirements   | 2 weeks   | Finalize requirements document   |
| Design         | 3 weeks   | Architecture design, wireframes  |
| Development    | 6 weeks   | Implement features, write tests  |
| Testing & QA   | 2 weeks   | Unit, integration, user testing  |
| Deployment     | 1 week    | Setup staging and production     |

Final Thoughts and Future Improvements

The AI Project Planning Team demonstrates how modern LLMs and agent orchestration frameworks can transform routine but critical tasks like project planning. By making this accessible through a clean UI, it empowers developers and managers to jumpstart projects confidently.

Possible future enhancements include:

Adding cost estimation and resource allocation AI agents
Integrating with project management tools like Jira or Trello
Supporting multiple languages and frameworks customization
More advanced timeline dependency visualizations

If you’re interested, the full code is available on GitHub. Feel free to explore, contribute, or raise issues!

Thanks for reading! If you try this tool or build on it, I’d love to hear your feedback or questions.

Made with ❤️ using CAMEL-AI and powered by Mistral AI.

I Built a Personal Tech News Assistant That Thinks Like a Team

Aech — Tue, 27 May 2025 12:14:33 +0000

Scraping, summarizing, and saving tech news using multi-agent automation — here's how I made it work with OWL and FireCrawl.

If you're anything like me, keeping up with tech news means juggling 5+ tabs, 3 newsletters, and forgetting half of it anyway.

I wanted something better — something that could scan, summarize, and deliver the latest tech news automatically every day.

So I built a digital assistant that does exactly that — using AI agents from the CAMEL-AI OWL framework and a scraping backend called FireCrawl MCP.

It’s like having two interns: one who surfs the web for news, and another who writes a daily digest for you.

What This Thing Actually Does

Visits sites like TechCrunch, The Verge, Wired
Extracts the latest headlines, summaries, and publish dates
Groups similar stories
Saves the results in a clean Markdown digest

How It Works (The Agent Flow)

Using the OWL framework, I created autonomous agents that talk to each other to get the job done. Here’s the workflow:

+------------------+        +----------------------+        +---------------------+
|  Curator Agent   | ----→  |  Research Agent      | ----→  |  FireCrawl Web API  |
| ("Get me news!") |        | ("I'll scrape it...")|        | (Scrapes tech sites)|
+------------------+        +----------------------+        +---------------------+
         ↑                             ↓
         +------------ Final digest in Markdown ------------+

They collaborate using natural language prompts internally — no need for hardcoded rules.

The Tools Behind It

🧠 OWL (Optimized Workforce Learning): Multi-agent orchestration framework
🔥 FireCrawl MCP: A scraping server that pulls structured data from URLs
🧰 CAMEL-AI: Lets you define agents and simulate workflows
🐍 Python: The glue tying it all together

Getting Started (Setup in 3 Steps)

Clone the OWL repo:

git clone https://github.com/camel-ai/owl.git
cd owl
pip install -r requirements.txt

Then navigate to the use case:

cd community_usecase/Mcp_use_Case
python content_curator.py

Make sure:

You have your .env file with API keys
mcp_servers_config.json is in the same folder as content_curator.py

A Sample Output Looks Like This

### Today’s Top Tech News

- **OpenAI Launches GPT-5 Preview** – The Verge  
  *OpenAI reveals GPT-5 with better memory and deeper reasoning.*

- **Apple Debuts M4 Chips for MacBooks** – TechCrunch  
  *Next-gen silicon promises speed and efficiency for developers.*

- **GitHub Copilot Workspace Rolls Out** – Wired  
  *Your AI coding buddy just got a major upgrade.*

You can set this up to run daily, and send the digest to Notion, WhatsApp, or email.

Customize It: Your Digest, Your Rules 🛠️

This is not a fixed tool — you can personalize every part:

🧾 Change the Task

Edit the prompt in content_curator.py:

default_task = "Curate today's top AI and startup news from Hacker News, Product Hunt, and Reddit r/MachineLearning."

🌐 Use Your Own Sites

FireCrawl supports scraping any website — simply modify the Research Agent’s prompt or provide custom URLs.

⚙️ Edit MCP Server Config

Want to use different MCPs (e.g., WhatsApp, Notion, Telegram)?
You can register and integrate them into OWL by editing mcp_servers_config.json.

Why I Made It

Honestly? I was drowning in tabs and newsletters. I wanted a focused, daily, and clean tech summary that didn’t waste my time.

Also, I wanted to explore what it's like building with OWL — and I was surprised how easy it was to simulate an AI team doing meaningful work.

What's Next?

Here’s what I’m planning to add next:

📆 Run it daily via cron
💬 WhatsApp alerts using the WhatsApp MCP
🔎 Include Hacker News, Reddit, and Dev.to scraping
🧠 Add sentiment + trend detection across headlines

Final Thoughts

This isn’t some fancy AGI project. It’s a practical, nerdy little tool that genuinely makes my mornings better.

If you’re into automating boring stuff with AI, I highly recommend trying out OWL + FireCrawl. You’ll feel like you’re building the future — one YAML config at a time.

📎 GitHub: CAMEL-AI OWL Repository
🧠 Framework Docs: camel-ai.org
💬 Questions or feedback? Drop a comment below!