Alfred is an AI-powered assistant designed to create impactful, measurable, and actionable bounties for climate and sustainability initiatives. Built with a modular multi-agent architecture, Alfred simplifies the traditionally manual process of planning, executing, and refining bounties, empowering funders and local communities to drive meaningful change.
See Paragraph for posts detailing the evolution of Alfred.
The Atlantis team has witnessed firsthand the devastating impacts of climate shocks in India, disrupting vital systems like water, energy, and food. The core challenge in addressing climate change isn’t technology—it’s coordination. Our pilot in rural India showed the potential of decentralized solutions to enhance resilience, affordability, and transparency. However, gaps in coordination, trust, and verification persist. Blockchain provides a transparent, scalable framework to align global resources with local climate action, empowering communities to take ownership and drive meaningful impact.
Impact Miner by Atlantis is an on-chain bounties platform (Android, iOS), that connects climate impact funders, sustainability organizations, and green gig workers. It ensures transparent, verifiable impact reporting, affordable certification, and efficient fund delivery. Contributions are tracked through NFTs backed by real-world validated actions.
Who benefits?
- Impact Funds: Deploy capital efficiently with real-time, transparent data.
- Sustainability Organizations & Workers: Gain easy access to funding with reduced costs.
- Local Communities: Receive more direct funding for scalable, impactful climate actions.
Creating effective bounties is time-consuming, requiring clear scopes, measurable success metrics, and significant manual effort—especially for newly onboarded communities. Through workshops with funders, we identified key questions to streamline bounty design:
- Is the goal well-defined, realistic, and aligned with local needs?
- Are success metrics measurable and actionable?
- Can the goal be broken into milestones?
- Does the design account for uncertainties and ensure transparency?
- Are all stakeholders included in the process, and do incentives encourage meaningful participation?
- Addressing these challenges is critical to making bounties impactful, scalable, and adaptive to real-world constraints.
Hands-on workshops for designing bounties are impactful but not scalable. To address diverse community needs at scale, we are leveraging AI-powered workflows to embed best practices, contextual insights, and dynamic decision-making into the bounty creation process. Our AI-assisted system will guide funders with:
- Contextual nudges for defining objectives, scopes, and milestones.
- Templates and recommendations informed by collective knowledge.
- A multi-modal assistant offering real-time guidance and seamless interaction. This approach lowers barriers, enhances scalability, and empowers stakeholders to create impactful bounties independently, transforming decentralized solutions into catalysts for collective action.
Alfred is an AI-powered assistant designed to simplify and streamline the creation of impactful bounties for climate and sustainability. Built with a modular, multi-agent architecture, Alfred leverages Agentic AI to transform complex, manual processes into an intuitive and scalable experience. Key Features:
- Planning Agent: Breaks aspirations into actionable milestones with measurable goals.
- Execution Agent: Provides localized, practical solutions to ensure feasibility.
- Feedback & Iteration: Dynamically refines plans for transparency and adaptability.
- Knowledge Base: Offers data-driven insights to enhance decision-making.
Alfred uses LangGraph, a robust orchestration framework, to power its stateful, multi-agent workflows. LangGraph’s key features include:
- Dynamic Flows: Supports complex logic like loops and conditionals for responsive agent behaviors.
- State Persistence: Automatically saves progress, enabling seamless pausing, resumption, and error recovery.
- Human-in-the-Loop: Allows critical interventions and approvals for enhanced decision-making.
- Real-Time Feedback: Provides streaming outputs for better user engagement and responsiveness.
- Ecosystem Integration: Works seamlessly with LangChain and LangSmith to leverage diverse tools and models.
- LangGraph ensures Alfred’s workflows are adaptable, scalable, and optimized for creating effective, measurable bounties.
We’re excited to have you join us in shaping Alfred into a powerful tool for climate and sustainability solutions. As we test our MVP (v0), your feedback and contributions are critical to its evolution. Whether you’re a developer, designer, researcher, or just someone passionate about climate action, there’s a way for you to contribute.
Running Locally
- Clone the repository:
git clone https://github.com/AtlantisDAO1/Alfred.gitcd Alfred - Install dependencies and run the application:
pip install -r requirements.txtstreamlit run ./app/alfred.py - Open your browser and navigate to
http://localhost:8501.
Deploying the App on Streamlit Cloud
- Log in to Streamlit Cloud.
- Create a new app and connect it to your GitHub repository.
- Add the
OPENAI_API_KEYandTAVILY_API_KEYin the "Secrets" section of the Streamlit Cloud dashboard. - Deploy the app.
Let us know about bugs, errors, or confusing user experiences by raising an issue in our Github.