Last updated 5 months ago
Cardano governance lacks an efficient way to explore and compare dRep voting rationales, making informed participation time-consuming, fragmented, and inaccessible to most community members.
We’ll build an AI assistant that extracts and clusters voting rationales, enabling questions, insight retrieval, and practical use of collective wisdom in Cardano governance.
This is the total amount allocated to Governance AI Assitant.
Please provide your proposal title
Governance AI Assitant
Enter the amount of funding you are requesting in ADA
90203
Please specify how many months you expect your project to last
12
Please indicate if your proposal has been auto-translated
No
Original Language
en
What is the problem you want to solve?
Cardano governance lacks an efficient way to explore and compare dRep voting rationales, making informed participation time-consuming, fragmented, and inaccessible to most community members.
Supporting links
Does your project have any dependencies on other organizations, technical or otherwise?
Yes
Describe any dependencies or write 'No dependencies'
This project depends on the Koios API to access structured, indexed governance data from the Cardano blockchain. While technically feasible to extract this data directly, doing so would require running a full node and building a custom indexer — incurring substantial complexity and costs. Koios provides a mature and reliable interface that allows us to focus on the unique value of our project: semantic analysis and AI-powered assistance for decentralized governance.
Will your project's outputs be fully open source?
Yes
License and Additional Information
MIT license
Please choose the most relevant theme and tag related to the outcomes of your proposal
AI
Mention your open source license and describe your open source license rationale.
We are releasing this project under the MIT License. This permissive open-source license encourages widespread reuse, modification, and integration of our code across the Cardano ecosystem and beyond. Our rationale is to maximize community adoption, transparency, and collaboration. By keeping the license simple and compatible, we ensure that other builders—technical or not—can freely explore, extend, or repurpose the governance dataset and AI tools we are developing.
How do you make sure your source code is accessible to the public from project start, and people are informed?
The project will be open source under the MIT License. A public GitHub repository will be created from the start, containing initial documentation and structured folders. All code and updates will be committed regularly. The repository link will be shared via social media. Milestones will include verifiable open-source deliverables, ensuring transparency and accessibility throughout the development process.
How will you provide high quality documentation?
All code will be published under the MIT License on a public GitHub repository from day one. Documentation will be collaboratively created and maintained on HackMD, ensuring clear, accessible, and version-controlled technical materials. GitHub Issues and Discussions will be used to track contributions and feedback, while HackMD will serve as the primary space for living documents and developer guides.
Please describe your proposed solution and how it addresses the problem
🧩 Solution Description
This proposal aims to develop an open-source infrastructure that simplifies how members of the Cardano community—especially dReps, voters, and researchers—access, understand, and compare on-chain rationales (i.e., justifications) attached to votes on Governance Actions / Proposals. While rationales are often embedded in metadata and stored in IPFS or GitHub, there is currently no streamlined way to navigate or extract insights from them at scale. This creates friction in collective decision-making and leaves valuable community knowledge underutilized.
We propose building a modular system consisting of a backend data extraction pipeline, a structured public dataset of rationales, and a lightweight AI assistant capable of answering governance-related queries using natural language. This system will enable scalable analysis of voting behavior, reveal semantic patterns across dRep rationales, and reduce the cognitive load required to participate meaningfully in governance.
The backend will begin by integrating with the Koios governance API, which allows programmatic access to votes cast on Governance Actions / Proposals submitted via CIP-1694. Each vote object includes a reference to a meta.url or meta.hash, which often points to rationale metadata stored off-chain (commonly in IPFS or GitHub).
We will create a backend service that:
Fetches vote data from the Koios API.
Identifies valid metadata links (IPFS, GitHub, or other repositories).
Retrieves and parses metadata files, extracting relevant content fields (especially body, which often contains the rationale).
Cleans and normalizes text fields for clustering and semantic search.
Stores all parsed data in a structured, public dataset that can be queried by researchers, developers, and governance participants.
To ensure long-term utility and openness, this pipeline will be fully documented and published under the MIT License on GitHub from day one.
The collected rationale dataset will power a lightweight AI assistant capable of responding to governance-related questions through natural language prompts. For example:
“What were the most common reasons for voting NO on GA xyz?”
“How did different dReps justify their votes on the 275M Budget Info Action?”
“Which rationales supported delegating authority to the Interim CC?”
This assistant will be powered by an open-source LLM interface and designed to:
Provide semantic clustering of rationales (e.g., by topic, sentiment, or alignment).
Allow filtering by Governance Action ID, dRep, or vote outcome.
Extract summary insights across dozens or hundreds of rationales.
All prompts, output formats, and models used will be carefully documented. The assistant will be offered through a minimal UI (either standalone or embedded within community tools).
All outputs will be released under the MIT License, including:
The rationale dataset (in CSV, JSON or other machine-readable formats)
The backend extraction pipeline
Frontend components (if built)
AI assistant code and prompts
Full documentation on GitHub and HackMD
This ensures the project is freely reusable, forkable, and auditable, enabling both transparency and community-driven improvements. Early access and technical documentation will be promoted via social media and governance forums, with community calls for feedback and testing.
While governance explorers like Koios, Tempo Vote, Adastat, and GovTool already offer access to on-chain voting data, they are primarily focused on transactional records—such as who voted, when, and how. These platforms do not aggregate the semantic content of rationales nor offer tools for clustering, summarizing, or interpreting justifications in a meaningful and scalable way.
During our research, we conducted a survey of publicly available tools and GitHub repositories, and found no evidence of any team currently building or planning a similar semantic rationale aggregation system. We recognize the importance of avoiding redundancy and inefficient duplication of work, especially when public treasury funds are involved.
As such, we emphasize that this project is designed to complement, not compete with, existing governance infrastructure. If we become aware of other initiatives or individuals working toward similar goals, we will be open to collaboration, code sharing, or integration efforts to maximize collective impact and reduce fragmentation.
Our approach is distinct in both technical scope and intent, focusing on building an open dataset and a lightweight assistant that enables "wisdom of the crowds" through structured insight mining, rather than simply visualizing raw governance data.
If successful, this project may expand into:
Sentiment analysis of rationale clusters
Delegation dashboards based on alignment of reasoning, not just voting behavior
Multilingual support for non-English rationales
Automatic summaries of each Governance Action
These directions will be considered based on feedback from early users and the broader dRep ecosystem.
Please define the positive impact your project will have on the wider Cardano community
🌍 Impact
Unlocking the Full Value of On-Chain Governance
Cardano is at the forefront of decentralized governance, with CIP-1694 bringing powerful tools like on-chain voting, metadata-linked rationales, and delegated representation (dReps). Yet, paradoxically, the very volume and complexity of this data make it inaccessible to many of the stakeholders it was meant to empower.
This proposal addresses a critical blind spot: governance overload. For every Governance Action (Proposal), there can be dozens or even hundreds of voting transactions, each optionally linked to rationale metadata, IPFS documents, and GitHub repositories. These rationales contain the core justifications, concerns, tradeoffs, and priorities that inform the collective will of the community.
But here’s the uncomfortable truth: almost no one reads them. Not community members. Often not even dReps.
⏱️ Time is the Real Bottleneck
Let’s be honest: if you’re a dRep or community member who actually wants to understand what’s going on before voting, you're looking at hours of unstructured reading per Governance Action. Multiply that by 10, 20, 30 proposals per epoch, and the result is clear—manual diligence is unsustainable.
Even the most committed participants must choose between being informed and being efficient. Most don’t choose at all—they just don’t participate meaningfully.
This project directly addresses that time-cost barrier by delivering a governance assistant powered by AI—capable of surfacing summarized insights, semantic clusters, and tailored answers to participant questions.
Instead of manually opening rationale links, reading 90 separate documents, and cross-referencing stake-weighted trends, the user will be able to ask simple, targeted questions like:
What were the main reasons given by dReps who voted NO?
How many rationales cited a lack of transparency or budget detail?
Are there any technical concerns raised repeatedly?
The assistant works—so you don’t have to.
🤝 Collective Intelligence at Scale
This project does not replace human judgment. It amplifies it.
By organizing and analyzing rationales semantically, the platform unlocks the latent collective intelligence hidden in voting metadata. It helps the Cardano ecosystem move toward the "wisdom of the crowds" model—not just by aggregating stake, but by aggregating arguments, insights, and divergent reasoning paths.
Today, rationales are scattered across metadata hashes, IPFS documents, markdown files, and off-chain links. They’re underutilized, inaccessible, and lost to time.
With this proposal, they become searchable, comparable, interpretable, and valuable.
__
🔍 Incentivizing Better Governance Practices
When rationales become more visible, interpretable, and influential, a subtle but powerful incentive emerges: people start writing better rationales.
Participants—especially dReps—will know that:
Their rationales are being read, interpreted, and indexed.
Their arguments may influence delegators, other dReps, and reputation.
Poor or empty rationales will be visible in aggregate analysis.
This type of soft pressure can improve the quality of deliberation without requiring coercive mechanisms or forced formats. Good governance becomes a cultural norm, not just a checkbox.
🧠 Lowering the Cognitive Barrier for Newcomers
For many new Cardano users, governance feels intimidating. The learning curve is steep, and the workflow to engage is fragmented:
Find Governance Actions.
Read the action metadata.
Decode the context.
Find and read dozens of rationale files.
Compare positions.
Decide how to vote.
This proposal simplifies that dramatically. A newcomer could ask:
“What is this Governance Action about, and what are the main arguments for and against it?”
And get a clear, informed response without navigating technical tooling or digging through metadata hashes.
This lowers the barrier to entry and helps Cardano maintain an inclusive, participatory governance model as it grows.
🔄 Reusable Infrastructure for Future Governance Cycles
The architecture of this project is designed with modularity and long-term use in mind:
The governance rationale database can serve as a public API for other builders.
The semantic clustering engine can be adapted to other governance ecosystems or Catalyst.
The AI assistant can be expanded to support other languages, filters, or data sources.
In short: this is not a single-cycle experiment. It’s a tooling foundation that will continue to add value as governance volume, complexity, and participation scale up.
📈 A New Layer of Transparency and Accountability
When rationales become semantically accessible, patterns emerge:
Which concerns are raised most often?
Which proposers receive repeat criticism?
How often do budget objections correlate with NO votes?
This opens the door to new types of transparency:
Proposers can address common concerns early.
Delegators can track how their dReps think and vote.
Analysts can identify systemic governance bottlenecks.
It creates a feedback loop between rationales and decision-making, enhancing accountability without requiring protocol-level changes.
😬 And Yes—Let’s Talk About dReps
Some dReps are excellent: transparent, analytical, consistent, and communicative.
Others... less so.
Many vote on dozens of Governance Actions with no visible rationale, or copy-paste statements, or links to generic documents. Sometimes, even those are missing.
We believe this tool will expose those patterns, not through shame, but through visibility.
When rationales are indexed and summarized:
Delegators can choose to reward diligence with delegation.
Lazy or opportunistic behavior becomes easier to detect.
Community expectations for performance become clearer and enforceable—socially, if not programmatically.
In short: this project creates a mirror for governance. And sometimes, mirrors are uncomfortable. But they’re necessary.
--
🎯 Conclusion: Scalable, Insightful, Humane Governance
Cardano’s governa---nce is ambitious and revolutionary. But if we want it to succeed, we must build tools that help people make decisions—not just store them on-chain.
This proposal directly contributes to:
Scalability, by lowering time and attention costs.
Participation, by making rationale access easier and more meaningful.
Quality, by surfacing insights and enabling argument comparison.
Transparency, by making governance more observable and analyzable.
It does not aim to replace debate or reflection. It aims to make them feasible again, even at scale.
What is your capability to deliver your project with high levels of trust and accountability? How do you intend to validate if your approach is feasible?
🚀 Project Leadership & Proven Delivery
Tis project will be led by Rodrigo Pacini, a recognized governance contributor in the Cardano ecosystem with a strong delivery record:
Funded Proposer in Project Catalyst Fund 10, with all deliverables completed and properly reported.
Delegated Representative (dRep) since early 2025, with more than 40 on-chain votes and fully published rationales on GitHub.
Founder of the Agora initiative, producing consistent governance research, proposal reviews, and voter guidance materials.
Serves as Milestone Reviewer and Moderator in Project Catalyst since Fund 11, overseeing proof-of-delivery submissions and maintaining quality control on proposal execution.
👥 Team Capabilities
This proposal brings together a seasoned team with diverse and complementary expertise:
🧑💻 Weslley da Costa — Main Developer
Lead developer responsible for core system architecture and implementation.
Experienced in AI agent design and process automation, having previously automated business workflows using artificial intelligence.
Currently developing a conceptual mockup of the assistant to be used in this project — which will later support Catalyst dReps and stakeholders by generating structured governance digests.
Works closely with the research team to implement data pipelines, dashboard infrastructure, and automated reporting tools, ensuring scalability and precision.
🧠 Victor Corcino - Developer
Veteran Community Advisor (vCA) and Proposal Mentor in Project Catalyst.
Elected representative for the 1st Catalyst Circle, representing Community Advisors in governance discussions.
Co-creator of the Community Tools, supporting proposers, CAs, vCAs, and voters through self-service platforms.
Catalyst Swarm core member, actively involved in grassroots coordination and educational outreach.
Background in engineering, science, development, and hands-on education, contributing to high-impact community engagement.
🧩 Phil Khoo — UI/UX Developer
Veteran Community Advisor (vCA) with deep involvement in Catalyst processes.
Co-creator of the Community Tools (including the Community Landing Page), used widely by proposers and voters.
Background in UI/UX design, finance, and business, with a strong focus on aligning tooling design with governance objectives.
🔍 Feasibility & Validation
The proposal includes a structured Statement of Milestones, with clearly defined deliverables, acceptance criteria, and public evidence requirements.
Scope is grounded in historical governance data (approx. 64 Governance Actions and ~200 budget proposals in the last 12 months), ensuring that the volume and complexity of the work are realistic and manageable.
Deliverables are modular and adaptable to different types of governance decisions (on-chain, off-chain, budgetary), enabling strategic flexibility.
All results will be published in open GitHub and Gitbook repositories, with version tracking, clear attribution, and full transparency.
Milestone Title
Data Infrastructure & Solution Architecture
Milestone Outputs
Objective: Establish the technical foundation of the project: data sources, rationale formats, ingestion pipeline, and overall assistant architecture.
Deliverables:
Full architecture map: input (Koios/metaUrl) → parsing → embedding → clustering → querying
Test script to download and process rationales via metaURL/IPFS
Initial SQLite database schema (Governance Action + Rationales)
Document explaining expected metadata and rationale structure
Conceptual interface mockup (already started by Weslley)
Technical blueprint of the data flow and assistant architecture
Acceptance Criteria
Architecture published on GitHub (with .md and .pdf versions)
Repository with functional extraction and parsing scripts
Conceptual UI mockup showing at least 3 simulated user flows
Technical architecture diagram (focused on data pipeline)
Evidence of Completion
Github repository with architecture files, diagrams, and scripts
Markdown + PDF publication explaining technical decisions
Screenshots of the mockup or a short video demo
Delivery Month
2
Cost
20000
Progress
20 %
Milestone Title
Semantic Database and Clustering
Milestone Outputs
Objective: Build the assistant’s semantic foundation: ingestion, embeddings, and similarity-based grouping.
Deliverables:
Full ingestion and parsing pipeline using Koios and metaUrl
Rationale embeddings using OpenAI via LangChain
Operational vector database (Qdrant) with indexed rationales
Structured SQLite database organized by Governance Action ID
Clustering algorithm implementation (HDBSCAN)
Metadata storage of clusters with vote position and rationale ID
Acceptance Criteria
Vector database with embeddings for at least 50 Governance Actions
Clusters showing thematic cohesion in at least 3 test cases
Scripts versioned and documented on GitHub
Tests showing relevant context being returned by semantic queries
Evidence of Completion
Repository with ingestion and embedding pipeline
SQLite dumps with clustered data
Execution logs with query examples and results
Delivery Month
4
Cost
40000
Progress
40 %
Milestone Title
MVP Assistant with Natural Language Queries
Milestone Outputs
Objective: Deliver a functional MVP assistant with web interface and contextualized responses.
Deliverables:
Python backend integrated with Qdrant + SQLite
Integration with ChatGPT via LangChain and Retrieval-Augmented Generation (RAG)
Prompt system tailored to different types of Governance Actions (funding, CCA, etc.)
Web interface with Governance Action ID filter and natural query input
Display of summarized output with source rationale references
Acceptance Criteria
Functional MVP accessible via browser (Streamlit)
At least 10 successful test queries across varied prompts
Fully open-source GitHub repo with backend logic and prompts
Basic setup and usage instructions in repository
Evidence of Completion
Public or restricted-access MVP link
Screenshots and video demonstrating working interface
Documented codebase and technical README
Delivery Month
8
Cost
20000
Progress
60 %
Milestone Title
Community Testing and Refinement
Milestone Outputs
Objective: Collect real-world feedback from dReps, adjust outputs, reduce errors, and polish UX.
Deliverables:
Testing script and feedback form (Google Forms or Notion)
Selection of 5–10 community testers (preferably dReps)
Prompt refinements and adjustments to display logic or cluster tuning
Logging and handling of bugs, hallucinations, or incorrect answers
Usability and reliability report based on real usage
Acceptance Criteria
Feedback collected from at least 5 distinct testers
At least 3 bugs or significant adjustments documented and resolved
Prompt system iterated based on real-world findings
Changelog and feedback summary published
Evidence of Completion
Logs from feedback sessions
Updated prompt versions and logic
Summary report (Markdown + PDF) of improvements and feedback received
Delivery Month
10
Cost
20000
Progress
80 %
Milestone Title
Final Deployment + Full Documentation Release + Project CLose-Out
Milestone Outputs
Objective: Release the final assistant version with comprehensive documentation, public code repositories and Project Close-Out reports
Deliverables:
Final deployment of the assistant (public Streamlit app or downloadable container)
GitHub repository fully organized with README, setup scripts, and modularized components
User guide tailored for dReps with example queries and use cases
Technical documentation covering pipeline, prompt system, and database structure
Project Close-Out Video and Report
Acceptance Criteria
Evidence of Completion
Link to public deployment
Well-documented GitHub repository with setup instructions
Final project report (Markdown + PDF) summarizing the initiative +PCV and PCR
Delivery Month
12
Cost
10203
Progress
100 %
Please provide a cost breakdown of the proposed work and resources
Personnel Costs
0.5 PTE - Software Developer and AI Prompt Engineer = $30,000 ($50/hour)
0.5 PTE - AI Prompt Engineer = $14,000($50/hour)
External developers & expert consultancy = $10,000
Project management
Milestone Statement (SoM): $500
Proof of Life: $50
Proof of Achievement (5 × 8h @ $40/h): $1,600
Project Close-Out Report & Video (50h @ $40/h): $2,000
Team Coordination (12 × $300): $3,600
🛠️ Tools & Subscriptions (ChatGPT, N8N, Google Meet, X Premium)
$100/month x 12 months = $1,000
Subtotal $62750
Contingency 15%
Total 72,162 or 90,203 ADA
How does the cost of the project represent value for the Cardano ecosystem?
Lorem ipsum dolor sit amet, consectetur adipiscing elit. Sed do eiusmod tempor incididunt.Lorem ipsum dolor sit amet, consectetur adipiscing elit. Sed do eiusmod tempor incididunt.
Lorem ipsum dolor sit amet, consectetur adipiscing elit. Sed do eiusmod tempor incididunt.
Terms and Conditions:
Yes
Rodrigo Pacini – Project Management, Prompt Engineer and QA
Researcher, active dRep, and project manager with extensive experience in Cardano governance. Active in Project Catalyst since Fund 2 as Veteran Community Advisor, Reviewer Level 2, Moderator, and Funded Proposer (Fund 10). Over 1,000 proposals reviewed, 3,000+ community reviews moderated, and 100+ milestones validated. Author of nine funded Challenge Settings mobilizing ~$3M in treasury funding. Former Cardano Ambassador Moderator.
Founder of AGORA,%5D(https://agora-cardano.carrd.co/),) an independent research and advocacy initiative focused on improving governance quality and decentralization in Cardano through frameworks, analysis, and community education.
Project manager of DReps LATAM – Brasil – Exploration & Community Sensing and co-lead of Manifesto Cardano Brasil,%5D(https://manifestocardanobrasil.gitbook.io/manifesto),) a funded research project exploring political identity, representation, and governance in the Brazilian Cardano community. Has participated in over 40 on-chain governance votes as a dRep. Background in economics, blockchain, and DeFi research since 2018, with a BTech in Naval Construction.
🔗 Links:
Weslley da Costa — Main Developer
Lead developer responsible for core system architecture and implementation.
Experienced in AI agent design and process automation, having previously automated business workflows using artificial intelligence.
Currently developing a conceptual mockup of the assistant to be used in this project — which will later support Catalyst dReps and stakeholders by generating structured governance digests.
Works closely with the research team to implement data pipelines, dashboard infrastructure, and automated reporting tools, ensuring scalability and precision.
🔗 Links:
https://www.linkedin.com/in/weslley-da-costa-dias/
**Victor Corcino **- Developer
Veteran Community Advisor (vCA) and Proposal Mentor in Project Catalyst.
Elected representative for the 1st Catalyst Circle, representing Community Advisors in governance discussions.
Co-creator of the Community Tools, supporting proposers, CAs, vCAs, and voters through self-service platforms.
Catalyst Swarm core member, actively involved in grassroots coordination and educational outreach.
Background in engineering, science, development, and hands-on education, contributing to high-impact community engagement.
🔗 Links:
https://www.linkedin.com/in/victorcorcino/
Phil Khoo — UI/UX Developer
Veteran Community Advisor (vCA) with deep involvement in Catalyst processes.
Co-creator of the Community Tools (including the Community Landing Page), used widely by proposers and voters.
Background in UI/UX design, finance, and business, with a strong focus on aligning tooling design with governance objectives.
**Javier Acosta **- AI & Technical Advisor
Javier's capability to deliver the Exura project goal is demonstrated through structured approach to development and deep understanding of the Cardano ecosystem and architecture, DeFi protocols, and the technical requirements for building a robust backend.
🔗 Links:
https://www.linkedin.com/in/sebastian-javier-acosta-stelzer-233924209/