Hello, everyone. This discussion post aims to gather insights and feedback from the community about a planned treasury proposal for dAppForge. dAppForge is an AI-powered plug-in for Polkadot developers that reduces development time.
Summary of Important Points
The developer experience on Polkadot is the ecosystem's #1 problem. Substrate and Ink! are very powerful frameworks, but they are very difficult to learn. Because of this, Polkadot developers always struggle to build better applications. The bad developer experience also makes it harder for new developers to enter the ecosystem. Polkadot needs a better developer experience to grow the developer ecosystem and enable developers to build better applications faster. Better applications will attract more users, which is a problem Polkdot has always had. One solution to the DevEx problem is dAppForge.
dAppForge is an AI-powered plug-in for Polkadot developers that reduces development time. The AI model will be explicitly fine-tuned for substrate and ink! We received $100,000 in funding from Angel Investor to build LLM's first iteration and the alpha version of the plug-in. You can also download the alpha version of the plugin from the VS code extension marketplace.
Since this is the alpha version, there is still much fine-tuning left to do. The funding received enabled us to complete the first of four milestones to build dAppForge V 1.0. dAppForge version 1.0 is divided into 4 milestones: Milestone 1: the first iteration of LLM and the alpha version of the VS code plugin. This was completed, but the product is at a level where developers won’t see a lot of value. This is because the model needs to be further fine-tuned. Milestone 2: This is the beta version of the product. It includes fine-tuning the LLM model, integrating ink!, and building a chat feature. This is where developers will see value in using the product. We are applying to the treasury to fund milestone 2. Milestones 3 and 4: We plan to integrate the Pop network to further improve the DevEx for ink developers! We will also work on implementing other features, such as reference documentation and pre-written prompts for runtime templates. The main task for these milestones will be releasing our API freemium model. The API will have a free tier, and developers will have access to basic code suggestions and a foundational knowledge graph to help them build their Polkadot projects. The subscription-based premium API will offer extra features, such as:
- Enhanced AI: More accurate, contextually relevant code suggestions powered by advanced models.
- Personalized Learning: Train the AI on your codebase for tailored suggestions.
- Up-to-date Knowledge Graph: Access the latest Polkadot knowledge and best practices with our regularly updated knowledge graph.
We completed milestone 1. Here is the GitHub link to the repository. It will be migrated to the dAppForge repository in the future, and the code for the plug-in will also be open-sourced. We are now seeking treasury funding for milestone two. Milestone 2 funds the product's beta version, which developers will see value in using. The code suggestions will be more accurate and help reduce some of the developers' pain points when building with Substrate. We will also incorporate ink!, making it easier for developers to use smart contracts to test their ideas quickly.
Neurons Lab, which developed the first version of AI in the previous development phase, has been commissioned again. As an AWS partner, Neurons Lab can leverage the AWS AI Production Ready Funding Program to obtain $75,000 in co-funding from AWS. This results in a total AI development budget of $150,000, with costs split equally between AWS and dAppForge. This will enable us to build a robust beta verison with a better fine-tuned AI model. AWS credits cover all the AWS hosting costs. The $75,000 does not include building the chat feature and integrating the AI model into the plug-in. Those are separate development costs. The comprehensive cost for milestone 2 is $105,000 and will take 12 weeks to complete.
The beta version of the product will be free to everyone. After completing milestone 2, developers will see the plug-in's value. The core of dAppForge will be open-source, meaning anyone can download the model on their local devices. After completing milestone 2, we plan to ask the treasury for funds for milestone 3 and then milestone 4.
Milestones 3 and 4 will each take about 12 weeks to complete and cost around $100,000 to $115,000 each. After completing milestone 4, dAppForge version 1.0 will be ready. The main part of milestones 3 and 4 will be implementing the API freemium feature. The subscription-based premium API is part of our monetization strategy. The cloud-based premium features will support dAppForge's ongoing development and improvement. This helps ensure that the free, open-source version is available and improved for everyone. The monetization strategy also eliminates the need to ask for continuous treasure funding after milestone 4.
Here is the link to the whole proposal
Looking forward to feedback from the community!
We feel partial to copilot development for Ink! And Polkadot-sdk developers. It should be useful for all types of auto-complete features as well as a learning tool for newcomers.
We have shared this feedback already in private but we would like to know if dAppForge will also be available outside of AWS. Something like Hugging Face or standalone tool so that developers are not bound to AWS.
Moreover, we would like to hear from developers within the community who are interested in becoming users of the Ink! / Polkadot-sdk Copilot tool.
In the meantime we have abstained but we will consider a revote once we have gathered proactively some feedback.