Storacha AI Agents Toolkit – March 20, 2025 – Eth Elder Ease – PLDG 2

[seetadev]

Storacha AI Agents Toolkit – March 20, 2025 – Eth Elder Ease – PLDG 2

Contributors

- @seetadev - @Chaitu-Tatipamula - @aspiringsecurity

Accomplished Tasks

• Developed knowledge base modules for AI agent framework using Storacha as the backend for Eth Elder Ease.
• Implemented secure data storage and retrieval for medical invoices for building a RAG knowledge base module.
• Created a real-time retrieval mechanism for decision-making.
• Established cross-platform communication for building knowledge base and AI agents (mobile + desktop).
• Developed token-gated access of AI agent using ERC-20 utility token on Filecoin. Launched the utility token on Sushiswap DEX.
• Configured libp2p gossipsub + QUIC for P2P message exchange between AI agents.
• Built a knowledge base using Storacha-backed RAG both for Web2 and Web3 AI agents. Compiled Alexa Skill for Web2 Agents.
• Integrated token gated access of save, email and print functions for medical invoice knowledge base using ERC-20 utility token.
• Mentored and reviewed publish and subscribe (pub-sub) implementation for P2P agents using Py-libp2p (code merged to main).
• Integrated LIT Protocol with Storacha to enable secure KYC verification for onboarding and managing manual assistance providers for elderly care.

Update

• Eth Elder Ease Integration:

  • Built AI agents to provide personalized assistance (e.g., medication reminders, emergency response).
  • Integrated secure data handling using Storacha’s decentralized storage.

• libp2p Integration:

  • Established secure peer-to-peer (P2P) communication using libp2p modules (QUIC + gossipsub).
  • Improved handshake reliability with Noise protocol adjustments.

• Token-Gated Access:

  • Successfully tested ERC-20 token-based access for secure retrieval from the knowledge base.
  • Used Filecoin’s FEVM for smart contract deployment.

• Challenges:

  • libp2p stream reliability issues when handling large real-time data payloads.
  • Latency issues with cross-platform AI agent communication.
  • Noise spec inconsistencies in initial handshake phase.

Demo Video

Questions

• How can we improve libp2p stream reliability for large data payloads?
• Are there best practices for optimizing FAISS-based retrieval over Storacha in high-load scenarios?
• How can we improve the AI agent’s reasoning capacity when accessing large datasets from Storacha?

Next Tasks

• Improve state synchronization between AI agents using libp2p’s gossipsub.
• Enhance real-time retrieval latency with FAISS + Storacha adjustments.
• Optimize token-gated access for role-based permissions in smart contracts.
• Integrate multi-agent collaboration using CrewAI and MetaGPT.

[jamiechicago312]

Take a look at types/options and check out on progress call back and that will give the status as it is uploaded
cc: @alanshaw

Will also look into this too 👍

[jamiechicago312]

https://github.com/storacha/w3up/blob/2ddb4042fee59e6b667057f6e82c711fa9e318ee/packages/upload-client/src/types.ts#L155

[seetadev]

@jamiechicago312 : Thank you so much for sharing this helpful reference from the Storacha's code repository. Appreciate it.

[jamiechicago312]

Of course

Alan shared this further in the discord server, places here as another resource if someone stumbled upon this.

Upload Progress Reporting

onUploadProgress callback can be passed as an option to uploadFile, uploadDirectory or uploadCAR.

https://github.com/storacha/w3up/blob/2ddb4042fee59e6b667057f6e82c711fa9e318ee/packages/upload-client/src/types.ts#L317

Alternatively onShardStored callback is called after a CAR shard has been stored and can be used to communicate progress for larger uploads.

https://github.com/storacha/w3up/blob/2ddb4042fee59e6b667057f6e82c711fa9e318ee/packages/upload-client/src/types.ts#L393C3-L393C16

Shards are max 128MB (although this can also be configured).

Advanced

You can pass a fetchWithUploadProgress option for more fine grained progress reporting in browser environments. It uses XHR for this.

Import and use like so (note: this is browser only).

import { fetchWithUploadProgress } from '@web3-storage/upload-client/fetch-with-upload-progress'

client.uploadFile(new Blob([]), { fetchWithUploadProgress })

[seetadev]

@jamiechicago312 : Hi Jamie. Wish to thank you and Alan for sharing such comprehensive insights and references from the Q&A session.

This is incredibly helpful, and we are delighted to learn about the level of functionality and flexibility provided by the @web3-storage/upload-client. The flexibility of this client opens up numerous possibilities, especially for applications involving large data sets or bandwidth-constrained environments. The modularity with CAR shards and UnixFS encoding can be instrumental in building performant and resilient storage solutions.

I’ll explore the provided links further and experiment with the advanced options like fetchWithUploadProgress and client-side shard tracking. If I encounter any edge cases or have suggestions for improvements, I’ll be sure to share them with the team.

Thank you once again for sharing here. This is indeed a set of very helpful references.