{"id":13525916,"url":"https://github.com/lysein/decentAI","last_synced_at":"2025-04-01T06:30:42.712Z","repository":{"id":220757764,"uuid":"752036912","full_name":"lysein/decentAI","owner":"lysein","description":"# decentAI (decentralizedAI)","archived":false,"fork":false,"pushed_at":"2024-02-04T13:56:07.000Z","size":113,"stargazers_count":6,"open_issues_count":1,"forks_count":1,"subscribers_count":1,"default_branch":"fixTfFetchIssue","last_synced_at":"2025-03-21T15:12:17.523Z","etag":null,"topics":[],"latest_commit_sha":null,"homepage":null,"language":"TypeScript","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"mit","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/lysein.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":"LICENSE","code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null}},"created_at":"2024-02-02T21:46:45.000Z","updated_at":"2025-01-27T06:24:20.000Z","dependencies_parsed_at":"2024-02-04T05:28:20.244Z","dependency_job_id":"3b96ef07-f2bc-4de7-9fdb-fd654f901de8","html_url":"https://github.com/lysein/decentAI","commit_stats":null,"previous_names":["carlosarturoceron/decentai","lysein/decentai"],"tags_count":0,"template":false,"template_full_name":null,"repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/lysein%2FdecentAI","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/lysein%2FdecentAI/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/lysein%2FdecentAI/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/lysein%2FdecentAI/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/lysein","download_url":"https://codeload.github.com/lysein/decentAI/tar.gz/refs/heads/fixTfFetchIssue","host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":246596569,"owners_count":20802846,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2022-07-04T15:15:14.044Z","host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":[],"created_at":"2024-08-01T06:01:23.478Z","updated_at":"2025-04-01T06:30:42.341Z","avatar_url":"https://github.com/lysein.png","language":"TypeScript","funding_links":[],"categories":["Decentralized AI"],"sub_categories":["Solana"],"readme":"# decentAI (decentralizedAI)\n\n## How can we think about AI?\nAI: software that presents one or more human capabilities.\n\nSeeing, reading, writing, speaking, analyzing, deciding.\n\n## The Problem: it's centralized\n\nIf you want to run a large scale AI without having the tech knowledge on how to train \u0026 deploy ML models for prediction you rely on consuming them trough something like this:\n\nhttps://bigCorpo.com/coolName/humanLikeBehavior (And they reserve the right of admission)\n\nThis is a HUGE problem:\n\nBecause AI is a catalyst for productivity. The problem is that one individual/company/nation can use AI and another cannot. \n\n* The AI community is aware of this and that's why most of the models are opensource.\n\nIn the most popular AI platform there are more than 10,000 pre-trained open source models.\n\nThe models are already opensource!\n\n`How can we decentralize them?`\n\n## Solution\n\nThe proposed solution leverages AZLE,  a TypeScript/JavaScript runtime for building applications (canisters) on the IC. With AZLE it will possible to build \u0026 deploy REST API's to deliver decentralized AI workloads that inherit the benefits of the IC.\n\nTo use a pretrained model you need to download the model to some storage \u0026 then load it, or load it directly from the web without storing it. \n\nIn python, this is as easy as doing:\n\n```python\nfrom sentence_transformers import SentenceTransformer\n# Specify the model name or path\nmodel_name = 'paraphrase-MiniLM-L6-v2'\n# Download the pre-trained model\nmodel = SentenceTransformer(model_name)\n# Example usage\nsentences = [\"This is an example sentence.\", \"Download a Sentence Transformer model.\"]\nembeddings = model.encode(sentences)\n```\n\nIn this hack we validate if we could write \"on chain\" a pretrained model and then rebuilt it on the running canister with the purpose of generating predictions. If this is possible we would be unlocking a potential milestone towards the decentralization of AI workloads!\n\nSome initial limitations were related to the size of model we could upload, initially tried with 900Mb, after knowing the limitation was closer to 90Mb I was able to find this model \u003c90kB. Note: 90mB models are extreamly capable not an issue, and if capacity keeps increasing it's promising.\n\nModel: https://www.kaggle.com/models/tensorflow/spam-detection/frameworks/tfJs\n\nTechnical conclusions: IT WORKED!\n\n## How to Run:\nFollow the steps at: https://github.com/demergent-labs/azle/blob/main/examples/hello_world/README.md\nCheckout: https://github.com/demergent-labs/azle/tree/main\n\nBuilt with \u003c3 at ETH 5 de mayo\nBig thanks to the ICP team \u0026 all mentors that helped me out.\n```","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Flysein%2FdecentAI","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Flysein%2FdecentAI","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Flysein%2FdecentAI/lists"}