Vecy: On-device AI & LLM APP for RAG
Fully private AI and LLM w/ documents/images on your device
7 followers
Fully private AI and LLM w/ documents/images on your device
7 followers
Vecy, Android AI APP, allows users to chat with fully-private LLM & files (documents, photos). Local files are automatically processed/indexed on the phone. Users can also directly upload files. 1 Index local files 2 chat w/ LLM & local files 3. image search
Vecy APP: Edge AI, Personalized AI, RAG, database and AI search. On your local device (PCs or Phones), perhaps100,000 or more (document) files should be processed before they can be used effectively with large language models (LLM, local or cloud).
The good news is that you now don't need wifi connection or a powerful desktop/workstation to enjoy AI and LLM. The APP (Vecy) is available in Google Play Store
https://play.google.com/store/ap...
It is basically the same APP announced 5 month ago
which was used to demo the effectiveness of the AI search and database system we have been developing.
The Vecy APP should work on all Android devices, although it has been tested mainly on the newest NPUs, Qualcomm 8gen4 (Samsung S25) and 8gen3 (Samsung S24).
How many documents your device may have? I checked my old laptop (with a 500GB SSD). It has 2,9 million files, among which 200,000 are "data files" (PDFs, Excels, PPTs, DOCs, Texts, etc). Nowadays, phones with 1TB storage are common. There is also trend that people may want to use phones (e.g., folding phones) for work and education, in addition to taking photos and entertainment.
Enterprise AI search will have to deal with (100s of) millions of data files, not just 100,000 data files. Industry practitioners will need a powerful database and AI search tools to process/manage the data files in order to feed them to AI/LLM, to enhance their businesses to the next level. This APP helps demo the database system we have been developing.
Vecy APP will automatically index your photos and documents. Users can also directly upload local files to the APP. For example, one can upload the Anthropic RAG benchmark dataset (90 documents, https://drive.google.com/file/d/... , for your convivence to download them) to the APP and ask the benchmark questions, for example
What feedbacks are used in this fuzzer?
On my Samsung 24 (last year's) phone, I get the correct response from LLM (after selecting the 90 documents for RAG) in about just 1 second (not 1 minute, which you might experience with similar APPs).
In summary, the APP does the following
1. Chat with local LLMs, no connection is needed.
2. Index your photo and document files
3. RAG, chat with local documents
4. Photo search
Considering that many users don't have an android device, we also replicate the functionalities of the APP on the cloud https://chat.vecml.com/ , where users can chat with local (7B) models (including Deep Seek R1) and files uploaded to the cloud. Users also have the option of choosing the size of RAG return, ranging from 400 tokens to 4000 tokens. Some users already discovered that https://chat.vecml.com/ also works well for tables and Excels.
Please feel free to contact us if you enterprises would like to try the AI search, RAG, and database system we have been developing.
Thank you all.
Best regards.
Ping Li
The above comment is from linkedin
https://www.linkedin.com/feed/update/urn:li:activity:7308844726080741376/