Retriever searches everything on your phone by describing it: photos, screenshots, documents (PDF, Word, Excel, PowerPoint, ebooks) and videos. Type "electricity bill from March" and it opens the right page; "the moment the cake comes out" opens the video at that moment. It also reads the words in photos and screenshots. Everything runs on the device with Google's EmbeddingGemma 2: no account, no server, no ads. Free and open source, on Android and iPhone.
Hi, I'm Dhruv. I built Retriever because I could never find things on my phone: the PDF bill from March, the whiteboard photo from Monday's meeting, one screenshot among thousands. Gallery search works by date and place, not by what's in the file.
What it does
Search by describing it: "electricity bill from March", "the moment the cake comes out"
Search the words inside photos, screenshots and documents
Search by name, or with a photo
Results open at the matching page or video moment
Turn on People, and a photo of someone finds every photo of them
Private by design
All of it runs on the phone: no account, no server, no ads. Nothing you search is sent anywhere.
On iOS the App Store label is "Data Not Collected". On Android, Google Play services' text recognizer may send Google its own diagnostics, and the privacy policy says so.
The search model, Google's EmbeddingGemma 2, is a one-time download (about 390 MB, more for the Pixel NPU build). After that everything works offline.
The hard part
It wasn't the model, it was the first run: a real phone has tens of thousands of files. The latest update makes search by description ready about 4x sooner (about 85 to 350 photos a minute in my tests on a Pixel Fold), and shows progress as steps that each tell you what you can search now.
One question: where does the search index live, and is it excluded from iCloud / Google backups? AFAIK backups can quietly copy app data off the phone.
My unfindable stuff is screenshots — error messages, code snippets, and receipts I screenshotted instead of saving properly. The on-device part is what sells it to me; I build privacy-first apps, and the "no account, no server" approach is exactly how I think about it too.
Two questions: for videos, is it matching on what's said or what's on screen? And with a 390 MB model plus tens of thousands of files to index on first run, how does it handle battery and storage — is indexing incremental in the background, or does it need one big first pass?
spelling out the play services diagnostics bit in the privacy policy is more honest than most apps bother with. does describing in japanese or vietnamese work as well as english? a lot of our team's screenshots are in those two
About Retriever on Product Hunt
“Search your phone's photos, documents and videos privately”
Retriever launched on Product Hunt on November 17th, 2026 and earned 72 upvotes and 7 comments, placing #35 on the daily leaderboard. Retriever searches everything on your phone by describing it: photos, screenshots, documents (PDF, Word, Excel, PowerPoint, ebooks) and videos. Type "electricity bill from March" and it opens the right page; "the moment the cake comes out" opens the video at that moment. It also reads the words in photos and screenshots. Everything runs on the device with Google's EmbeddingGemma 2: no account, no server, no ads. Free and open source, on Android and iPhone.
Retriever was featured in Android (57.8k followers), iOS (111.2k followers), Privacy (11.5k followers) and GitHub (41.5k followers) on Product Hunt. Together, these topics include over 141.3k products, making this a competitive space to launch in.
Who hunted Retriever?
Retriever was hunted by Dhruv Gupta. A “hunter” on Product Hunt is the community member who submits a product to the platform — uploading the images, the link, and tagging the makers behind it. Hunters typically write the first comment explaining why a product is worth attention, and their followers are notified the moment they post. Around 79% of featured launches on Product Hunt are self-hunted by their makers, but a well-known hunter still acts as a signal of quality to the rest of the community. See the full all-time top hunters leaderboard to discover who is shaping the Product Hunt ecosystem.
Want to see how Retriever stacked up against nearby launches in real time? Check out the live launch dashboard for upvote speed charts, proximity comparisons, and more analytics.
Hi, I'm Dhruv. I built Retriever because I could never find things on my phone: the PDF bill from March, the whiteboard photo from Monday's meeting, one screenshot among thousands. Gallery search works by date and place, not by what's in the file.
What it does
Search by describing it: "electricity bill from March", "the moment the cake comes out"
Search the words inside photos, screenshots and documents
Search by name, or with a photo
Results open at the matching page or video moment
Turn on People, and a photo of someone finds every photo of them
Private by design
All of it runs on the phone: no account, no server, no ads. Nothing you search is sent anywhere.
On iOS the App Store label is "Data Not Collected". On Android, Google Play services' text recognizer may send Google its own diagnostics, and the privacy policy says so.
The search model, Google's EmbeddingGemma 2, is a one-time download (about 390 MB, more for the Pixel NPU build). After that everything works offline.
The hard part
It wasn't the model, it was the first run: a real phone has tens of thousands of files. The latest update makes search by description ready about 4x sooner (about 85 to 350 photos a minute in my tests on a Pixel Fold), and shows progress as steps that each tell you what you can search now.
It's free and open source (Apache-2.0), with Kotlin Multiplatform shared between the Android and iPhone apps: https://github.com/dhruv2295/Retriever
What do you search for on your phone that you can never find? I'd love to hear where Retriever misses.