About
AnalogDB provides a large collection of curated analog photographs to users through a REST API interface. Beyond just returning photos, AnalogDB tags each image with the camera and film it was shot on, writes a caption describing it, enables discovery of similar images, adds keyword labels, extracts dominant colors, and allows for filtering, sorting and searching by phrase across all images.

Design
AnalogDB makes use of several technologies and services to enable a full featured product.

Data is scraped from reddit with praw and ingested through a pipeline orchestrated by Dagster. Along the way, an LLM reads each post title to pull out the camera, film and lens, a vision model writes a caption for each image, and keywords and dominant colors are extracted. This pipeline is also responsible for transforming raw images, uploading to AWS S3, and creation of resources through the backend api. Images from S3 are served from CloudFront CDN for quick and reliable delivery.
The core backend application is written in Go and makes use of chi as the HTTP router. It exposes versioned handlers under /v1 that are responsible for parsing authentication headers, filtering incoming requests, querying databases, and returning JSON responses. Posts live in PostgreSQL, with Redis caching hot responses. Upon upload, all images are embedded with CLIP and stored in a Weaviate vector database, which powers both similarity search and searching photos by phrase. The backend is packaged as several docker containers and hosted on a VPS.
The frontend web application is built with Next.js using the App Router, making use of server-side rendering for quick loading pages. Alongside the gallery there are pages for search and for browsing by film and camera. Components come from Mantine and styles are written with CSS Modules.
The backend also publishes analytics events to Kafka, where a small Go consumer writes them into ClickHouse. Everything is instrumented with OpenTelemetry and viewed through Grafana.
API
Full documentation for the API: https://api.analogdb.com/
Example
curl "https://api.analogdb.com/v1/posts?page_size=1"
{
"meta": {
"total_posts": 38246,
"page_size": 1,
"next_cursor": "eyJzIjoidGltZSIsInYiOjE3OTQwMDM1MzUsImlkIjo1NTQ2OX0",
"next_page_url": "/v1/posts?cursor=eyJzIjoidGltZSIsInYiOjE3OTQwMDM1MzUsImlkIjo1NTQ2OX0&page_size=1&sort=time"
},
"posts": [
{
"id": 55469,
"title": "NYC - Halloween 2026 [ Nikon F2 | 28mm f/2.8 | Cinestill 800T ]",
"author": "maxt21",
"permalink": "https://www.reddit.com/r/analog/comments/1xppww4/nyc_halloween_2026_nikon_f2_28mm_f28_cinestill/",
"description": "Had fun shooting Halloween this year in Brooklyn. Last year I did the same and it was a lot of fun. This year was also a lot of fun",
"caption": "A man with yellow face paint and a small anchor decoration on his head smiles at the camera in a subway station.",
"score": 108,
"nsfw": false,
"grayscale": false,
"timestamp": 1794003535,
"sprocket": false,
"camera_make": "nikon",
"camera_model": "f2",
"film_make": "cinestill",
"film_type": "800t",
"film_speed": 800,
"focal_length": 28,
"aperture": "f/2.8",
"images": [
{
"resolution": "low",
"url": "https://i.analogdb.com/4044cdd36cc7e86d8065660c-low.jpeg",
"width": 720,
"height": 720
},
{
"resolution": "medium",
"url": "https://i.analogdb.com/4044cdd36cc7e86d8065660c-medium.jpeg",
"width": 1080,
"height": 1079
}
],
"colors": [
{
"hex": "#101111",
"css": "black",
"html": "black",
"percent": 0.24719358
},
{
"hex": "#2f2d33",
"css": "darkslategray",
"html": "black",
"percent": 0.22062495
}
],
"keywords": [
{ "word": "man", "weight": 1 },
{ "word": "halloween", "weight": 0.9583 }
]
}
]
}
The response has been truncated here, real posts come with four image sizes, five colors and more keywords.
Deploying
There are prebuilt docker images on Docker Hub:
evanofslack/analogdbfor the backend APIevanofslack/analogdb-webfor the frontendevanofslack/analogdb-dagsterfor the scraping pipelineevanofslack/analogdb-consumerfor the analytics consumer
Please see the docker-compose.yaml for each service (backend, web, scrape and consumer) for an example deployment with necessary variables and services.
