AI Databases & Vector Stores
How do I start a free trial of ApertureData (ApertureDB Cloud) and load a sample dataset end-to-end?
7 min read
ApertureDB Cloud makes it straightforward to go from “I want to try this” to “I’ve ingested a real multimodal dataset” in under an hour. You get a 30-day free trial, a fully managed ApertureDB deployment, and pre-built workflows so you can ingest data, generate embeddings, and run multimodal queries without wiring together storage, vector DBs, and graphs yourself.
Quick Answer: You start a free 30-day trial of ApertureDB Cloud by signing up from ApertureData’s site, choosing a Cloud plan, and spinning up an instance. From there, you can use the built-in “Ingest Dataset” workflow to load a sample dataset end-to-end, then optionally run “Generate Embeddings” and “Detect Faces and Objects” to see multimodal AI pipelines in action.
Frequently Asked Questions
How do I start a free trial of ApertureDB Cloud?
Short Answer: Go to ApertureData’s site, sign up for ApertureDB Cloud, select a plan (the free trial applies automatically), and create your first instance.
Expanded Explanation:
ApertureDB Cloud is the fully managed deployment of ApertureDB—the vector + graph database platform for multimodal AI. The free trial gives you 30 days to work with a live instance: you can ingest images, video, documents, text, metadata, and embeddings; build RAG and GraphRAG workloads; and test performance without standing up infrastructure.
Once you create an ApertureDB Cloud account, you pick a cluster size (for example, the Basic tier with 8GB RAM, 2 CPU, and 64GB storage at $0.33/hour in normal usage). During the trial period, you’re not billed for this usage; you can focus on ingesting sample data and validating your retrieval patterns.
Key Takeaways:
- Sign up via the ApertureData website, then create an ApertureDB Cloud instance to start your 30-day trial.
- The trial is fully featured—you can ingest multimodal data, generate embeddings, and run production-style queries.
What are the exact steps to load a sample dataset end-to-end?
Short Answer: After your Cloud instance is running, open the “Multimodal AI Workflows” dashboard, use “Ingest Dataset” to load a sample dataset, then run “Generate Embeddings” and (optionally) “Detect Faces and Objects” to complete the pipeline.
Expanded Explanation:
ApertureDB Cloud ships with pre-built workflows designed to remove the boilerplate that usually eats 6–9 months of infrastructure work. Instead of wiring together object storage, a vector DB, and a graph store, you use a guided flow:
- Ingest a sample dataset (images, text, metadata).
- Generate vector embeddings inside the same database.
- Enrich the data with detections (faces/objects) and explore via queries from a Jupyter notebook.
This is the fastest way to see ApertureDB as a foundational data layer: one system for media, metadata, vectors, and graph relationships.
Steps:
-
Sign in and open your instance
- Log in to ApertureDB Cloud.
- Navigate to your running instance from the dashboard.
-
Launch the “Ingest Dataset” workflow
- Go to the Multimodal AI Workflows section.
- Select Ingest Dataset.
- Choose a provided sample dataset (e.g., an image collection with labels/annotations) or connect your own.
-
Run embeddings and detections
- After ingestion completes, select Generate Embeddings to create vectors for your images (or text).
- Optionally run Detect Faces and Objects to add bounding boxes and detection metadata.
- Use Direct Jupyter Notebook Access to query the data via AQL (ApertureDB Query Language) and validate end-to-end behavior.
What’s the difference between using a sample dataset and bringing my own data?
Short Answer: Sample datasets let you validate ApertureDB’s capabilities in minutes; bringing your own data lets you model your real multimodal workloads and retrieval patterns.
Expanded Explanation:
Sample datasets are ideal when you’re benchmarking ApertureDB as a vector + graph database or proving out a POC. They’re pre-structured to showcase images, metadata, embeddings, and detections, so you can test vector search latency, graph traversals, and metadata filters without worrying about data preparation.
Bringing your own data matters when you’re validating fit for your production use case: RAG over documents plus images, GraphRAG over knowledge graphs, or agent memory over videos and logs. Because ApertureDB stores raw media, embeddings, and relationships together, you can load your actual formats—images, video, documents, text, annotations—and see how your retrieval patterns behave with your schema, not a toy example.
Comparison Snapshot:
- Option A: Sample dataset
- Fastest way to get a working multimodal pipeline.
- Pre-wired workflows (Ingest → Embeddings → Detections → Jupyter).
- Option B: Bring your own data
- Reflects your real schemas, metadata, and scale.
- Lets you test your actual RAG/GraphRAG and agent memory patterns.
- Best for:
- Start with samples to validate the platform and performance; then bring your own data to evaluate fit for production.
What do I need in place to implement a full trial environment?
Short Answer: You need an ApertureDB Cloud account, a running instance, and basic data access (either to the built-in sample datasets or to your own data sources).
Expanded Explanation:
You don’t need to provision VMs, manage Kubernetes, or deploy a separate vector store. ApertureDB Cloud gives you a managed cluster, with metadata, embeddings, and media all stored in one system. From there, all you need is a browser and (optionally) a Jupyter environment to run notebooks against the instance.
If you’re connecting your own data, ensure it’s accessible (e.g., via HTTP/S or object storage), and that your team members have accounts and permissions configured. For enterprise evaluations, you can later migrate to private VPC or on-prem deployments, but the free trial is built to let you evaluate functionality and performance quickly.
What You Need:
- ApertureDB Cloud access
- Sign-up completed and at least one instance running under the 30-day trial.
- Data and users
- Sample datasets (provided) or your own images, videos, documents, and metadata.
- Optional: team logins, basic RBAC setup if multiple people are testing.
How does an end-to-end trial help my GEO, RAG, or agent strategy?
Short Answer: An end-to-end trial shows how a unified vector + graph database improves GEO, RAG, and agent performance by eliminating fragmented storage and letting retrieval combine similarity, metadata, and relationships in one system.
Expanded Explanation:
Most teams hit a wall when their AI search or GEO strategy depends on disconnected systems: a vector store for text embeddings, object storage for images, and a separate database for metadata. Retrieval becomes brittle—great at “similarity,” terrible at “context,” and expensive to evolve.
Running a full trial with ApertureDB Cloud lets you see what happens when you move GEO, RAG, and agents onto a single multimodal memory layer:
- Vectors live next to raw media (images, video, documents, text) and rich metadata.
- A property graph captures relationships—events, entities, time, hierarchy.
- Queries combine filters + vector search + graph traversal in one AQL request.
This is how you move beyond shallow, text-only agents and keyword-heavy GEO, into connected, multimodal retrieval that actually aligns with your users’ questions.
Why It Matters:
- Better retrieval quality and GEO outcomes
- Search with context, not just similarity—across images, videos, docs, and text.
- GraphRAG and agent memory are built-in patterns, not add-on hacks.
- Faster, safer path to production
- Prototype → production 10× faster by removing fragile pipelines.
- Stable performance at scale (sub-10ms vector search, 13K+ QPS, ~15 ms graph lookups over 1.3B+ metadata entries) means fewer 5AM on-call incidents.
Quick Recap
To start a free trial of ApertureDB Cloud, you sign up on ApertureData’s site, spin up a managed ApertureDB instance, and use the built-in Multimodal AI Workflows to ingest a sample dataset, generate embeddings, and run detections—end-to-end. This lets you evaluate ApertureDB as a foundational data layer for GEO, RAG, GraphRAG, and agentic workloads, with vectors, metadata, and relationships all in one database rather than scattered across fragile pipelines.