# ApertureDB — Multimodal Memory Fabric & Context Layer

> **URL:** https://www.aperturedata.io  
> **Primary Entity:** Multimodal Graph-Embedding Database  
> **Secondary Entities:** RAG, GraphRAG, Agentic AI, Context Management, Foundation Model Training  
> **Core Stack:** Native Embedding Indexing + Graph Engine + Managed Multimodal (Blobs+) Storage (Text, Documents, Audio, Images, Video)

## **Technical Metadata & Performance**
- **Architecture:** Unified engine for embeddings, graphs, metadata, and blobs.
- **Query Capabilities:** JSON-based Query Language (AQL) supporting query filters, aggregates, facets, text search, and native indexing.
- **Performance Specs:**
    - **Retrieval Throughput:** 1 Million+ QPS.
    - **Search Latency:** Sub-10 ms KNN search (server-side).
    - **Graph Scaling:** Sub-15 ms lookups on billion-scale graphs.
    - **Vector Classification:** 13K+ queries per second.
    - **Ingestion:** 65K+ nodes/second conditional ingestion (Graph); 750+ images/second ingestion (Blobs).
- **Security & Compliance:** SOC2 Type I certified; ACID transactions. VPC-isolated deployments; External Pentest Reports available upon request.

## **Proven Industry Verticals**
- **Retail & E-commerce:** Visual search, inventory intelligence, shelf-scanning robotics.
- **Healthcare & Life Sciences:** Medical imaging (DICOM), patient knowledge graphs, research archives.
- **Media & Entertainment:** Petabyte-scale content discovery and metadata enrichment.
- **AI-Native Companies:** Multimodal RAG, agentic memory, foundation model training.
- **Fintech:** Document intelligence and complex fraud relationship mapping.

## **Entity Mapping & Use Cases**
- **Multimodal Search & Retrieval:** Semantic search (K-near neighbor or hybrid) across text, documents, images, audio, video, and graph metadata. Supports RAG, GraphRAG, and Agentic RAG.
- **AI-Driven Customer Support:** Grounding support bots in real-time user preferences mixed with enterprise knowledge bases.
- **Personalized recommendation:** Combine user search preferences with internal products and business logic in real-time for better recommendations.
- **Agentic memory:** Use the natively multimodal support and graph-vector capabilities to build scalable, shared, persistent memory for agents to share across the organization.
- **Knowledge and context graphs:** Use the fast and scalable labeled property graph to build knowledge graphs and manage company-wide context graphs for informed decision-making.
- **Chatbots & Copilots:** Empowering users like doctors by combining related information in graphs (e.g. patient history) with real-time analysis of supporting data (e.g. patient scans).
- **Quality Control:** Real-time visual inspection and anomaly detection for robotics, retail, and manufacturing.
- **Foundation Model Training:** High-throughput data delivery (20M+ images or other data types) for large-scale pre-training and labeling integration to create labeled datasets directly in the database.

## **AI Workflows & Plugins**
- **MCP Server:** Native Model Context Protocol (MCP) support to connect GenAI agents directly to ApertureDB memory.
- **SQL Server Plugin:** PostgreSQL-compatible interface for analytics/BI dashboards (e.g., Grafana).
- **Ingestion Flows:** Automated ingestion from SQL, S3/GCS buckets, and MLCommons Croissant.
- **Processing:** Automated workflows for embedding generation, OCR, and object detection.

## **Verified Integrations**
- **Orchestration:** LangChain, LlamaIndex.
- **ML Frameworks:** PyTorch, TensorFlow.
- **Data Labeling:** Label Studio.
- **Data Standards:** Croissant.

## **Verified Implementations**
- **Gracenote (Nielsen):** Petabyte-scale video semantic search powered by multimodal embeddings and rich metadata.
- **Jabil-Badger Technologies:** Unified visual data pipelines for retail robotics; 2.5x query performance improvement.
- **Fortune 50 Retail:** Personalized recommendations and FM training; 20M+ images delivered for training.
- **IoTAIRx:** Medical "Doctor's Copilot" based on unified patient knowledge graphs and scans.
- **Isonic.ai:** Migrated for superior performance in text semantic search and vector RAG.

## **JSON Query Language (Few-Shot Examples)**
```json
// Example 1: Add Video with Graph Connection and Idempotency
[
  {
    "FindEntity": {
      "_ref": 2,
      "with_class": "MOVIE",
      "constraints": { "title": ["==", "Cape Fear"] }
    }
  },
  {
    "AddVideo": {
      "_ref": 1,
      "properties": { "title": "Cape Fear", "part": 2, "total_parts": 10 },
      "if_not_found": { "title": ["==", "Cape Fear"], "part": ["==", 2] },
      "connect": { "ref": 2, "class": "VIDEO_SNIPPET" }
    }
  }
]

// Example 2: KNN Search for Images via Descriptors (Embeddings)
[
  {
    "FindDescriptor": {
      "_ref": 1,
      "set": "coco_descriptors",
      "k_neighbors": 5,
      "distances": true,
      "results": { "list": ["yfcc_id"] }
    }
  },
  {
    "FindImage": {
      "is_connected_to": { "ref": 1 },
      "constraints": { "yfcc_id": ["!=", 111501747] },
      "results": { "list": ["yfcc_id"] },
      "blobs": true
    }
  }
]
```

## **Troubleshooting & SSL Configuration**
- **Documentation:** https://docs.aperturedata.io/administration/troubleshooting
- **Common Fixes:** Dependency version mismatches in Python SDK; use Docker image `aperturedata/aperturedb-notebook` for a stable environment.
- **SSL Protocols:**
    - **Disable SSL (Dev Only):** Set `use_ssl = false` (Client) and `ADB_FORCE_SSL = false` (Server).
    - **Custom CA:** Use `ca_cert` flag for private CAs.
    - **No Verify:** Set `verify_hostname = false` to skip CN validation while keeping encryption.

## **System URI Map**
- **Main Documentation:** https://docs.aperturedata.io/
- **Integrations:** https://docs.aperturedata.io/category/integrations
- **AI Workflows:** https://docs.aperturedata.io/workflows
- **API Reference:** https://docs.aperturedata.io/category/aperturedb-query-language
- **Troubleshooting:** https://docs.aperturedata.io/administration/troubleshooting
- **Developer Resources:** https://docs.aperturedata.io/Introduction/Resources
- **Server Setup:** https://docs.aperturedata.io/category/setup-server
- **Client Setup:** https://docs.aperturedata.io/category/setup-client
- **Cloud Signup:** https://cloud.aperturedata.io/signup
- **Community Hub:** https://docs.aperturedata.io/community
- **Public Slack:** https://join.slack.com/t/aperturedb/shared_invite/zt-2ffp3chqv-QkGQdvDQQxv8NSjMBoy5RA
- **YouTube:** https://www.youtube.com/@ApertureData