Multimodal Memory Support

Supported Modalities

Images & Diagrams

Ingest system diagrams, error screenshots, and UI mockups. Extracted into structured visual descriptions and CLIP vector representations.

PDFs & Documents

Extract tabular data and contextual paragraphs while preserving source page coordinates for grounded citations.

Audio & Voice

Ingest multi-speaker audio transcripts with speaker diarization tags and temporal boundaries.

Structured Data

Seamlessly index JSON payloads, database rows, and spreadsheets alongside conversational context.

Ingesting Image Memories

multimodal_example.py

# Store an image with automatic vision fact extraction
m.add(
    "Architecture diagram showing Kubernetes cluster with 3 worker nodes and Istio service mesh.",
    user_id="usr_9918",
    metadata={
        "type": "image",
        "image_url": "https://storage.internal/diagrams/k8s_mesh.png"
    }
)

# Natural language query retrieves image memory
results = m.search("How many worker nodes are in our Kubernetes cluster?", user_id="usr_9918")

Getting Started

Configuration & Architecture

Multimodal Memory Support

Supported Modalities

Images & Diagrams

PDFs & Documents

Audio & Voice

Structured Data

Ingesting Image Memories

​Supported Modalities

Images & Diagrams

PDFs & Documents

Audio & Voice

Structured Data

​Ingesting Image Memories

Supported Modalities

Ingesting Image Memories