Deep Lake
AI data lake with multimodal tensor storage, vector search, and serverless SQL — stream datasets directly to PyTorch and TensorFlow.
Deep Lake
Deep Lake (by Activeloop) is an AI-native data runtime that stores multimodal datasets — images, videos, text, audio, annotations, and embeddings — as chunked tensors in cloud or local storage. It exposes a vector store API for RAG pipelines and streams data directly into PyTorch/TensorFlow DataLoaders without full dataset downloads.
Key Features
- Multimodal tensor storage — one dataset can contain images, text, embeddings, bounding boxes, and labels
- Vector search — cosine, L2, and dot-product ANN search over embedding tensors; hybrid text+vector search
- Serverless SQL — query datasets via TQL (Tensor Query Language) with no data movement
- Streaming DataLoader — pull mini-batches for training directly from S3/GCS without local copies
- Data versioning — branch, commit, checkout datasets like git; full history tracking
- Integrations — LangChain, LlamaIndex, PyTorch, TensorFlow, and HuggingFace compatible
Quick Start
pip install deeplake
import deeplake
ds = deeplake.dataset("hub://activeloop/coco-train")
for sample in ds.pytorch(batch_size=4):
images = sample["images"] # stream from cloud
Install via ai-supply
npx ai-supply add deeplake-multimodal-data-lake
Curated mirror of the open-source Deep Lake (Apache-2.0). Get it from the source.
Compromise signals — malicious or tampered code (leaked secrets, backdoors, a dropped executable) — reduce the score, and known dependency CVEs carry a bounded penalty (they warrant review but never QUARANTINE — update the dependency to clear). Other dangerous-by-capability traits are risk surface, expected for some capabilities. Every finding is mapped to its OWASP control below.
Findings mapped to the OWASP Top 10 for LLM Applications (2025) and the OWASP Machine Learning Security Top 10. Expand any flagged control for the exact findings — compromise reduces the score; expected/risk-surface do not, except a known CVE, which carries a small bounded penalty (high/critical → Review).
The same gate an agent runs before installing (POST /api/v1/trust/deeplake-multimodal-data-lake/check). Click a policy:
Consume Deep Lake programmatically. Authenticate with an API key or session — see Authorize an agent.
# Agents: CHECK BEFORE YOU INSTALL (no auth) — score, grade, level, capability manifest
curl https://ai-supply.store/api/v1/trust/deeplake-multimodal-data-lake
# Gate against your org policy (returns { pass, violations })
curl -X POST https://ai-supply.store/api/v1/trust/deeplake-multimodal-data-lake/check \
-H "Content-Type: application/json" \
-d '{"minGrade":"B","denyPermissions":["shell"],"denyUnknownEgress":true}'
# CLI
npx ai-supply add deeplake-multimodal-data-lake
# REST (install → download)
curl -X POST https://ai-supply.store/api/v1/listings/deeplake-multimodal-data-lake/install \
-H "Authorization: Bearer $AIM_KEY"
# MCP tool
install_listing({ "slug": "deeplake-multimodal-data-lake" })OpenAPI spec →Curated mirror — latest upstream source. See the repository for tagged releases.