Clustering

Unsupervised Clustering & Theme Discovery

Clusters content into semantic groups using HDBSCAN, surfacing themes, variants, and outliers. Turns raw corpora into navigable structure without labels.

video

image

text

audio

Multi-Stage

54.0K runs

Deploy Recipe

Why This Matters

Clustering is operational insight infrastructure. Once computed, clusters become queryable resources for navigation, QA, and theme-based retrieval.

from mixpeek import Mixpeek

client = Mixpeek(api_key="your-api-key")

# Create collection with embeddings
collection = client.collections.create(
    collection_name="unlabeled_corpus",
    feature_extractor={
        "feature_extractor_name": "multimodal_extractor",
        "version": "v1"
    }
)

# Run clustering
clusters = client.analytics.cluster(
    collection_id=collection.id,
    algorithm="hdbscan",
    min_cluster_size=15,
    return_outliers=True
)

# Generate cluster summaries with LLM
for cluster in clusters:
    summary = client.llm.summarize(
        cluster_id=cluster.id,
        sample_size=10
    )

# Search within a cluster
results = client.retrievers.execute(
    retriever_id="cluster-search",
    inputs={
        "cluster_id": "cluster_42",
        "query_text": "specific theme"
    }
)

Feature Extractors

Image Embedding Text Embedding Video Embedding Audio Embedding

Feature Extractors

Image Embedding

Generate visual embeddings for similarity search and clustering

752K runs

Text Embedding

Extract semantic embeddings from documents, transcripts and text content

827K runs

Video Embedding

Generate vector embeddings for video content

610K runs

Audio Embedding

Extract semantic embeddings from audio content for similarity search

420K runs

Retriever Stages

Enrichment Resources

Clustering

Analytics

Documentation

Clustering