Chroma Integration

inertialai-chroma is the official InertialAI embedding function for ChromaDB. It connects InertialAI's inertial-embed-alpha foundation model directly to any Chroma collection, enabling semantic search over text, raw time-series sensor data, or both simultaneously — all through Chroma's standard Python API.

How It Works

InertialAIEmbeddingFunction implements Chroma's EmbeddingFunction[Documents] interface. When you attach it to a collection, Chroma calls it automatically on every add() and query() operation. The package handles authentication, request formatting, and parsing the response into the float32 vectors Chroma expects — you never interact with InertialAI's HTTP API directly.

import chromadb
from inertialai_chroma import InertialAIEmbeddingFunction

client = chromadb.HttpClient(host="localhost", port=8000)
ef = InertialAIEmbeddingFunction()  # reads INERTIALAI_API_KEY from env

collection = client.create_collection("my-sensors", embedding_function=ef)

collection.add(
    documents=[
        "temperature spike at noon",
        "stable overnight readings",
        "pressure anomaly detected",
    ],
    ids=["doc-1", "doc-2", "doc-3"],
)

results = collection.query(query_texts=["unusual thermal event"], n_results=2)
print(results["documents"])

InertialAI's API is called in the background on every add() and query().

Key Features

Feature	Details
Input modes	Text-only, time-series-only, and multi-modal (text + time-series) in the same collection
Multi-modal format	Pass json.dumps({"text": ..., "time_series": ...}) as standard Chroma document strings
Distance spaces	Cosine (default), L2, and inner product
Secure persistence	Stores the env var name, never the key value — serialized collections are safe to commit
Index protection	model_name and dimensions are immutable after collection creation
Full interface	Implements all 8 methods of Chroma's EmbeddingFunction[Documents] protocol
Python support	Python 3.11, 3.12, and 3.13

Installation

pip install inertialai-chroma

or with uv:

uv add inertialai-chroma

Prerequisites

Before using this integration you need:

• Python 3.11 or later
• An InertialAI API key — sign up at app.inertialai.com and generate a key in your dashboard. See Authentication for details.
• A running Chroma instance — follow the Chroma Docker deployment guide to start one locally.

In This Section

Page	Description
Quickstart	Install, connect to Chroma, and run your first multi-modal similarity search.
Multi-Modal Embeddings	Embed time-series sensor data alongside text in the same Chroma collection.
Collection Persistence	Understand how credentials are handled when Chroma serializes collections to disk.
Configuration Reference	Constructor parameters, distance spaces, and immutability constraints.
ECG5000 Demo	Run a local end-to-end demo with real ECG data, a Chroma server, and a Streamlit search UI.

Resources

• GitHub repository — source code, issue tracker, and contribution guide
• PyPI package — release history and package metadata
• Chroma EmbeddingFunction docs — Chroma's embedding function interface reference
• Chroma Docker deployment — run a Chroma instance locally