ChromaDB C++

Easily interact with ChromaDB Vector Database in C++

Description

ChromaDB is an open-source vector database designed for managing and querying high-dimensional vector data efficiently.

ChromaDB C++ lets you easily interact with the ChromaDB Vector Database:

• Collection Management: Create, retrieve, update, and delete collections
• Embedding Management: Add, get, update, upsert, and delete embeddings
• Querying: Perform advanced searches on collections using documents or embeddings
• Authorization Handling: Connect securely with authorization tokens

Example

#include "ChromaDB/ChromaDB.h"

int main()
{
	chromadb::Client client("http", "localhost", "8080");

	std::cout << client.GetVersion() << std::endl;
	std::cout << client.GetHeartbeat() << std::endl;

	chromadb::Collection collection = client.CreateCollection("test_collection");

	std::cout << collection.GetId() << std::endl;

	std::vector<std::string> ids = { "ID1", "ID2", "ID3" };
	std::vector<std::vector<double>> embeddings = { { 1.0, 2.0, 3.0 }, { 4.0, 5.0, 6.0 }, { 7.0, 8.0, 9.0 } };
	std::vector<std::unordered_map<std::string, std::string>> metadatas = { { {"key1", "value1"} }, { {"key1", "value2"} }, { {"key1", "value3"} } };

	client.AddEmbeddings(collection, ids, embeddings, metadatas);

	std::vector<chromadb::QueryResponseResource> queryResponse = client.Query(collection, {}, { { 1.0, 2.0, 3.0 } }, 2);

	std::cout << queryResponse[0].ids[0] << std::endl; // ID1
	std::cout << queryResponse[0].metadatas->at(0).at("key1") << std::endl; // value1

	client.DeleteCollection(collection);
}

Running ChromaDB

To use this library, you need to have ChromaDB running either locally or in the cloud.

To run ChromaDB in a docker container, you can use the following command:

docker container run -p 8080:8000 chromadb/chroma

Alternatively, you can use a docker-compose file:

services:
    chromadb:
        image: 'chromadb/chroma'
        ports:
            - '8080:8000'
        volumes:
            - chroma-data:/chroma/chroma

volumes:
    chroma-data:
        driver: local

Run it with:

docker-compose up

Once running, you can access ChromaDB at: http://localhost:8080

Installation

This guide will help you build and install the ChromaDB library and run an example project on both Linux and Windows systems.

Prerequisites

• CMake (version 3.10 or higher)
• A C++ compiler (supporting C++20)
• OpenSSL development libraries

Building ChromaDB Library

Make sure to run the following command to initialize and update the submodules on both Linux and Windows if not done already:

git submodule update --init --recursive

Linux

1. Create a build directory

mkdir build
cd build

2. Run CMake

cmake ..

3. Build the project

make

4. Install the library

sudo make install

Windows

1. Create a build directory

mkdir build
cd build

2. Run CMake

cmake ..

3. Build the project

cmake --build . --config Release

4. Install the library (you may need administrator privileges)

cmake --install .

Running the example project

Linux

1. Navigate to the examples directory

cd examples

2. Create a build directory

mkdir build
cd build

3. Run CMake

cmake ..

4. Build the example project

make

5. Run the example

./ExampleProject

Windows

1. Navigate to the examples directory

cd examples

2. Create a build directory

mkdir build
cd build

3. Run CMake

cmake ..

4. Open the generated Visual Studio solution and build it
   • Open ExampleProject.sln in Visual Studio
   • Build the project (ensure you build in Release mode)

Notes

• On Windows, ensure that the chromadb.dll is copied to the output directory where the ExampleProject executable resides. This is handled by the CMake script with a post-build command.
• You may need to adjust the CMAKE_PREFIX_PATH in the examples CMakeLists.txt if the library and include paths for ChromaDB are different on your system.
• If you encounter any issues with finding the OpenSSL libraries, ensure the OPENSSL_ROOT_DIR environment variable is set to the correct path of your OpenSSL installation. You can download OpenSSL pre-built binaries here.

Testing

To run tests included in the ChromaDB project:

cd build
ctest -C Release

This will execute the tests defined in the tests directory, using Google Test.

Usage

To use ChromaDB in your project, simply include the header file "ChromaDB/ChromaDB.h". This header provides access to the full suite of functionalities offered by the ChromaDB C++ library.

Connecting to ChromaDB

To connect to a ChromaDB server, you need to create an instance of the Client class. You can specify the connection details such as the scheme, host, port, database, tenant, and an authorization token.

#include "ChromaDB/ChromaDB.h"

int main()
{
    // Define connection parameters
    std::string scheme = "http";
    std::string host = "localhost";
    std::string port = "8080";
    std::string database = "my_database";
    std::string tenant = "my_tenant";
    std::string authToken = "my_auth_token";

    // Create a ChromaDB client
    chromadb::Client client(scheme, host, port, database, tenant, authToken);
}

Parameters

• scheme: The protocol to use (http or https).
• host: The hostname or IP address of the ChromaDB server.
• port: The port number on which the ChromaDB server is running.
• database: (Optional) The database to use (defaults to default_database). If it does not exist, it will be created.
• tenant: (Optional) The tenant to use (defaults to default_tenant). If it does not exist, it will be created.
• authToken: (Optional) The authorization token to use.

Authentication

ChromaDB supports static token-based authentication. To use it, start the Chroma server with the necessary environment variables as specified in the documentation. For Docker Compose, you can use the env_file option or pass the environment variables directly as shown below:

services:
    chroma-with-auth:
        image: 'chromadb/chroma'
        ports:
            - '8081:8000'
        volumes:
            - chroma-data:/chroma/chroma
        env_file: 
            - .env
        # Alternatively, you can specify the environment variables directly
        environment:
            CHROMA_SERVER_AUTHN_CREDENTIALS: 'authToken'
            CHROMA_SERVER_AUTHN_PROVIDER: 'chromadb.auth.token_authn.TokenAuthenticationServerProvider'

volumes:
    chroma-data:
        driver: local

Create a Collection

To create a new collection in ChromaDB, use the CreateCollection method. This method allows you to specify the name of the collection, optional metadata, and an optional embedding function.

#include "ChromaDB/ChromaDB.h"

int main()
{
    std::unordered_map<std::string, std::string> metadata = { {"key1", "value1"}, {"key2", "value2"} };
    std::shared_ptr<chromadb::EmbeddingFunction> embeddingFunction = std::make_shared<chromadb::OpenAIEmbeddingFunction>("openai-api-key");

    chromadb::Collection collection = client.CreateCollection("test_collection", metadata, embeddingFunction);
}

Parameters

• name: The name of the collection to create.
• metadata: (Optional) A map of metadata key-value pairs for the collection.
• embeddingFunction: (Optional) A shared pointer to an embedding function for the collection.

Get a Collection

To retrieve an existing collection in ChromaDB, use the GetCollection method. This method allows you to specify the name of the collection and an optional embedding function.

#include "ChromaDB/ChromaDB.h"

int main()
{
    chromadb::Collection collection = client.GetCollection("test_collection");
    std::cout << "Collection name: " << collection.GetName() << std::endl;
    std::cout << "Collection id: " << collection.GetId() << std::endl;

    std::cout << "Collection metadata: " << std::endl;
    for (const auto& [key, value] : collection.GetMetadata())
	std::cout << key << ": " << value << std::endl;
}

Parameters

• name: The name of the collection to retrieve.
• embeddingFunction: (Optional) A shared pointer to an embedding function for the collection.

Get or Create a Collection

To retrieve an existing collection or create a new one in ChromaDB, use the GetOrCreateCollection method. This method allows you to specify the name of the collection, optional metadata, and an optional embedding function.

#include "ChromaDB/ChromaDB.h"

int main()
{
    chromadb::Collection collection = client.GetOrCreateCollection("test_collection");
}

Parameters

• name: The name of the collection to retrieve.
• metadata: (Optional) A map of metadata key-value pairs to associate with the collection.
• embeddingFunction: (Optional) A shared pointer to an embedding function for the collection.

Get all Collections

To retrieve all existing collections in ChromaDB, use the GetCollections method. This method allows you to specify an optional embedding function that applies to all collections.

#include "ChromaDB/ChromaDB.h"

int main()
{
    std::vector<chromadb::Collection> collections = client.GetCollections();

    for (chromadb::Collection& collection : collections)
        std::cout << "Collection name: " << collection.GetName() << std::endl;
}

Parameters

• embeddingFunction: (Optional) A shared pointer to an embedding function for the collections.

Get Collection Count

To get the total number of collections in ChromaDB, use the GetCollectionCount method.

#include "ChromaDB/ChromaDB.h"

int main()
{
    size_t collectionCount = client.GetCollectionCount();
    std::cout << "Collection count: " << collectionCount << std::endl;
}

Update a Collection

To update an existing collection in ChromaDB, use the UpdateCollection method. This method allows you to change the name and metadata of a collection.

#include "ChromaDB/ChromaDB.h"

int main()
{
    std::string newName = "test_collection_updated";
    std::unordered_map<std::string, std::string> newMetadata = { {"key3", "value3"}, {"key4", "value4"} };
    chromadb::Collection updatedCollection = client.UpdateCollection("test_collection", newName, newMetadata);

    std::cout << updatedCollection.GetName() << std::endl; // "test_collection_updated"
}

Parameters

• oldName: The old name of the collection to update.
• newName: The new name for the collection.
• newMetadata: (Optional) A map of new metadata key-value pairs for the collection.

Delete a Collection

To delete an existing collection in ChromaDB, use the DeleteCollection method. This method allows you to specify the collection you want to remove.

#include "ChromaDB/ChromaDB.h"

int main()
{
    chromadb::Collection collection = client.CreateCollection("test_collection");
    client.DeleteCollection(collection);

    collection.GetName(); // Throws exception
}

Parameters

• collection: The collection to delete.

Add Embeddings to a Collection

To add embeddings to an existing collection in ChromaDB, use the AddEmbeddings method. This method allows you to specify the collection, the IDs of the embeddings, and optionally, the embeddings themselves, metadata, and documents associated with the embeddings.

#include "ChromaDB/ChromaDB.h"

int main()
{
    std::vector<std::string> ids = { "ID1", "ID2", "ID3" };
    std::vector<std::vector<double>> embeddings = { { 1.0, 2.0, 3.0 }, { 4.0, 5.0, 6.0 }, { 7.0, 8.0, 9.0 } };
    std::vector<std::unordered_map<std::string, std::string>> metadatas = { { {"key1", "value1"} }, { {"key2", "value2"} }, { {"key3", "value3"} } };

    client.AddEmbeddings(collection, ids, embeddings, metadatas, {});
}

Parameters

• collection: The collection to which embeddings will be added.
• ids: The unique IDs of the embeddings.
• embeddings: (Optional) A vector of embeddings.
• metadata: (Optional) A vector of metadata key-value pairs associated with the embeddings.
• documents: (Optional) A vector of documents associated with the embeddings

If you do not provide embeddings, you must provide documents along with an embedding function.

Add Embeddings with Embedding Function

When adding embeddings to a collection in ChromaDB, you can utilize an embedding function to generate embeddings from documents. You need to either call embeddingFunction->Generate() to generate embeddings before manually passing them to AddEmbeddings, or simply pass an embedding function directly to a collection, and the embeddings will be generated automatically.

Manual Generation of Embeddings

#include "ChromaDB/ChromaDB.h"

int main()
{
    std::shared_ptr<chromadb::OpenAIEmbeddingFunction> embeddingFunction = std::make_shared<chromadb::OpenAIEmbeddingFunction>("openai-api-key");

    chromadb::Collection collection = client.GetCollection("test_collection");

    std::vector<std::string> ids = { "ID1", "ID2", "ID3" };
    std::vector<std::string> documents = { "document1", "document2", "document3" };
    auto embeddings = embeddingFunction->Generate(documents);

    client.AddEmbeddings(collection, ids, embeddings, {}, documents);
}

Automatic Generation of Embeddings

#include "ChromaDB/ChromaDB.h"

int main()
{
    std::shared_ptr<chromadb::OpenAIEmbeddingFunction> embeddingFunction = std::make_shared<chromadb::OpenAIEmbeddingFunction>("openai-api-key");

    chromadb::Collection collection = client.GetCollection("test_collection", embeddingFunction);

    std::vector<std::string> ids = { "ID1", "ID2", "ID3" };
    std::vector<std::string> documents = { "document1", "document2", "document3" };

    client.AddEmbeddings(collection, ids, {}, {}, documents);
}

We currently supports JinaEmbeddingFunction, OpenAIEmbeddingFunction, CohereEmbeddingFunction, VoyageAIEmbeddingFunction and TogetherAIEmbeddingFunction for this purpose.

JinaEmbeddingFunction

#include "ChromaDB/ChromaDB.h"

int main()
{
    std::shared_ptr<chromadb::JinaEmbeddingFunction> jinaEmbeddingFunction = std::make_shared<chromadb::JinaEmbeddingFunction>("jina-api-key");
}

Parameters

• apiKey: The API key to access the API.
• model: (Optional) The model to use for generating embeddings. Defaults to jina-embeddings-v2-base-en.
• baseUrl: (Optional) The base URL of the API server. Defaults to api.jina.ai.
• path: (Optional) The path of the endpoint for generating embeddings. Defaults to /v1/embeddings.

Note: You can get started immediately by obtaining a free Jina API Key with 1M Tokens here

OpenAIEmbeddingFunction

#include "ChromaDB/ChromaDB.h"

int main()
{
    std::shared_ptr<chromadb::OpenAIEmbeddingFunction> openAIEmbeddingFunction = std::make_shared<chromadb::OpenAIEmbeddingFunction>("openai-api-key");
}

Parameters

• apiKey: The API key to access the API.
• model: (Optional) The model to use for generating embeddings. Defaults to text-embedding-3-small.
• dimensions: (Optional) The number of dimensions of the embeddings. Defaults to 1536.
• baseUrl: (Optional) The base URL of the API server. Defaults to api.openai.com.
• path: (Optional) The path of the endpoint for generating embeddings. Defaults to /v1/embeddings.

CohereEmbeddingFunction

#include "ChromaDB/ChromaDB.h"

int main()
{
    std::shared_ptr<chromadb::CohereEmbeddingFunction> cohereEmbeddingFunction = std::make_shared<chromadb::CohereEmbeddingFunction>("cohere-api-key");
}

Parameters

• apiKey: The API key to access the API.
• model: (Optional) The model to use for generating embeddings. Defaults to embed-english-v3.0.
• inputType: (Optional) The input type passed to the model. Defaults to classification.
• baseUrl: (Optional) The base URL of the API server. Defaults to api.cohere.com.
• path: (Optional) The path of the endpoint for generating embeddings. Defaults to /v1/embed.

VoyageAIEmbeddingFunction

#include "ChromaDB/ChromaDB.h"

int main()
{
    std::shared_ptr<chromadb::VoyageAIEmbeddingFunction> voyageAIEmbeddingFunction = std::make_shared<chromadb::VoyageAIEmbeddingFunction>("voyageai-api-key");
}

Parameters

• apiKey: The API key to access the API.
• model: (Optional) The model to use for generating embeddings. Defaults to voyage-2.
• inputType: (Optional) The input type passed to the model. Defaults to document.
• baseUrl: (Optional) The base URL of the API server. Defaults to api.voyageai.com.
• path: (Optional) The path of the endpoint for generating embeddings. Defaults to /v1/embeddings.

TogetherAIEmbeddingFunction

#include "ChromaDB/ChromaDB.h"

int main()
{
    std::shared_ptr<chromadb::TogetherAIEmbeddingFunction> togetherAIEmbeddingFunction = std::make_shared<chromadb::TogetherAIEmbeddingFunction>("togetherai-api-key");
}

Parameters

• apiKey: The API key to access the API.
• model: (Optional) The model to use for generating embeddings. Defaults to togethercomputer/m2-bert-80M-8k-retrieval.
• baseUrl: (Optional) The base URL of the API server. Defaults to api.together.xyz.
• path: (Optional) The path of the endpoint for generating embeddings. Defaults to /v1/embeddings.

Getting Additional Metadata from Embedding Requests

If you generated embeddings using an embedding function that calls a provider (e.g., OpenAI or Jina), you can retrieve additional information about the most recent request using the embeddingFunction->GetRequestMetadata() method. This function returns specific metadata for a provider.

#include "ChromaDB/ChromaDB.h"

int main()
{
    std::shared_ptr<chromadb::JinaEmbeddingFunction> embeddingFunction = std::make_shared<chromadb::JinaEmbeddingFunction>("jina-api-key");

    chromadb::Collection collection = client.GetOrCreateCollection("test_collection", {}, embeddingFunction);

    std::vector<std::string> ids = { "ID1", "ID2" };
    std::vector<std::string> documents = { "document1", "document2" };

    client.AddEmbeddings(collection, ids, {}, {}, documents);

    std::cout << embeddingFunction->GetRequestMetadata() << std::endl;
    // Example output (specific for the Jina API): 
    // {
    //     "model": "jina-embeddings-v2-base-en",
    //     "object": "list",
    //     "usage": {
    //         "prompt_tokens": 8,
    //         "total_tokens": 8
    //     }
    // }

    // Access specific fields in the additional metadata
    std::cout << embeddingFunction->GetRequestMetadata()["usage"]["prompt_tokens"] << std::endl; // 8
}

Get Embeddings from a Collection

To retrieve embeddings from an existing collection in ChromaDB, use the GetEmbeddings method. This method allows you to specify the collection, optional IDs of the embeddings, and optional filters and fields to include in the result.

#include "ChromaDB/ChromaDB.h"

int main()
{
    chromadb::Collection collection = client.GetCollection("test_collection");

    std::vector<std::string> ids = { "ID1", "ID2", "ID3" };
    std::vector<std::vector<double>> embeddings = { { 1.0, 2.0, 3.0 }, { 4.0, 5.0, 6.0 }, { 7.0, 8.0, 9.0 } };
    std::vector<std::string> documents = { { "Document1" }, { "Document2" }, { "Document3" } };
    std::vector<std::unordered_map<std::string, std::string>> metadatas = { { {"key1", "value1"} }, { {"key1", "value2"} }, { {"key1", "value3"} } };

    client.AddEmbeddings(collection, ids, embeddings, metadatas, documents);

    auto queryResponse = client.GetEmbeddings(collection, {}, { "embeddings", "documents", "metadatas" }, {}, {});

    std::cout << queryResponse.size() << std::endl; // 3
    std::cout << queryResponse[0].id << std::endl; // ID1
    std::cout << queryResponse[0].embeddings->at(0) << std::endl; // 1.0
    std::cout << queryResponse[1].metadata->at("key1") << std::endl; // value2
    std::cout << queryResponse[2].document->c_str() << std::endl; // Document3
}

Parameters

• collection: The collection from which to retrieve embeddings.
• ids: (Optional) The IDs of the embeddings to retrieve.
• include: (Optional) The fields to include in the result (e.g., "metadatas", "documents", "embeddings").
• where_document: (Optional) The where clause for filtering documents.
• where: (Optional) The where clause for filtering metadata.

You can also filter the results based on their metadata using a where clause:

#include "ChromaDB/ChromaDB.h"

int main()
{
    nlohmann::json where = { {"key1", "value2"} };
    auto queryResponse = client.GetEmbeddings(collection, {}, { "embeddings", "documents", "metadatas" }, {}, where);

    std::cout << queryResponse.size() << std::endl; // 1
    std::cout << queryResponse[0].id << std::endl; // ID2
}

The where clause must be an array of key-value pairs. The key must be a string, and the value can be a string or a nested objects of valid filter values:

• $eq: Equals
• $ne: Not equals
• $in: In
• $nin: Not In
• $and: And
• $or: Or

The $gt, $gte, $lt, and $lte filters are also officially supported in ChromaDB. However, this C++ driver only supports strings for metadata values, so these operators are not available in this implementation yet.

#include "ChromaDB/ChromaDB.h"

int main()
{
    nlohmann::json where = { {"key1", { {"$ne", "value1"} }} };
    auto queryResponse = client.GetEmbeddings(collection, {}, { "embeddings", "documents", "metadatas" }, {}, where);

    std::cout << queryResponse.size() << std::endl; // 2
    std::cout << queryResponse[0].id << std::endl; // ID2
    std::cout << queryResponse[1].id << std::endl; // ID3
}

#include "ChromaDB/ChromaDB.h"

int main()
{
    nlohmann::json where = { {"key1", { {"$nin", { "value1", "value3" }} }} };
    auto queryResponse = client.GetEmbeddings(collection, {}, { "embeddings", "documents", "metadatas" }, {}, where);

    std::cout << queryResponse.size() << std::endl; // 1
    std::cout << queryResponse[0].id << std::endl; // ID2
}

#include "ChromaDB/ChromaDB.h"

int main()
{
    nlohmann::json where = { { "$or", { { {"key1", "value1"} }, { {"key1", "value3"} } } } };
    auto queryResponse = client.GetEmbeddings(collection, {}, { "embeddings", "documents", "metadatas" }, {}, where);

    std::cout << queryResponse.size() << std::endl; // 2
    std::cout << queryResponse[0].id << std::endl; // ID1
    std::cout << queryResponse[1].id << std::endl; // ID3
}

The where_document filter works similarly to the where filter but for filtering documents. It additionally supports the $contains filter value for more advanced document filtering:

#include "ChromaDB/ChromaDB.h"

int main()
{
    chromadb::Collection collection = client.CreateCollection("test_collection");

    std::vector<std::string> ids = { "ID1", "ID2", "ID3", "ID4" };
    std::vector<std::vector<double>> embeddings = { { 1.0, 2.0, 3.0 }, { 4.0, 5.0, 6.0 }, { 7.0, 8.0, 9.0 }, { 10.0, 11.0, 12.0 } };
    std::vector<std::string> documents = { { "Document1" }, { "Document2Special" }, { "Document3" }, { "Document4" } };
    std::vector<std::unordered_map<std::string, std::string>> metadatas = { { {"key1", "value1"}, {"key2", "value2"} }, { {"key1", "value2"}, {"key2", "value3"} }, { {"key1", "value3"}, {"key2", "value4"} }, { {"key1", "value4"}, {"key2", "value5"} } };

    client.AddEmbeddings(collection, ids, embeddings, metadatas, documents);

    nlohmann::json where_document = { {"$and", { {{"$contains", "Document2"}}, {{"$contains", "Special"}} }} };
    auto queryResponse = client.GetEmbeddings(collection, {}, { "embeddings", "documents", "metadatas" }, where_document, {});
	
    std::cout << queryResponse.size() << std::endl; // 1
    std::cout << queryResponse[0].id << std::endl; // ID2
    std::cout << queryResponse[0].embeddings->at(0) << std::endl; // 4.0
    std::cout << queryResponse[0].embeddings->at(1) << std::endl; // 5.0
    std::cout << queryResponse[0].embeddings->at(2) << std::endl; // 6.0
    std::cout << *queryResponse[0].document << std::endl; // Document2Special
    std::cout << queryResponse[0].metadata->at("key1") << std::endl; // value2
    std::cout << queryResponse[0].metadata->at("key2") << std::endl; // value3
}

If you need guidance on structuring filters such as the where clause, refer to tests/test_client.cpp for numerous examples.

Get Embedding Count from a Collection

To retrieve the count of embeddings from an existing collection in ChromaDB, use the GetEmbeddingCount method. This method provides a straightforward way to determine the number of embeddings in a collection.

#include "ChromaDB/ChromaDB.h"

int main()
{
    chromadb::Collection collection = client.CreateCollection("test_collection");

    std::cout << client.GetEmbeddingCount(collection) << std::endl;
}

Parameters

• collection: The collection from which to retrieve the count of embeddings.

Update Embeddings in a collection

To update embeddings in an existing collection in ChromaDB, use the UpdateEmbeddings method. This method allows you to specify the collection, the IDs of the embeddings to update, and optionally, the new embeddings, metadata, and documents associated with the embeddings.

#include "ChromaDB/ChromaDB.h"

int main()
{
    chromadb::Collection collection = client.GetCollection("test_collection");

    std::vector<std::string> ids = { "ID1", "ID2" };
    std::vector<std::string> new_documents = { "NewDocument1", "NewDocument2" };
    std::vector<std::unordered_map<std::string, std::string>> new_metadatas = { { {"key1", "NewValue3"}, {"key2", "NewValue4"} }, { {"key1", "NewValue4"}, {"key2", "NewValue5"} } };

    client.UpdateEmbeddings(collection, ids, {}, new_metadatas, new_documents);
}

Parameters

• collection: The collection in which to update embeddings.
• ids: The IDs of the embeddings to update.
• embeddings: (Optional) The new embeddings.
• metadata: (Optional) The new metadata associated with the embeddings.
• documents: (Optional) The new documents associated with the embeddings.

Delete Embeddings from a Collection

To delete embeddings from an existing collection in ChromaDB, use the DeleteEmbeddings method. This method allows you to specify the collection, the IDs of the embeddings to delete, and optionally, the where_document and where clauses to filter which embeddings to delete based on document or metadata criteria.

#include "ChromaDB/ChromaDB.h"

int main()
{
    chromadb::Collection collection = client.GetCollection("test_collection");

    client.DeleteEmbeddings(collection, { "ID1", "ID3" });
}

Parameters

• collection: The collection from which to delete embeddings.
• ids: The IDs of the embeddings to delete.
• where_document: (Optional) The where clause for filtering which documents to delete.
• where: (Optional) The where clause for filtering which metadata to delete.

Query a collection

To query an existing collection in ChromaDB, use the Query method. This method allows you to specify the collection, optional query documents, query embeddings, number of results, fields to include in the results, and optional where_document and where clauses to filter the query based on document or metadata criteria.

You MUST either provide queryEmbeddings OR queryDocuments. If you provide queryDocuments, you also need to pass an Embedding Function to the collection.

#include "ChromaDB/ChromaDB.h"

int main()
{
    std::shared_ptr<chromadb::JinaEmbeddingFunction> embeddingFunction = std::make_shared<chromadb::JinaEmbeddingFunction>("jina-api-key");

    chromadb::Collection collection = client.GetCollection("test_collection", embeddingFunction); // or collection.SetEmbeddingFunction(embeddingFunction);

    auto queryResponse = client.Query(collection, { "This is a query document" }, {}, 3, { "metadatas", "documents", "embeddings", "distances" });

    for (auto& response : queryResponse)
    {
        for (size_t i = 0; i < response.ids.size(); i++)
        {
            std::cout << response.ids[i] << std::endl;
            std::cout << response.documents->at(i).c_str() << std::endl;

            for (const auto& [key, value] : response.metadatas->at(i))
            {
                std::cout << key << ": " << value << std::endl;
            }
        }
    }
}

Parameters

• collection: The collection to query.
• queryDocuments: (Optional) The documents to query.
• queryEmbeddings: (Optional) The embeddings to query.
• nResults: (Optional) The number of results to return. Defaults to 10.
• include: (Optional) The fields to include in the results (e.g., "metadatas", "documents", "embeddings", "distances"). Defaults to "metadatas" and "embeddings".
• where_document: (Optional) The where clause for filtering documents.
• where: (Optional) The where clause for filtering metadata.

Reset ChromaDB

To reset ChromaDB, use the Reset method. This will delete all collections and entries.

Resetting may not be allowed by your configuration. To enable it, include ALLOW_RESET=TRUE in your environment variables.

#include "ChromaDB/ChromaDB.h"

int main()
{
    client.Reset();
}

License

This project is licensed under the MIT License.