AgentWithMemory_Step10_MemoryUsingChroma

Agent with Memory Using Chroma

This sample persists chat history in Chroma and recalls relevant messages in a new agent session. It uses ChatHistoryMemoryProvider with ChromaVectorStore from CommunityToolkit.VectorData.Chroma.

Features Demonstrated

• Authenticating to Microsoft Foundry with DefaultAzureCredential
• Storing chat messages in a Chroma vector store
• Creating the chat-history collection when it does not exist
• Recalling relevant chat history across agent sessions

Prerequisites

1. .NET 10 SDK

2. A Microsoft Foundry project with:

   • A chat model deployment (the default is gpt-5.4-mini)
   • A text-embedding-3-large deployment with 3,072 dimensions

3. A running Chroma server, version 1.5.0 or later. To start the instance defined in the compose file, run docker compose up -d chroma at the root of the repository. Or you can run one with Docker:

   docker run -d --name chroma -p 8000:8000 chromadb/chroma:1.5.9

4. Azure CLI authentication (az login) with an identity that has the Foundry User role on the Foundry resource

The Foundry project endpoint does not serve embeddings. The sample generates them through the endpoint of the Foundry resource that hosts the project, using the same credential.

Configuration

Set the following environment variables:

Variable	Description	Default
FOUNDRY_PROJECT_ENDPOINT	Microsoft Foundry project endpoint	(required)
FOUNDRY_MODEL	Chat model deployment name	gpt-5.4-mini
FOUNDRY_EMBEDDING_MODEL	Embedding model deployment name	text-embedding-3-large
FOUNDRY_EMBEDDING_DIMENSIONS	Number of dimensions produced by the embedding deployment	3072
CHROMA_ENDPOINT	Chroma server endpoint	http://localhost:8000

Run the Sample

dotnet run

The first session stores the user's preference for pirate jokes. The second session uses a different AgentSession but the same per-run user search scope. This lets the agent retrieve that preference from Chroma without recalling data from earlier sample runs.