Azure OpenAI Atlas Vector Search
Parent: MongoDB Atlas on Azure · Published reference · snapshot 2026-10-23 · skill mongodb-cohort-2618e44302ce
↓ Facts as markdownall context files
Source-anchored research for Azure OpenAI Atlas Vector Search. Run status: COMPLETED.
These notes link each claim to its source. Some citations identify archived research reports that are no longer hosted on this site. A published reference means the content is available; it does not certify independent review or accuracy.Read the editorial policy and follow the sources before relying on a claim.
Comparisons and alternatives
- A June 19, 2025 n8n user reported "Server selection timed out after 30000 ms" in an Atlas Vector Store node paired with Azure OpenAI embeddings, while chat-memory access in the same cluster worked. A July 3 reply requested credential, permission and read-versus-insert checks; the thread provides no confirmed diagnosis or fix. (low confidence) [source]
Facts and statements
- Azure OpenAI can generate document and query embeddings while Atlas stores document vectors and retrieves similar content; the retrieved content supplies context for model generation. (high confidence) [source]
- Microsoft states that Azure OpenAI On Your Data retired on October 14, 2026. Its connector instructions are historical and unsupported for new deployments; ordinary Azure OpenAI chat completions without On Your Data configuration are outside that retirement. (low confidence) (verified as of 2026-10-14) [source]
- Microsoft’s migration sample does not migrate MongoDB sources automatically. It proposes application-side source retrieval followed by generation, or separately loading an Azure AI Search index; embeddings, access controls, and connector behavior need explicit migration work. (low confidence) (verified as of 2026-10-14) [source]
- Historically, On Your Data connected an Atlas vector index through the Foundry portal, API or SDK. Setup identified the database, collection and vector index, mapped content and vector fields, and used Azure OpenAI ada002 embeddings. (medium confidence) [source]
- The historical Microsoft-managed Atlas connector supported public network access only. This connector restriction does not establish the networking capabilities of application-managed Atlas retrieval. (low confidence) [source]
- Confluent documents a Flink SQL implementation that creates an Azure OpenAI embedding connection and a MongoDB Atlas external table, generates query embeddings with AI_EMBEDDING, and queries Atlas with VECTOR_SEARCH_AGG. Its MongoDB connector documentation restricts retrieval to Atlas approximate nearest-neighbor search and says MongoDB filtering is unsupported. These are connector-specific limits. (low confidence) (verified as of 2026-10-23) [source]
- A published Python example explicitly composes AzureOpenAIEmbeddings with MongoDBAtlasVectorSearch.from_connection_string and separately invokes AzureChatOpenAI on retrieved documents. It demonstrates application-managed composition, but its 2024 imports and preview API version are historical example code, not evidence of current SDK compatibility. (low confidence) [source]
- MongoDB announced its managed Azure OpenAI integration at Ignite in November 2024. Its tutorial describes selecting an Atlas vector index in the Foundry chat playground or supplying collection and index information through an API, with citations identifying retrieved chunks. (medium confidence) [source]
- The historical Atlas connector required public database access, Azure OpenAI ada002 embeddings, database/collection/index identifiers, and content/vector field mappings. These connector restrictions do not establish equivalent restrictions for application-managed Atlas retrieval. (low confidence) (verified as of 2026-10-14) [source]
- A framework integration can compose Spring AI’s MongoDBAtlasVectorStore, which accepts an EmbeddingModel, with its AzureOpenAiEmbeddingModel implementation. This is an application-managed alternative to the legacy managed connector, inferred from the two documented interfaces. (medium confidence) [source]
- Spring AI’s Atlas integration requires an embedding model and configured collection/index. Schema initialization is opt-in through spring.ai.vectorstore.mongodb.initialize-schema=true; otherwise an application must arrange the index separately. (low confidence) [source]
- MongoDB’s Leafy Hospital example uses Azure OpenAI text-embedding-ada-002 on clinical notes, stores embeddings in Atlas, and matches query vectors through Atlas Vector Search. This is an implementation example, not independent evidence of clinical effectiveness. (low confidence) [source]
Children
- No children recorded.