// the find
milvus-io/bootcamp
Dealing with all unstructured data, such as reverse image search, audio search, molecular search, video analysis, question and answer systems, NLP, etc.
A collection of Jupyter notebooks and demo applications showing how to use Milvus (the vector database) for tasks like image search, RAG pipelines, question answering, and semantic search. It's aimed at developers who want working examples of Milvus integration patterns rather than production-ready code.
The breadth of covered use cases is genuinely useful — hybrid search combining dense and sparse embeddings, Graph RAG, RAGAS evaluation, and Spark integration are all present and not just trivial toy examples. The notebooks are organized by use case rather than by API surface, which makes it easier to find something relevant to your actual problem. CI workflows actually run the Jupyter notebooks on PRs, so they're less likely to be silently broken than typical example repos. The devcontainer configuration means you can get a working environment without fighting Python dependency hell.
The repo is heavily oriented toward Zilliz Cloud (the managed Milvus offering) — several notebooks push you toward their hosted service rather than self-hosted Milvus, which is a conflict of interest worth knowing about. The applications directory (reverse image search, QA system) is noticeably older than the bootcamp notebooks and uses patterns that are now considered outdated (separate MySQL for metadata, no pymilvus 2.x client patterns). There's no coherent versioning story — notebooks written for different Milvus versions sit side by side without clear markers, so you can hit SDK breakage without warning. The 'tutorials' are mostly linear walkthroughs with no discussion of failure modes, production considerations, or when you'd choose Milvus over a simpler alternative.