Native offline semantic search for Unity powered by MiniLM. Understands meaning instead of keywords with zero servers, zero APIs, and blazing-fast on-device AI inference.Semantic Search is a fully native AI-powered semantic search engine built exclusively for Unity. Unlike traditional keyword search, it understands the meaning behind user queries, allowing users to find relevant content even when exact words don't match. Everything runs completely offline using a lightweight MiniLM model through Unity Sentis, eliminating the need for cloud services, API keys, or backend infrastructure.Designed for games, XR applications, enterprise software, education, and AI-powered tools, Semantic Search provides intelligent document retrieval with an incredibly simple workflow. Generate embeddings directly inside the Unity Editor, drop a prefab into your scene, and start searching using an intuitive async C# API.Whether you're building an NPC knowledge system, offline documentation, product search, or Retrieval-Augmented Generation (RAG) pipeline, Semantic Search delivers fast, accurate, and privacy-friendly AI search on every supported platform.Key FeaturesAI-powered semantic search that understands intent instead of keywordsFully offline with no internet, cloud services, or API keys requiredNative Unity implementation using Unity SentisLightweight MiniLM model optimized for on-device inferenceOne-click embedding generation directly inside the Unity EditorSupports TXT, Markdown, HTML, JSON, DOCX, and PDF documentsAutomatic document chunking for improved retrieval qualityRuntime document indexing and semantic searchFlexible storage using Resources, StreamingAssets, or PersistentDataPathFast CPU execution with optional GPU accelerationQuantized and full-precision model supportClean async C# API with simple prefab integrationBuilt-in Editor tools, index manager, and search testing windowDemo scenes included for rapid learning and integrationPerfect ForNPC dialogue and knowledge systemsOffline help centers and FAQ applicationsAI-powered RAG retrieval pipelinesIn-game inventory and item searchProduct catalogs and e-commerce searchEnterprise document searchEducational applicationsMedical and legal reference systemsChat history and conversation searchXR, VR, and AR applicationsSmart content recommendation systemsMinimum Unity Version: Unity 2022.3 LTS or newer (including Unity 6.x)Render Pipelines: Built-in, URP, HDRPSupported Platforms: Windows, macOS, Linux, Android, iOS, WebGL, VisionOSProgramming Language: C#AI Runtime: Unity Sentis (Unity Inference Engine)Machine Learning Model: MiniLM EmbeddingsModel Formats: ONNX and Sentis (.sentis)Model Size: Approximately 45 MB (Quantized) / 90 MB (Standard)Inference: Fully On-DeviceInternet Required: NoBackend Server Required: NoCloud API Required: NoOffline Support: YesGPU Required: No (CPU supported)Runtime Indexing: YesEditor Embedding Generation: YesAutomatic Text Chunking: YesCosine Similarity Ranking: YesSupported Document Formats: TXT, MD, HTML, JSON, DOCX, PDFStorage Options: Resources, StreamingAssets, PersistentDataPathPrefab Included: YesDemo Scenes Included: YesDocumentation Included: YesExternal Dependencies:Unity Sentis package (required)TextMeshPro (demo scenes only)This package includes documentation and promotional materials generated with AI assistance (ChatGPT, DALL·E). All scripts, logic, and core functionality were manually developed by the author.
