orama

Implements full-text search using a radix tree data structure combined with BM25 ranking algorithm, with built-in support for typo tolerance via Levenshtein distance matching and linguistic normalization through stemming and stop-word removal. The engine tokenizes input text, applies language-specific stemmers (English, Italian, French, Spanish, German, Portuguese, Dutch, Swedish, Norwegian, Danish, Russian, Arabic, Chinese, Japanese), and matches against indexed terms with configurable edit-distance thresholds to handle misspellings without requiring external spell-check services.

Implements approximate nearest neighbor (ANN) search using a flat vector index with cosine similarity scoring, supporting integration with external embedding providers (OpenAI, Hugging Face, Ollama) via a pluggable embeddings system. The engine stores dense vectors alongside documents, performs similarity calculations in-memory, and allows custom embedding models through the plugin architecture without requiring changes to core search logic.

Provides a pluggable embeddings abstraction that integrates with external embedding providers (OpenAI, Hugging Face, Ollama, custom endpoints) to automatically generate vector embeddings for documents and queries. The plugin handles API communication, caching of embeddings, batch processing for efficiency, and fallback strategies if embedding generation fails, allowing seamless integration of vector search without vendor lock-in.

Provides a plugin that automatically tracks search metrics including query frequency, result click-through rates, query latency, and zero-result queries. Collects metrics in-memory or forwards them to external analytics services, enabling monitoring of search quality and user behavior without modifying application code. Metrics can be queried programmatically or exported for analysis.

Provides a plugin that identifies and highlights matched terms in search results by analyzing which terms matched in full-text search and wrapping them with configurable HTML tags (default: `<mark>` elements). The plugin tracks match positions during search, reconstructs the original text with highlights, and supports custom highlight templates for styling matched terms differently based on match type (exact, fuzzy, stemmed).

Provides a plugin that proxies search requests to Orama Cloud infrastructure, allowing applications to use cloud-hosted search indexes while maintaining the same local API. The plugin handles authentication, request forwarding, response transformation, and fallback to local search if cloud is unavailable, enabling hybrid deployments where some searches use cloud infrastructure and others use local indexes.

Provides a plugin that automatically extracts searchable content from various document formats (Markdown, HTML, PDF, JSON) during indexing, handling format-specific parsing, metadata extraction, and content normalization. The plugin supports custom parsers for domain-specific formats and integrates with framework plugins to extract content from documentation source files.

Provides language-specific tokenization for full-text indexing, with specialized support for Chinese, Japanese, and Korean (CJK) languages that don't use whitespace-based word boundaries. Implements dictionary-based and statistical tokenization algorithms for CJK, falls back to whitespace tokenization for other languages, and allows custom tokenizers per language for domain-specific needs.

Provides language-specific stemming algorithms that reduce words to their root forms (e.g., 'running', 'runs', 'ran' → 'run') during indexing and search, improving recall by matching morphological variants. Includes pre-built stemmers for English, Italian, French, Spanish, German, Portuguese, Dutch, Swedish, Norwegian, Danish, Russian, and Arabic, with support for custom stemmers for unsupported languages.

Provides language-specific stop word lists (common words like 'the', 'a', 'and') that are excluded from full-text indexing to reduce index size and improve relevance. Includes pre-built stop word lists for 20+ languages, allowing selective filtering per language without modifying search logic. Stop words are removed during tokenization, reducing index size by 20-30% for typical documents.

Merges full-text and vector search results using a configurable scoring algorithm that normalizes and weights both ranking signals. The engine executes both search paths in parallel, applies separate relevance scoring (BM25 for full-text, cosine similarity for vectors), normalizes scores to a common scale, and combines them using a weighted formula (configurable via `hybrid_weight` parameter) to produce a single ranked result set.

Provides a TypeScript-first schema system that defines document structure, field types (string, number, boolean, enum), and indexing behavior (searchable, sortable, filterable). The schema is validated at insert time, enforces type safety across the codebase, and enables the engine to optimize indexing strategies per field type — full-text indexing for strings, numeric range indexing for numbers, and facet indexing for categorical fields.

Implements faceted navigation by building inverted indexes for categorical fields, allowing queries to return aggregated counts of values per field. The engine tracks document membership in facet categories during indexing, executes facet aggregation queries in parallel with search, and returns both search results and facet metadata (value counts, available options) in a single response for building filter UIs.

Allows explicit pinning of specific documents to top positions in search results, overriding algorithmic ranking. The engine maintains a pin list per search query, executes the search normally, then reorders results to place pinned documents first while preserving relative ranking of unpinned results. Useful for promoting important documents (e.g., official documentation, sponsored content) regardless of relevance score.

Enables exporting the entire in-memory search index (including full-text index, vector embeddings, facet data, and metadata) to a binary format that can be persisted to disk or transmitted over the network, and importing it back to reconstruct the index without reindexing. Uses a custom binary serialization format optimized for size and deserialization speed, supporting both Node.js Buffer and browser Blob formats.

Provides a hook-based plugin system allowing third-party code to extend Orama functionality at multiple integration points: before/after indexing, before/after search, custom tokenization, embedding generation, and result post-processing. Plugins are registered at database creation time, receive context objects with access to internal state, and can modify documents, search parameters, or results without forking the core library.

Provides pre-built plugins for popular static site generators (Docusaurus, VitePress, Nextra, Astro) that automatically extract content from documentation sites, build search indexes during build time, and inject search UI components into the generated site. Each plugin handles framework-specific content parsing, metadata extraction, and asset bundling to produce a ready-to-use search interface.

Provides a plugin that automatically persists search indexes to disk at configurable intervals or on-demand, enabling recovery of indexes across application restarts without reindexing. Supports multiple storage backends (file system, IndexedDB in browsers) and handles serialization/deserialization transparently, allowing developers to treat persistence as a transparent layer.