#1
An open-source AI platform for enterprise search and chat that indexes knowledge from 50+ apps and works with any LLM, deployable on your own infrastructure.
- GitHub stars
- 33k
- Last commit
- today
- Latest release
- v5.1.2
- Self-hosted
- Yes
onyx.app
Onyx is an open-source AI platform that acts as a knowledge and context layer for teams and their AI agents. It connects to more than 50 applications, pulls in documents along with their metadata and permissions, and builds an internal representation so that LLMs can answer questions grounded in what your company actually knows.
The project argues that this indexed approach gives more reliable, lower-latency and cheaper context than approaches that have an agent search many tools iteratively through MCP. Features include agentic RAG on a hybrid index, a Deep Research mode for multi-step reports, custom agents with their own knowledge, instructions and actions, and live web search through providers such as Serper, Google PSE, Brave and SearXNG. Sandboxes and skills extend what models can do, and the chat interface works with every major LLM.
Onyx emphasizes data sovereignty through flexible self-hosted deployments, with a single-command deploy option. The code is written in Python and Next.js, and the license is listed as 'Other' on GitHub, so check the terms for the enterprise features. It suits organizations that want a private, ChatGPT-style assistant connected to internal documents.
Key features
- Connectors to 50+ workplace applications
- Hybrid index with agentic RAG
- Deep Research multi-step reports
- Custom agents with their own knowledge
- Web search via Serper, Brave and others
- Works with any LLM
- Indexing that respects source permissions
Pricing: Business costs $25 per user per month on monthly billing, or $20 per user per month billed annually, with a free trial. Enterprise pricing and deployment options are by contact.
#2
A distributed, RESTful search and analytics engine written in Java, used for full-text search, log analysis and vector search across large datasets.
- GitHub stars
- 79k
- Last commit
- today
- Latest release
- v9.5.6
- Self-hosted
- Yes
- Hosted version
- Available
elastic.co
Elasticsearch is a distributed, RESTful search engine written in Java. It stores documents as JSON, indexes them so they can be searched quickly, and exposes everything through an HTTP API, which makes it a common backbone for site search, application search and log analytics. It is built on the Apache Lucene library.
Beyond full-text queries it supports filtering, aggregations for analytics and, in recent versions, vector search for semantic and AI-assisted retrieval. Data is spread across nodes in a cluster using shards and replicas, so capacity and resilience grow by adding machines. It is the core of the Elastic Stack, usually paired with Kibana for exploration and with ingestion tools for logs and metrics.
The repository describes the project as free and open source, while its license metadata is listed as 'Other', so review the current license terms before relying on it commercially. It can be self-managed on your own servers or used as a managed service through Elastic Cloud.
Key features
- Distributed full-text search over JSON documents
- RESTful HTTP API
- Aggregations for analytics
- Vector search for semantic retrieval
- Clustering with shards and replicas
#3
A fast search engine API written in Rust that adds typo-tolerant, filterable and hybrid semantic search to websites and apps, available self-hosted or as a cloud service.
- GitHub stars
- 60k
- Last commit
- yesterday
- Latest release
- v1.56.2
- Self-hosted
- Yes
- Hosted version
- Available
meilisearch.com
Meilisearch is a search engine exposed through an API, written in Rust and designed to be dropped into applications and websites. It aims to give developers a good search experience quickly, with features that work out of the box rather than needing extensive tuning.
Its capabilities include hybrid search that combines semantic and full-text results, search-as-you-type with a target response time under 50 milliseconds, typo tolerance, filtering and faceted search, sorting and geosearch. The topics also describe it as a vector database for AI use cases. The project maintains example applications for movies, ecommerce with facets and filters, conversational home booking, personalization and a multi-tenant CRM, which show how the features fit together.
The license is listed as 'Other', so confirm the terms for your use case. Meilisearch can be run on your own infrastructure, and Meilisearch Cloud offers it as a managed service. It is a common pick for site search, product catalogs and in-app search where a lighter alternative to a large cluster is attractive.
Key features
- Hybrid semantic and full-text search
- Instant search-as-you-type results
- Typo tolerance and fuzzy matching
- Filtering and faceted search
- Sorting and geosearch
- Vector storage for AI search
- REST API with example demo apps
Pricing: Self-hosting is free under the MIT license. Meilisearch Cloud starts at $20 per month with a 14-day free trial, billed by usage or instance resources; Enterprise is quoted on request.
#4
A vector database and similarity search engine written in Rust, offering filtering over stored vectors and payloads, available self-hosted or as Qdrant Cloud.
- GitHub stars
- 36k
- Last commit
- today
- Latest release
- v1.20.0
- Licence
- Apache-2.0
- Self-hosted
- Yes
- Hosted version
- Available
qdrant.tech
Qdrant is a vector database and similarity search engine. It stores points, which are vectors plus an attached payload, and exposes an API to store, search and manage them. That makes it a building block for semantic search, recommendations, image search and other applications that rely on embeddings from neural networks.
A key selling point is extended filtering support, so a vector query can be combined with conditions on the payload, which is useful for faceted search and matching. The engine is written in Rust and the maintainers point to benchmarks for its speed and reliability under load. Topics mention HNSW indexing and nearest-neighbor search, and there are client libraries, demo projects, integrations and a collection of agent skills that bring vector search into AI coding assistants.
Qdrant is Apache-2.0 licensed and can run on your own infrastructure. It is also available as the fully managed Qdrant Cloud, which includes a free tier. Teams building retrieval-augmented generation, recommendation and similarity features use it as the storage and retrieval layer for their embeddings.
Key features
- Vector storage with attached payloads
- Filtering combined with vector search
- HNSW-based nearest-neighbor search
- REST-style API with client libraries
- Agent skills for AI coding assistants
- Self-hosted or managed Qdrant Cloud
Pricing: A free-forever cloud tier is available. The Standard tier is usage-based, from $25; Premium requires a minimum spend and Hybrid and Private Cloud are quoted by the team.
#5
A fast, typo-tolerant, in-memory search engine for site and app search, pitched as an open-source Algolia alternative and an easier-to-run Elasticsearch alternative.
- GitHub stars
- 27k
- Last commit
- 2 days ago
- Latest release
- v30.2
- Licence
- GPL-3.0
- Self-hosted
- Yes
- Hosted version
- Available
typesense.org
Typesense is an open-source search engine for building fast, typo-tolerant search experiences in websites and applications. It is pitched as an open-source substitute for Algolia and Pinecone, and as a simpler option than Elasticsearch. It keeps its index in memory and is written in C++.
The project's feature topics include typo tolerance and fuzzy matching, full-text search, faceting, geosearch, synonyms, merchandising rules and instant search, and it supports semantic and hybrid search. Public demos let you try it on large datasets, such as 32 million songs from MusicBrainz, 28 million books from OpenLibrary, a recipe set, Linux kernel commit messages, an ecommerce store experience, a geosearch listing browser and semantic search over Hacker News comments.
Typesense is licensed under GPL-3.0 and can be run on your own servers, and a managed Typesense Cloud service is also available. Documentation, a roadmap, a Slack community and video introductions are provided. It suits developers who want Algolia-style search with the option to self-host and keep costs predictable.
Key features
- Typo-tolerant, in-memory search
- Faceting, filtering and synonyms
- Geosearch for location-based results
- Semantic and hybrid search
- Merchandising controls for ranking
- Self-hosted or Typesense Cloud
#6
An Apache-licensed, distributed and RESTful search engine and observability suite for searching, analyzing and monitoring large volumes of unstructured data.
- GitHub stars
- 14k
- Last commit
- today
- Latest release
- 3.9.0
- Licence
- Apache-2.0
- Self-hosted
- Yes
- Hosted version
- Available
opensearch.org
OpenSearch is an open-source search and observability suite that makes large amounts of unstructured data searchable and analyzable. It is a distributed engine with a RESTful API, written in Java, and developed in the open by the OpenSearch community under the Apache 2.0 licence.
Teams use it for full-text search, log and event analytics, and monitoring of applications and infrastructure. The project provides downloads, documentation, forums and Slack channels, and it follows a formal release and maintainer process. Because it is distributed, clusters can be scaled across multiple nodes as data volumes grow.
OpenSearch includes certain Apache-licensed code derived from Elasticsearch, as its trademark notice explains. You can run it on your own servers, or use managed services offered by cloud providers. Security issues are reported privately by email rather than through public issues, and the project maintains a code of conduct for contributors.
Key features
- Distributed search with a RESTful API
- Full-text search over unstructured data
- Log and event analytics
- Observability and monitoring use cases
- Scales across nodes in a cluster
- Apache 2.0 licensed Java codebase
Pricing: Free and open source under Apache 2.0; managed hosting is available from cloud providers.
#7
A cloud-native vector database that stores objects and embeddings and combines semantic search with keyword filtering, RAG and reranking in one query interface.
- GitHub stars
- 17k
- Last commit
- yesterday
- Latest release
- v1.40.5
- Self-hosted
- Yes
- Hosted version
- Available
weaviate.io
Weaviate is a vector database that stores both data objects and their vector embeddings, enabling semantic search at scale. One query interface can blend vector similarity search, keyword filtering, retrieval-augmented generation and reranking, and it is written in Go as a cloud-native system.
Common applications are RAG pipelines, semantic and image search, recommenders, chatbots and content classification. Vectors can be created automatically on import by integrated models from providers such as OpenAI, Cohere and Hugging Face, or you can import pre-computed embeddings. For production, it offers built-in multi-tenancy, replication and role-based access control. Topics reference approximate nearest neighbor search and HNSW indexing.
Weaviate can be deployed with Docker, on Kubernetes, or through the managed Weaviate Cloud, with guides for AWS and GCP, and the documentation includes quickstarts for both cloud and a local Docker instance. Its license is listed as 'Other' on GitHub, so check the repository terms. It suits teams building AI applications that need a scalable store for embeddings alongside structured filters.
Key features
- Stores objects and vectors together
- Hybrid vector and keyword search
- Integrated vectorizer models or bring your own
- Multi-tenancy and replication
- Role-based access control
- Docker, Kubernetes and Weaviate Cloud options
#8
Vespa is an open-source platform for searching, inferring over and organizing vectors, tensors, text and structured data at serving time and at any scale.
- GitHub stars
- 7.2k
- Last commit
- today
- Latest release
- v8.781.11
- Licence
- Apache-2.0
- Self-hosted
- Yes
- Hosted version
- Available
Vespa is a platform that searches, makes inferences in and organizes vectors, tensors, text and structured data, at serving time and at any scale. The GitHub repository holds all the code needed to build and run Vespa, licensed under Apache 2.0. Its topics include search engine, vector database and RAG, and the main language is Java.
The README frames the problem as use cases such as search, recommendation and personalization that must select a subset of a large corpus, evaluate machine-learned models over it, then organize, aggregate and return the results, typically in under 100 milliseconds while the data keeps changing. Vespa does these operations with high availability and performance, including across multiple nodes in parallel.
Applications can be deployed to the Vespa cloud service or on your own instance. The getting-started application is described as fully functional and production-ready, with more nodes advised for redundancy. Java components can be added to an application. A new release is built from the master branch every morning CET, Monday through Thursday, and the README says it serves hundreds of thousands of queries per second on several large internet services.
Key features
- Search and inference over vectors, tensors, text and structured data
- Evaluates machine-learned models over selected data
- High availability and distribution over multiple nodes
- Deploy to a cloud service or run your own instance
- Custom Java components in an application
- Frequent releases from the master branch
Pricing: A free trial for running applications in the cloud is offered; no paid prices are listed.
#9
Sphinx is a documentation generator that supports reStructuredText and MyST Markdown, with features like cross-referencing, multiple output formats, themes, extensions, and internationalization.
- GitHub stars
- 8.1k
- Last commit
- 4 days ago
- Latest release
- v9.2.0
sphinx-doc.org
Sphinx is a documentation generator that allows authors to write in reStructuredText or MyST Markdown to create structured technical documents. It supports tables, highlighted code blocks, mathematical notations, and cross-references within and across projects.
The tool can generate documentation in formats such as HTML, LaTeX (for PDF), ePub, and Texinfo. It offers built-in and third-party HTML themes, extension mechanisms for custom functionality, and automatic API documentation for Python, C++, and other software domains.
Key features
- Rich text formatting with reStructuredText or MyST Markdown
- Cross-referencing within and across projects
- Multiple output formats: HTML, LaTeX, ePub, Texinfo
- Extensive theme support with built-in and third-party themes
- Extensible via built-in and third-party extensions
- Automatic API documentation for Python, C++, and other domains
#10
Federated AI search and RAG that queries your existing apps live, ranks results, and returns cited answers without copying data into a vector database.
- GitHub stars
- 3k
- Last commit
- 17 days ago
- Latest release
- v4.5.0.7
- Licence
- Apache-2.0
- Self-hosted
- Yes
swirlaiconnect.com
SWIRL is an AI search and retrieval-augmented generation tool that works without moving your data. Instead of copying content into a vector database, it queries your sources live with each user's own permissions, re-ranks the results, and can generate an answer with citations using the language model of your choice. The Community edition in this repository is released under the Apache-2.0 license.
Connectors reach many business applications, and the project's tagline refers to more than 100 apps. The approach avoids building ETL pipelines, standing up a vector store, or maintaining a second copy of data that needs security and audit controls, since permissions are enforced at the source. The stack is Python and Django, and a Docker quick start gets an instance running in about two minutes.
A separate SWIRL Enterprise edition adds a three-pass reranker, canonical answers, an MCP server for agents, and managed support. The Community edition is free to self-host. It suits IT and data teams wanting enterprise search across many systems while keeping data in place.
Key features
- Federated search across connected apps
- Live queries with source-level permissions
- No vector database or ETL needed
- Cited answers from your choice of language model
- Django-based with Docker quick start
- Enterprise tier adds reranking and MCP server
Pricing: Annual platform licences: Departmental $24K, Business $72K and Enterprise from $150K per year. No free plan is listed; a paid $12,000 pilot is credited to the first annual contract.
4 more IBM Watson Discovery alternatives
IBM Watson Discovery alternatives: questions
- What is the best open-source alternative to IBM Watson Discovery?
- Onyx is the top-ranked open-source alternative to IBM Watson Discovery on Enlisted: An open-source AI platform for enterprise search and chat that indexes knowledge from 50+ apps and works with any LLM, deployable on your own infrastructure. Other strong options are Elasticsearch, Meilisearch, Qdrant and Typesense.
- What are the best commercial alternatives to IBM Watson Discovery?
- Mindbreeze, Amazon Kendra, SearchBlox, Sinequa and Azure AI Search are among the hosted, closed-source alternatives to IBM Watson Discovery listed on Enlisted, which lists 16 in total. Compare their pricing and features on each tool page.
- Are these IBM Watson Discovery alternatives free?
- All 14 are open source, so the code is free to use under its licence, and all of them can be self-hosted on your own server or computer. 7 also offer a paid or managed cloud version if you'd rather not host it yourself.
- How is this list of IBM Watson Discovery alternatives ranked?
- By a score built from GitHub stars, star growth over the last 30 days and how recently the code changed. 12 of these projects shipped code in the last 30 days. Data is refreshed daily, and nobody can pay to move up.