Swirl
Swirl is an open-source enterprise search and AI tool that enables users to search multiple enterprise applications and sites with a single...
Last verified:
What is Swirl?
Swirl is a federated AI search engine that connects generative AI and LLMs to over 100 enterprise data platforms, applications, and services without any data copying, ingestion, or indexing. It runs in your Docker or Kubernetes environment, enabling secure, real-time access to authorized data for personalized AI-driven insights. Setup is quick via simple configuration, bypassing developers, ETL projects, or legal hurdles.
Key features include hybrid metasearch with relevancy re-ranking using embeddings from configured LLMs, asynchronous querying across silos, and state-of-the-art RAG for summarization, question answering, and visualizations. The AI Search Assistant acts as a conversational chatbot that identifies optimal sources, generates SQL/SPARQL queries, and provides inline insights with citations. It supports Galaxy UI, Swagger APIs, SSO/OAuth2, and extraction from 1500+ file formats including images and tables.
Designed for enterprises with siloed data, Swirl suits IT teams, analysts, and leaders needing fast visibility without migrations. Ideal for air-gapped deployments, compliance-heavy sectors like legal, and organizations using tools like Snowflake, SharePoint, Salesforce, or any LLM from OpenAI to Ollama.
Swirl Enterprise excels where authentication, multi-LLM support, and advanced RAG on complex docs are required, while Community Edition fits simple, unauthenticated RAG across repos under Apache 2.0.
Swirl pricing
Pricing model: Free
Community Edition is free and open-source under Apache 2.0 for unauthenticated RAG. Enterprise Edition pricing varies by deployment type, features, and support; available via Azure Marketplace offers for quick VM installs or contact for license keys/Docker/Helm charts. Detailed plans at swirlaconnect.com/pricing.
Swirl pros
- Zero data migration or ingestion
- Connects to 100+ enterprise platforms
- Re-ranks results across sources for relevancy
- Supports any LLM including on-premises Ollama
- Deploys in minutes with Docker Compose
- SSO and OAuth2 for secure access
- RAG on 1500+ file formats and images
- AI Search Assistant for conversational queries
- Real-time data access without ETL
- On-premises and air-gapped compatible
- No code required for setup
- Swagger APIs for easy integration
- Generates SQL/SPARQL via AI
- Maintains existing permissions
- Reduces costs like $3M in case studies
- Universal search outperforming vector DBs
Swirl cons
- Response time depends on slowest source
- Requires Docker/K8s infrastructure
- Community lacks SSO/multi-LLM support
- Enterprise pricing not public
- Needs config for new sources
- Async processing may delay insights
- Limited to API-queryable endpoints
- No native support for non-API silos
- Relies on LLM quality for re-ranking
Frequently asked questions about Swirl
What is SWIRL AI Search?
SWIRL AI Search is a hybrid metasearch engine that connects Generative AI or LLMs to enterprise data platforms without copying, ingesting, or indexing data. It runs in Docker/Kubernetes, queries authorized APIs asynchronously, re-ranks unified results using LLM embeddings, and enables RAG for insights like summarization and QA.
What is SWIRL AI Search Assistant?
SWIRL AI Search Assistant is a chatbot that converses to pinpoint best data sources and crafts precise queries, including SQL/SPARQL. It triggers RAG on relevant results for inline insights, maintains session-only context, and respects user authorizations via SSO.
How does SWIRL avoid data ingestion?
SWIRL sends queries to source APIs, normalizes/ re-ranks responses via vector similarity (spaCy/embeddings), and fetches full text only for RAG prompts. No data is stored; it leverages Xethub-proven re-ranking over vector DBs for QA tasks.
What systems does SWIRL integrate with?
Over 100 including AI/LLMs (Anthropic, Grok, Ollama), data platforms (Snowflake, Postgres, Elastic), enterprise apps (Salesforce, Jira, SharePoint), and collaboration tools (Google Drive, Slack). Full list at swirlaconnect.com/connectors.
How to connect a new data source?
Create a SearchProvider record defining API interactions, JSONPath/XPath normalization. For LLMs, configure AIProvider with roles for chat, embeddings, RAG. Guides in docs cover custom integrations.
What differs Community vs Enterprise Edition?
Community: Free for unauthenticated RAG, single AI support. Enterprise: Adds AI Assistant, multi-LLM, SSO/OAuth2, auth PageFetcher, image/table extraction, configurable prompts; for authenticated/complex use cases.
How is SWIRL deployed?
Via Docker Compose for quick setup; Kubernetes images for scale. Enterprise available on Azure Marketplace VMs or private clouds/on-premises with chosen LLM.
What technology stack powers SWIRL?
Python/Django/Celery/Redis core, PostgreSQL recommended DB. Supports async tasks, APIs, and embeddings for re-ranking/RAG.
Does SWIRL support on-premises LLMs?
Yes, integrates Ollama and any custom API LLM on-premises/private clouds, ensuring IP security and air-gapped ops while using existing SSO.
How to get SWIRL Enterprise?
Use Azure Marketplace for Legal/Enterprise offers on private VMs, or contact for license keys/Docker/Helm on other platforms. Enterprise support via tickets.