What Sonic is
Sonic is a free, open-source, self-hosted search backend for indexing text with identifier tuples. It returns identifiers for matching documents, leaving applications to retrieve the corresponding records from an external database; direct textual data is not stored in the index. The backend can normalize natural-language queries, correct typos when exact matches fail, and suggest completions while users type. Its lexer supports more than 80 languages and can remove stop words after detecting a language. Terms are organized into collections and buckets, including user-specific buckets. Search, ingestion and object management use Sonic Channel, a TCP protocol, rather than HTTP. Official clients are listed for Node.js, Rust and PHP, with community libraries for several other languages. Installation options include releases, Cargo, source, Docker and Debian packages targeting 64-bit Debian 12. Sonic uses the MPL-2.0 license.
Who it suits
Sonic suits developers who need a lightweight search backend they can run themselves on Linux. It fits applications that resolve returned identifiers through their own database.
Good to know, and what to watch
- Self-hosted and open source
- Supports more than 80 languages
- Typo correction and query completion included
- Official Node.js, Rust and PHP clients
- Multiple installation methods documented
- No HTTP endpoint is provided
- Applications must resolve identifiers externally
- Official Debian packages target 64-bit Debian 12
- Result windows retain only recent pushed results
Plans
Every plan and priceSonic
Open source
The verdict
Sonic provides text search while keeping document data in an external store. Its TCP protocol, self-hosted deployment and language support make integration choices important before adoption.
What the maker publishes
- Deployment options
- self-hosted source
- Deployment
- The project documents installation from releases, Cargo, source, Docker, and Debian-based packages; its official packages target 64-bit Debian 12. source
- Headquarters
- Nantes, France source
- Index model
- It returns identifiers for matching documents, which applications can resolve in an external database; Sonic does not store direct textual data in its index. source
- Integrations
- The project lists official client libraries for Node.js, Rust, and PHP, plus community libraries for languages including Python, Ruby, Go, Java, Deno, Bun, and .NET. source
- Intended use
- The project presents Sonic as a lightweight alternative to heavyweight search backends for some use cases and names Crisp and Scrumpy among its users. source
- Language support
- Its lexing system supports over 80 languages, with stop-word removal after language detection; unsupported languages can be indexed without stop-word filtering. source
- Maker location
- The README says Sonic was crafted in Nantes, France. source
- Organization
- Search terms are stored in collections and buckets, including separate buckets for user-specific indexes. source
- Protocol
- Search, ingestion, and object management use Sonic Channel, a TCP-based protocol; Sonic does not provide an HTTP endpoint. source
- Purpose
- Sonic is a fast, lightweight, schema-less search backend that indexes search text and identifier tuples for querying. source
- Result window
- For each word, Sonic retains only the N most recently pushed results, with the sliding-window width configurable. source
- Search features
- It can normalize natural-language queries, correct typos when exact matches are insufficient, and suggest word completions in real time. source
- Security
- The repository identifies its license as MPL-2.0; the reviewed project pages do not state security certifications or compliance claims. source
- Storage limit
- Each bucket uses a 32-bit internal identifier, theoretically allowing about 4.2 billion indexed objects per bucket. source
Checked against the maker's pages on 10 Oct 2026.