Frequently asked questions
Straight answers, including where the answer is "not yet."
- Is Tachyon production ready?
- Tachyon is under active development. Every feature documented is tested end to end, but you should read Persistence for the exact durability guarantees before relying on it for data you can't afford to lose, and check the Roadmap for what hasn't shipped yet — clustering and replication in particular, which affect how much redundancy you have in production.
- How much RAM does Tachyon need?
- There's no fixed number — memory usage tracks how much of the corpus is actively being queried, not total corpus size, because segments are memory-mapped and decoded lazily. See Persistence for how that works, and the preliminary benchmark for measured memory figures at 100K/1M/5M documents as a reference point.
- Does Tachyon support vector or semantic search?
- No. Tachyon does lexical (BM25) search, not vector or hybrid search — it is explicitly not a vector database. See the Roadmap for the current status; it is not on the near-term roadmap today.
- Can Tachyon replace Elasticsearch?
- For application search specifically — searching, filtering, faceting, sorting a document collection — often yes. For log analytics, aggregations, or anything that depends on the broader ELK/Kibana ecosystem, no; Tachyon doesn't have an aggregation framework or a dashboard product. See Tachyon vs. Elasticsearch for the detailed comparison.
- Is Tachyon free? What license is it under?
- Yes. Tachyon is open source under the Apache-2.0 license, free to self-host with no usage limits, no per-record or per-search cost, and no copyleft obligations on how you use it.
- How does Tachyon compare to Typesense?
- They're the closest architectural match on the market: both typo-tolerant, self-hosted, single-binary engines. The biggest differences are memory model (Typesense holds its index in RAM; Tachyon mmaps segments) and license (GPL-3.0 vs. Apache-2.0). See Tachyon vs. Typesense for the full comparison, including a preliminary benchmark.
- Does Tachyon support clustering or replication?
- Not yet. Tachyon runs as a single node today, by design — see Architecture → Future Architecture. Distributed search, replication, and sharding are the next major milestones on the Roadmap.
- What languages does Tachyon have official SDKs for?
- TypeScript/JavaScript, Python, and C#/.NET — see SDKs. Anything else can use the REST API directly; it's plain JSON over HTTP with no SDK required.
- How do I run Tachyon in production?
- As a single Docker container with a persistent volume for `--data-dir` — see Deployment. There's no official Compose or Kubernetes manifest, but the Docker Compose and Kubernetes guides walk through adapting Tachyon's image and flags to either.
- Does Tachyon support typo tolerance out of the box?
- Yes, on by default, using Damerau-Levenshtein edit distance scaled to query-token length. See Typo Tolerance for the exact thresholds and how to tune or disable it per collection.