A Lightweight Self-Hosted Search Engine with Zero Cluster Tax
Most software teams need fast, reliable keyword and phrase search—not the operational nightmare of managing complex, resource-heavy multi-node search clusters. AveLynx Stretchy was engineered to provide fast search on Cloudflare's edge network, combining hybrid keyword (BM25 full-text) and semantic vector search in a lightweight footprint.
Why Legacy Distributed Search Clusters are Painful for Most Teams
Distributed search clusters are complex systems, but for 90% of web apps, telemetry tools, and enterprise document vaults, they introduce massive operational overhead:
• Exorbitant Memory Requirements: Running production distributed search comfortably requires 16GB to 64GB of RAM per node across multiple shard replicas to avoid out-of-memory (OOM) heap crashes.
• Mapping Exceptions: If a microservice sends an integer in a field that was previously indexed as a float or string, legacy engines throw strict parsing exceptions and drop incoming documents.
• Split-Brain & Cluster Maintenance: Maintaining master nodes, shard rebalancing during node restarts, and JVM garbage collection tuning requires dedicated DevOps time.
• Skyrocketing Infrastructure Bills: Hosting multi-node clusters on cloud infrastructure frequently drives monthly costs between $1,200 and $4,500 for moderate query loads.
The Stretchy Alternative: Fast Hybrid Search with Low Overhead
Stretchy provides hybrid keyword (BM25 full-text) + semantic vector search running on Cloudflare's network. Instead of requiring a distributed cluster, Stretchy is designed as a standalone, zero-coordination search engine:
| Feature | Legacy Distributed Clusters | AveLynx Stretchy |
|---|---|---|
| Base RAM Footprint | 16 GB – 64 GB per node | < 150 MB total |
| Nodes Required | 3+ (Master, Data, Ingest) | 1 (Single container or Serverless) |
| Search Speed | High latency under load | Fast Edge Execution |
| Startup Time | 45 – 90 seconds | < 2 seconds |
| Type Collision Handling | Throws errors & drops events | Auto type coercion & raw storage |
| Air-Gapped Operation | High complexity | Native (Zero internet dependencies) |
How Much Can You Save?
By eliminating cluster coordination overhead and massive memory footprints, teams running search on Stretchy significantly reduce infrastructure complexity and cloud bills.
Instead of spending thousands of dollars each month maintaining a multi-node cluster, you can run Stretchy on a single lightweight VPS, inside your existing Kubernetes cluster, or on edge serverless functions.
Get Started with Stretchy
Test Stretchy in seconds with Docker or integrate using our open-source TypeScript and Rust SDKs:
• Explore the Stretchy search engine features.
• Review the published latency benchmarks.
• Read the architecture and TCO whitepaper.
• Check out the client libraries on GitHub.
Common Questions
Can Stretchy replace complex distributed search clusters for logging and telemetry?
Yes. Stretchy features built-in probabilistic anomaly detection using Welford's algorithm and Chebyshev statistical thresholds, flagging spikes and anomalies without requiring heavy machine learning plugins.
How do I migrate my queries from legacy search engines?
Stretchy features a visual query builder and clean REST API supporting boolean clauses (MUST, SHOULD, MUST_NOT), numeric range filtering, prefix matching, and wildcard queries with zero raw syntax headaches.