AccelerateSearch

An open-source, Rust-native search engine built for speed, scale, and self-hosting.

---

Warning

This project is currently in active development. APIs, on-disk formats, configuration keys, and CLI flags may change without notice. Pin a specific commit or release tag if you need stability, and expect breaking changes between minor versions until 1.0.0.

---

What is AccelerateSearch?

AccelerateSearch is a production-grade, self-hosted search engine written in Rust. It speaks a clean REST API, ships as a single binary, and provides the developer experience of Meilisearch with the power of Elasticsearch, the simplicity of Typesense, and the vector capabilities of Qdrant, all in one Rust-native package.

It targets engineers who want to embed a search engine in their app, build it into a website, or wire it into an internal toolchain, without operating a JVM, an Elasticsearch cluster, or a managed service.

Why AccelerateSearch?

• Single binary with no JVM, no service mesh, no sidecar, and no separate database.
• Sub-10 ms p99 search on commodity hardware for collections under 1 M documents.
• First-class vector search with HNSW indexing, cosine, dot, and Euclidean similarity.
• Hybrid ranking that fuses BM25 and vector signals with Reciprocal Rank Fusion.
• Meilisearch-style developer experience with tasks, settings, bulk ingestion, and scoped API keys.
• Apache-2.0 licensed for use in commercial or open-source projects without strings.

Note

Sub-second BM25 search, vector KNN, hybrid ranking, typo tolerance, synonyms, faceting, highlighting, a single binary that replaces your search stack.

Built in Rust. Runs anywhere. Zero external services required.

Installation

Pick whichever path fits your workflow. All four methods install the same binary; the difference is only in how the source gets onto your machine and how the binary lands on your PATH.

1. Download a Pre-built Binary (recommended)

The fastest path. Grab the latest release for your platform from GitHub Releases:

Platform	File
Linux (x86_64)	accelerate-linux-amd64
Linux (ARM64)	accelerate-linux-arm64
macOS (Intel)	accelerate-macos-amd64
macOS (Apple Silicon)	accelerate-macos-arm64
Windows	accelerate-windows-amd64.exe

# Linux / macOS
curl -L -o accelerate \
  https://github.com/muhammad-fiaz/AccelerateSearch/releases/latest/download/accelerate-$(uname -s | tr '[:upper:]' '[:lower:]')-amd64
chmod +x accelerate
sudo mv accelerate /usr/local/bin/accelerate
accelerate version

# Windows (PowerShell)
Invoke-WebRequest -Uri "https://github.com/muhammad-fiaz/AccelerateSearch/releases/latest/download/accelerate-windows-amd64.exe" `
  -OutFile "$env:ProgramFiles\accelerate.exe"
& "$env:ProgramFiles\accelerate.exe" version

2. Install with cargo install (from the Git repo)

If you already have the Rust toolchain installed, you can install AccelerateSearch directly from the GitHub repository without cloning it first:

# Latest commit on the default branch
cargo install --git https://github.com/muhammad-fiaz/AccelerateSearch --bin accelerate

# A specific tag (recommended for reproducible builds)
cargo install --git https://github.com/muhammad-fiaz/AccelerateSearch \
              --tag v0.0.0 \
              --bin accelerate

# A specific commit (for pinning in CI)
cargo install --git https://github.com/muhammad-fiaz/AccelerateSearch \
              --rev <commit-sha> \
              --bin accelerate

The compiled binary lands in ~/.cargo/bin/accelerate (add ~/.cargo/bin to your PATH if it is not already).

3. Clone and Build from Source

Use this path if you want to hack on the code, run the test suite, or produce a debug build.

# Clone the repository
git clone https://github.com/muhammad-fiaz/AccelerateSearch
cd AccelerateSearch

# Debug build (fast compile, slower runtime)
cargo build

# Release build (slow compile, fast runtime — the binary the
# release artefacts ship)
cargo build --release

# Optional: install the freshly built binary onto your PATH
cargo install --path . --bin accelerate

# Run the test suite
cargo test --workspace --no-fail-fast

After cargo build --release the binary is at target/release/accelerate (or accelerate.exe on Windows). Move it to a directory on your PATH:

# Linux / macOS
sudo cp target/release/accelerate /usr/local/bin/accelerate

# Windows
copy target\release\accelerate.exe C:\Windows\System32\accelerate.exe

4. Docker

Use the published image, or build it yourself from the included Dockerfile:

# Pull (when published)
docker pull ghcr.io/muhammad-fiaz/accelerate:latest

# Or build locally
docker build -t accelerate:local .

# Run, persisting the data dir on the host
docker run --rm -p 7700:7700 \
  -v "$PWD/data:/data" \
  accelerate:local

The full docker-compose.yml is in the repository root for a one-command bring-up.

Quick Start

Start the Server

# Start with default config (binds to http://localhost:7700)
accelerate start

# Start with custom options
accelerate start \
  --host 0.0.0.0 \
  --port 7700 \
  --data-dir /var/lib/accelerate \
  --log-level info \
  --master-key $(openssl rand -hex 32)

Tip

The default bind address is localhost (loopback only). Pass --host 0.0.0.0 to listen on every interface, or --host 127.0.0.1 to keep the same loopback-only behaviour. The server exposes a full REST API under /api/v1/, ships a Swagger UI at /swagger-ui/, and prints a structured banner with version, environment, and data dir on startup.

CLI Surface

accelerate --help
accelerate start [--config FILE] [--host ADDR] [--port N]
                 [--data-dir DIR] [--master-key KEY] [--log-level LEVEL]
                 [--env ENV] [--no-color] [--quiet] [--no-banner]
                 [--no-console] [--no-file] [--disable-auth]
                 [--no-swagger] [--no-openapi]
accelerate version
accelerate health [--url URL]
accelerate snapshot create | list | info NAME | delete NAME | restore --path FILE

Flag	Description
--no-color	Strip ANSI colour from every console surface.
--quiet	Silence every non-error log line and skip the banner.
--no-banner	Skip the ASCII art banner at startup.
--no-console	Disable console log output entirely.
--no-file	Disable the rotating file log appender.
--no-swagger	Disable the Swagger UI at /swagger-ui.
--no-openapi	Disable the OpenAPI spec at /api-docs/openapi.json.

Warning

Every flag is also exposed as an ACCELERATE_* environment variable, and the TOML configuration can override defaults at the lowest precedence.

Documentation

The user guide is an mdbook site in docs/ and is published to GitHub Pages at https://muhammad-fiaz.github.io/AccelerateSearch/.

Build the docs locally with cargo:

# Install the toolchain helper (one-off)
cargo install mdbook --locked --version 0.4.43

# Build the mdbook user guide into docs/book/
(cd docs && mdbook build)

# Build the Rust API reference into target/doc/
cargo doc --no-deps --workspace --target-dir target

# Combine both into a single ./site/ directory ready for GitHub Pages
mkdir -p site
cp -R docs/book/. site/
mkdir -p site/rust-api
cp -R target/doc/. site/rust-api/
touch site/.nojekyll

The full reference (deployed to GitHub Pages):

• architecture.md: module dependency graph and data flow.
• api.md: full REST API reference.
• configuration.md: every TOML key with type, default, and description.
• deployment.md: Docker, systemd, and bare-metal deployment.
• development.md: build, test, benchmark, contribute.
• rust-api.md: links to the generated cargo doc output.

Note

The OpenAPI spec is served at runtime from /api-docs/openapi.json and rendered interactively at /swagger-ui.

Every push to main rebuilds the docs site and deploys it to gh-pages via the .github/workflows/docs.yml workflow.

The full architecture (crate dependency graph, data flow, storage layout) lives in docs/architecture.md.

Testing

cargo test --workspace

Tip

All tests are inline in the source files (#[cfg(test)] modules). No separate tests/ folder is used.

Benchmarks

cargo bench --bench evaluate

The unified benches/evaluate.rs suite exercises every critical path: tokenisation, indexing, search, filter evaluation, faceting, highlighting, typo tolerance, vector KNN, hybrid RRF, collection CRUD, and HTTP round-trips. Results are written to target/criterion/.

Contributing

See CONTRIBUTING.md for development setup, code style, and the pull-request workflow. Issues and feature requests are tracked on GitHub.

License

Copyright 2026 Muhammad Fiaz. Licensed under the Apache License, Version 2.0.

Community & Support

• Open an issue
• Request a feature
• Ask for help
• Star the repository