srake

README

SRAKE - SRA Knowledge Engine

Pronounced like Japanese sake (酒) — "srah-keh"

SRAKE is a tool for ingesting, searching, and serving NCBI SRA (Sequence Read Archive) metadata. It streams multi-gigabyte compressed archives directly into a local SQLite database, supports full-text and semantic vector search, and exposes results via CLI, REST API, or MCP.

Pre-alpha software — developed at BioHackathon 2025, Mie, Japan. Not production-ready. APIs, schemas, and behavior may change without notice.

Installation

From source (requires Go 1.25+, CGO, SQLite3)

git clone https://github.com/nishad/srake.git
cd srake
go build -tags "sqlite_fts5,search" -o srake ./cmd/srake

Pre-built binaries

Download from the releases page. Available for linux/amd64, linux/arm64, darwin/amd64, and darwin/arm64.

Docker

docker pull ghcr.io/nishad/srake:latest
docker run -v $(pwd)/data:/data ghcr.io/nishad/srake:latest --help

Quick start

# Ingest SRA metadata (auto-selects best source from NCBI)
srake ingest --auto

# Ingest a specific local archive
srake ingest --file /path/to/archive.tar.gz

# Search
srake search "homo sapiens" --limit 10

# Start the API server
srake server --port 8080

# Start MCP server for AI assistants
srake mcp --transport stdio

Commands

srake ingest — Ingest SRA metadata

Streams tar.gz archives from NCBI (or local files) directly into SQLite without intermediate extraction.

srake ingest --auto                # auto-select best source
srake ingest --daily               # latest daily update
srake ingest --monthly             # full monthly dataset
srake ingest --file archive.tar.gz # local file
srake ingest --list                # list available files on NCBI

Filtering during ingest:

srake ingest --file archive.tar.gz \
  --taxon-ids 9606 \
  --platforms ILLUMINA \
  --strategies RNA-Seq \
  --date-from 2024-01-01 \
  --min-reads 1000000

# Preview what would be ingested without inserting
srake ingest --file archive.tar.gz --taxon-ids 9606 --stats-only

srake search — Search metadata

Supports multiple search modes: database (SQLite FTS5), text (Bleve), vector (semantic embeddings), hybrid (combined), and auto (default).

srake search "breast cancer" --limit 20
srake search "RNA-Seq" --organism "homo sapiens" --platform ILLUMINA
srake search "tumor gene expression" --search-mode vector
srake search "mouse brain" --similarity-threshold 0.7 --show-confidence
srake search "covid" --format json --output results.json

Output formats: table (default), json, csv, tsv.

srake server — REST API server

srake server --port 8080 --enable-cors

Endpoints:

Method	Path	Description
GET	/api/v1/search?query=...	Search metadata
GET	/api/v1/studies/{accession}	Get study
GET	/api/v1/experiments/{accession}	Get experiment
GET	/api/v1/samples/{accession}	Get sample
GET	/api/v1/runs/{accession}	Get run
GET	/api/v1/stats	Database statistics
GET	/api/v1/health	Health check
POST	/api/v1/export	Export results

srake mcp — MCP server for AI assistants

Implements Model Context Protocol over stdio or HTTP.

srake mcp --transport stdio    # for Claude Desktop, etc.
srake mcp --transport http --port 8081

Provides tools: search_sra, get_metadata, find_similar, export_results.

srake metadata — Accession lookup

srake metadata SRR12345678 --format json
srake metadata SRP123456 SRX654321 --format yaml

srake db — Database management

srake db info                  # show database statistics
srake db stats --rebuild       # rebuild pre-computed statistics
srake db export -o out.sqlite  # export to SRAmetadb-compatible format
srake db export -o out.sqlite --fts-version 3  # FTS3 for legacy compatibility

srake models — Embedding model management

srake models list
srake models download Xenova/SapBERT-from-PubMedBERT-fulltext
srake models test <model-id> "sample text"

Vector search

SRAKE supports semantic search using biomedical embeddings (SapBERT via ONNX Runtime). This requires downloading a model and building an index with embeddings enabled.

srake models download Xenova/SapBERT-from-PubMedBERT-fulltext
srake search "metabolic pathway analysis" --search-mode vector

Environment variables

Variable	Description
SRAKE_DB_PATH	Path to metadata database
SRAKE_INDEX_PATH	Path to search index directory
SRAKE_CONFIG_DIR	Configuration directory (default: ~/.config/srake)
SRAKE_DATA_DIR	Data directory (default: ~/.local/share/srake)
SRAKE_CACHE_DIR	Cache directory (default: ~/.cache/srake)
SRAKE_MODEL_VARIANT	Embedding model variant (full, quantized)
NO_COLOR	Disable colored output

Follows XDG Base Directory Specification.

Database schema

Core tables: studies, experiments, samples, runs, submissions, analyses. Junction tables: experiment_samples, statistics. Full-text search via SQLite FTS5 virtual tables.

Development

go build -tags "sqlite_fts5,search" ./...
go test -tags "sqlite_fts5,search" ./...

# Build with version injection
go build -tags "sqlite_fts5,search" \
  -ldflags="-X main.Version=$(git describe --tags) -X main.Commit=$(git rev-parse --short HEAD) -X main.BuildDate=$(date -u +%Y-%m-%dT%H:%M:%SZ)" \
  -o srake ./cmd/srake

License

MIT — Nishad Thalhath, 2025