This is an high-performance Rust client that decodes Solana shred packets directly from the Turbine block propagation layer. Unlike Geyser RPC or standard RPC requests, this decoder ensures you receive transactions the instant they are propagated by the leader validator, providing an unmatched speed advantage for time-sensitive trading operations.
This advanced version supports 11 major Solana DeFi protocols with comprehensive instruction decoding, Address Lookup Table (ALT) resolution, and intelligent caching mechanisms.
- Jemalloc Memory Allocator – Memory management with reduced fragmentation and improved allocation performance compared to default allocator
- Optimized UDP Buffer (256KB) – Prevents packet loss by configuring
SO_RCVBUFto handle high-volume shred bursts without dropping packets - Physical CPU Thread Pool – Rayon thread pool configured to use
num_cpus - 1physical cores, reserving one for system operations - Parallel Transaction Processing – Leverages Rayon for CPU-bound deserialization for maximum throughput
- Async I/O with Tokio – Non-blocking network operations for receiving shreds and serving gRPC streams
- DashMap Concurrent Storage – Lock-free concurrent hash maps for FEC blocks and processed block tracking
- Zero-Copy Shred Handling – Efficient
Box<[u8]>usage to minimize memory allocations
- Automatic Data Shred Recovery – When data shreds are missing, the decoder uses coding shreds to reconstruct the complete payload
- Shredder Integration – Uses Solana's ReedSolomon recovery erasure code processing
- Complete Block Reconstruction – Ensures 99%+ transaction capture rate even with packet loss
- Intelligent Recovery Logging – Detailed logs showing recovery attempts and success rates per FEC set
- RPC-Backed Cache – Fetches and caches Address Lookup Tables from RPC endpoint with configurable TTL (default: 300 seconds)
- Automatic Refresh on Extension – Detects
ExtendLookupTableinstructions and invalidates cache to prevent stale data - Out-of-Bounds Protection – Automatically refreshes lookup tables when indices exceed current size
- Fallback Mechanism – Uses stale cache entries if RPC fetch fails, preventing transaction loss
- Async Resolution – Pre-resolves all lookup tables before parallel processing to maintain async context
- V0 Transaction Support – Full support for versioned transactions with address table lookups
- Periodic FEC Block Cleanup – Removes FEC blocks older than 20 seconds every 10 seconds
- Processed Block Tracking – DashSet-based tracking prevents duplicate processing
- Sub-Millisecond Latency – Broadcasts decoded transactions via gRPC with <0.1ms latency on localhost
- Broadcast Channel Architecture – 1000-message buffer with
tokio::sync::broadcastfor multiple subscribers - Tonic Framework – Production-ready gRPC implementation with Protocol Buffers
- Automatic Stream Management – Handles client connections, disconnections, and backpressure
- Timestamp Precision – Microsecond-precision timestamps for each transaction
- Debug Mode Enhancements – Thread IDs, names, line numbers, and targets in debug builds
- Performance Metrics – Detailed timing for shred collection, FEC completion, and transaction extraction
- Slot Statistics – Tracks FEC blocks processed, transactions extracted, and completion rates per slot
- Error Context – Rich error messages with full context for debugging protocol-specific issues
- Standardized Output Format – Uniform JSON structure across all protocols with:
- Program ID and instruction name
- Protocol identifier
- Raw instruction data (hex)
- Mapped accounts with signer/writable flags
- Parsed instruction arguments
- Borsh Deserialization – Efficient binary deserialization for all instruction types
- Account Safety Checks – Validates account indices before access to prevent panics
- Unknown Account Handling – Gracefully handles missing accounts with "Unknown" placeholders
- Packet Reception – Receives raw shred packets from configured source
- Validation & Assembly – Validates and assembles shreds into complete blocks
- Data Recovery – Reconstructs missing data using advanced error correction
- Transaction Extraction – Extracts and deserializes transactions from assembled blocks
- Protocol Decoding – Identifies and decodes protocol-specific instructions
- Output Streaming – Broadcasts decoded transactions via gRPC to connected clients
- Jemalloc Global Allocator – Optimized for multi-threaded workloads
- Arc-Based Sharing – Shared ownership of FEC blocks and sockets across tasks
- Scoped Lifetimes – Automatic cleanup when FEC blocks complete
- Bounded Channels – Prevents unbounded memory growth in broadcast queue
- Throughput – Handles 1000+ transactions per slot without problems
- CPU Utilization – Scales linearly with available physical cores
- Memory Footprint – ~200-500MB typical usage with automatic GC
- Cache Hit Rate – 95%+ ALT cache hits after warm-up period
- MEV Searchers – Fastest possible transaction feed for arbitrage and liquidations
- Sniper Bots – Catch token launches across 11 protocols instantly
- Market Makers – Real-time pool creation and liquidity events
- Validators – Decode shreds locally without external RPC dependencies
- Analytics Platforms – Comprehensive multi-protocol transaction indexing
- Trading Firms – Production-grade infrastructure for high-frequency strategies
Create a .env file with the following variables:
# RPC endpoint example for Address Lookup Table resolution
RPC_ENDPOINT=https://mainnet.helius-rpc.com/?api-key=YOUR_API_KEY
# ALT cache TTL in seconds (default: 300)
LOOKUP_TABLE_CACHE_TTL_SECONDS=300
# UDP socket for receiving shreds
# Use 8001 or 8002 depending on Jito Shredstream Proxy configuration
# Or use your validator's TVU port for direct connection
UDP_BUFFER_SOCKET=0.0.0.0:8002
# gRPC server endpoint for streaming transactions
GRPC_SERVER_ENDPOINT=0.0.0.0:50051RUSTFLAGS="-C target-cpu=native" cargo build --releaseRUST_LOG=info cargo run --releaseRUST_LOG=debug cargo run --releaseThe release profile in Cargo.toml is configured for maximum performance:
[profile.release]
opt-level = 3 # Maximum optimizations
lto = "fat" # Link-time optimization across all crates
codegen-units = 1 # Single codegen unit for better optimization
panic = "abort" # Smaller binary, faster panics
strip = true # Remove debug symbols
debug = false # No debug info in releaseTransactions are streamed as JSON with the following structure:
{
"signatures": ["5Xm..."],
"slot": 123456789,
"message": {
"header": {
"numRequiredSignatures": 1,
"numReadonlySignedAccounts": 0,
"numReadonlyUnsignedAccounts": 5
},
"recentBlockhash": "9Xm...",
"instructions": [
{
"program_id": "6EF8rrecthR5Dkzon8Nwu78hRvfCKubJ14M5uBEwF6P",
"instruction_name": "Create",
"protocol": "Pumpfun",
"raw_data": "181ec828051c0777...",
"accounts": [
{
"index": 0,
"pubkey": "7Xm...",
"signer": true,
"writable": true
}
],
"parsed_data": {
"name": "MyToken",
"symbol": "MTK",
"uri": "https://...",
"creator": "7Xm..."
}
}
]
}
}- ERROR – Critical failures (RPC errors, deserialization failures)
- WARN – Recoverable issues (missing accounts, cache misses)
- INFO – Key events (FEC completion, cache initialization, GC stats)
- DEBUG – Detailed tracing (shred collection, timing metrics)
- Increase
UDP_BUFFER_SOCKETbuffer size if experiencing packet loss - Adjust
LOOKUP_TABLE_CACHE_TTL_SECONDSbased on network latency - Monitor CPU usage and adjust Rayon thread pool if needed
If you are really interested in the source code, please contact me for details and demo on Discord: .xanr
