Building your own transaction sending logic is the best way to ensure maximum performance, control, and reliability for your application. While the Helius SDK provides a convenient wrapper for getting started, understanding and implementing this manual workflow is highly recommended for production systems. This guide will walk you through the necessary steps to build your own solution.

​
The Manual Workflow

Manually sending a transaction involves the following steps:
1

Build the Initial Transaction

Assemble your instructions and sign the transaction so it can be simulated.
2

Optimize Compute Units

Simulate the transaction to determine the precise CUs needed and add a small buffer.
3

Add Priority Fees

Get a fee estimate from the Helius Priority Fee API and add it to your transaction.
4

Send and Re-broadcast

Send the final transaction and implement a robust polling strategy to handle confirmation.
The Helius SDKs are open-source. You can view the underlying code for the sendSmartTransaction method in our Node.js SDK and Rust SDK to see a production-grade implementation of this workflow.

​
1. Build the Initial Transaction

First, gather all the instructions you want to include in your transaction. Then, create a Transaction or VersionedTransaction object. You will also need to fetch a recent blockhash. This example prepares a versioned transaction. At this stage, you must also sign it so that it can be simulated in the next step. 
import {
  Connection,
  Keypair,
  TransactionMessage,
  VersionedTransaction,
  SystemProgram,
  LAMPORTS_PER_SOL,
} from "@solana/web3.js";

const connection = new Connection("YOUR_RPC_URL");
const fromKeypair = Keypair.generate(); // Assume this is funded
const toPubkey = Keypair.generate().publicKey;

// 1. Build your instructions
const instructions = [
  SystemProgram.transfer({
    fromPubkey: fromKeypair.publicKey,
    toPubkey: toPubkey,
    lamports: 0.001 * LAMPORTS_PER_SOL,
  }),
];

// 2. Get a recent blockhash
const { blockhash } = await connection.getLatestBlockhash();

// 3. Compile the transaction message
const messageV0 = new TransactionMessage({
  payerKey: fromKeypair.publicKey,
  recentBlockhash: blockhash,
  instructions,
}).compileToV0Message();

// 4. Create and sign the transaction
const transaction = new VersionedTransaction(messageV0);
transaction.sign([fromKeypair]);

​
2. Optimize Compute Unit (CU) Usage

To avoid wasting fees or having your transaction fail, you should set the compute unit (CU) limit as precisely as possible. You can do this by simulating the transaction using the simulateTransaction RPC method. It’s a best practice to first simulate with a high CU limit to ensure the simulation itself succeeds, and then use the unitsConsumed from the response to set your actual limit. 
import { ComputeBudgetProgram } from "@solana/web3.js";

// Create a test transaction with a high compute limit to ensure simulation succeeds
const testInstructions = [
    ComputeBudgetProgram.setComputeUnitLimit({ units: 1_400_000 }),
    ...instructions, // Your original instructions
];
const testMessage = new TransactionMessage({
    payerKey: fromKeypair.publicKey,
    recentBlockhash: blockhash,
    instructions: testInstructions,
}).compileToV0Message();
const testTransaction = new VersionedTransaction(testMessage);
testTransaction.sign([fromKeypair]);


// Simulate the transaction to get the exact CUs consumed
const { value: simulationResult } = await connection.simulateTransaction(testTransaction);

if (!simulationResult.unitsConsumed) {
  throw new Error("Simulation failed to return unitsConsumed");
}

// Add a 10% buffer to the CU estimate
const computeUnitLimit = Math.ceil(simulationResult.unitsConsumed * 1.1);

// Create the instruction to set the CU limit
const setCuLimitInstruction = ComputeBudgetProgram.setComputeUnitLimit({
    units: computeUnitLimit,
});
Now you have an instruction that sets the compute limit precisely. You will add this to your final transaction.

​
3. Set the Right Priority Fee

Next, determine the optimal priority fee to add to your transaction. Using the Helius Priority Fee API is the best way to get a real-time estimate based on current network conditions. You’ll need to call the getPriorityFeeEstimate RPC method. For the highest chance of inclusion via Helius’s staked connections, use the recommended: true option. 
// The transaction needs to be serialized and base58 encoded
const serializedTransaction = bs58.encode(transaction.serialize());

const response = await fetch("YOUR_RPC_URL", {
    method: "POST",
    headers: { "Content-Type": "application/json" },
    body: JSON.stringify({
        jsonrpc: "2.0",
        id: "1",
        method: "getPriorityFeeEstimate",
        params: [
            {
                // Pass the serialized transaction
                transaction: serializedTransaction, 
                // Use 'recommended' for Helius's staked connections
                options: { recommended: true },
            },
        ],   
    }),
});
const data = await response.json();

if (!data.result || !data.result.priorityFeeEstimate) {
    throw new Error("Failed to get priority fee estimate");
}

const priorityFeeEstimate = data.result.priorityFeeEstimate;

// Create the instruction to set the priority fee
const setPriorityFeeInstruction = ComputeBudgetProgram.setComputeUnitPrice({
    microLamports: priorityFeeEstimate,
});

​
4. Build, Send, and Confirm

Now, assemble the final transaction with the new compute budget instructions, send it, and implement a robust polling mechanism to confirm it has landed.
Do not rely on the RPC provider’s default retry logic (maxRetries in sendTransaction). While Helius’s staked connections forward your transaction directly to the leader, it can still be dropped. You must implement your own rebroadcasting logic for reliable confirmation.
A common pattern is to re-send the same transaction periodically until the blockhash expires. Only re-sign the transaction if you are also fetching a new blockhash. Re-signing with the same blockhash can lead to duplicate transactions being confirmed. 
// 1. Add the new instructions to your original set
const finalInstructions = [
  setCuLimitInstruction,
  setPriorityFeeInstruction,
  ...instructions,
];

// 2. Re-build and re-sign the transaction with the final instructions
const { blockhash: latestBlockhash, lastValidBlockHeight } = await connection.getLatestBlockhash();

const finalMessage = new TransactionMessage({
  payerKey: fromKeypair.publicKey,
  recentBlockhash: latestBlockhash,
  instructions: finalInstructions,
}).compileToV0Message();

const finalTransaction = new VersionedTransaction(finalMessage);
finalTransaction.sign([fromKeypair]);

// 3. Send the transaction
const signature = await connection.sendTransaction(finalTransaction, {
  skipPreflight: true, // Optional: useful for bypassing client-side checks
});

// 4. Implement a polling loop to confirm the transaction
let confirmed = false;
while (!confirmed) {
    const statuses = await connection.getSignatureStatuses([signature]);
    const status = statuses && statuses.value && statuses.value[0];

    if (status && (status.confirmationStatus === 'confirmed' || status.confirmationStatus === 'finalized')) {
        console.log('Transaction confirmed!');
        confirmed = true;
    }

    // Check if the blockhash has expired
    const currentBlockHeight = await connection.getBlockHeight();
    if (currentBlockHeight > lastValidBlockHeight) {
        console.log('Blockhash expired, transaction failed.');
        break;
    }
    
    // Wait for a short period before polling again
    await new Promise(resolve => setTimeout(resolve, 2000)); 
}
This example provides a basic polling loop. A production-grade application would require more sophisticated logic, including handling different confirmation statuses and potential timeouts.