Windsurf

Tari Ootle is a decentralized smart contract platform. You write templates in Rust, compile to WASM, and deploy to the Tari L2 network. Deployed instances are called components and hold on-chain state. Follow these rules exactly.

Overview

Tari Ootle is a decentralized application platform built on the Tari Layer 2 network. You build templates (smart contracts) in Rust, compile them to WASM (wasm32-unknown-unknown), publish them to the network, and interact with deployed components (instances of templates) via transactions.

Key concepts:

• Template — A Rust module annotated with #[template] that defines the logic and state structure. Compiled to WASM and deployed to the network.
• Component — A live instance of a template on-chain. Holds state (struct fields) and exposes public methods.
• Resource — A native digital asset (fungible token or non-fungible NFT). Created with ResourceBuilder. Cannot be copied or accidentally destroyed.
• Vault — An on-chain container that holds exactly one type of resource. Must be stored in a component before the function returns.
• Bucket — A temporary container for resources during a transaction. Used to move assets between vaults or to/from method calls.
• Transaction — A set of instructions (CallFunction, CallMethod, etc.) that are signed, submitted, and executed atomically.

Crate ecosystem:

Important macro distinction:

• args! (from tari_template_lib::prelude) — used inside templates for cross-template calls (alias for invoke_args!)
• args! (from tari_ootle_transaction) — used in test code and client code for TransactionBuilder, TemplateTest::call_function, and call_method. Produces Vec<NamedArg>.

Getting Started Workflow

The typical development workflow for building on Tari Ootle:

1. Generate a template project from the official starter repo:

   cargo generate https://github.com/tari-project/wasm-template
   

   This repo contains:

   • wasm_templates/ — Blank template starters (e.g., wasm_templates/empty)
   • examples/ — Complete working examples with templates and client apps (e.g., examples/guessing_game/template, examples/guessing_game/cli)

   When prompted, select the subfolder matching your needs.

2. Write your template in src/lib.rs inside the #[template] module.

3. Build to WASM:

   cargo build --target wasm32-unknown-unknown --release
   

4. Test locally using tari_template_test_tooling (see Testing Templates).

5. Publish to the network via the Wallet Web UI (see Publishing Templates).

6. Interact with your component using a client app built with ootle-rs, the Wallet CLI, or a pre-built example CLI from cargo generate.

Tip: For the guessing game and other example templates, pre-published template addresses are available on the Esmeralda testnet. Check the Tari Ootle guides for current addresses — you can skip publishing and go straight to interacting.

CLI App Workflow Order (IMPORTANT)

When running a CLI client app (e.g., the guessing game CLI), operations MUST happen in this order:

1. Initialize wallet — Create keypairs, select network, connect to indexer
2. Fund admin account — Get tTARI from the faucet
3. Publish the template — Deploy the WASM to the network. Direct the user to the Wallet Web UI at http://127.0.0.1:5100 to publish. Do NOT write custom publish code — the Web UI handles fee estimation, upload, and provides the template address.
4. Create/deploy the game component — Instantiate the template on-chain (requires the template address from step 3)
5. Register players — Add player accounts AFTER the game component exists
6. Play — Start rounds, make guesses, end games

CRITICAL: Never register players or add users before the template is published and the game component is deployed. Players need a component to interact with. Do not add a publish command to an example CLI unless the user asks for one; for a one-off publish, direct them to the Wallet Web UI.

Tooling Requirements

The only tools needed for Ootle development are:

• rustup with the wasm32-unknown-unknown target
• cargo-generate (for scaffolding)
• Standard Rust toolchain (cargo build, cargo test)
• wasm-opt — optional, for checking or shrinking the binary yourself (see Minifying the WASM)

Do NOT install rust-analyzer extensions, cargo-expand, wasm-pack, wasm-bindgen, or other WASM/Rust analysis tools. They are unnecessary for Ootle development and add bloat. The wasm32-unknown-unknown target and standard cargo build are sufficient.

Non-Interactive CLI Usage

The generated CLI examples use dialoguer for interactive prompts (Select, Input), which requires a real TTY. Some agent runners execute commands in a non-interactive shell without a TTY.

When running CLI commands that have interactive prompts:

• Tell the user to run the command in their terminal rather than trying to run it through the agent runner
• Do NOT try to pipe input, use expect, or wrap with script — these are fragile workarounds
• Do NOT modify the CLI to add non-interactive flags unless the user specifically asks for it
• If the CLI already supports --flag style arguments that bypass prompts, use those

Writing a Template

Project Setup

# Install cargo-generate if not already installed
cargo install cargo-generate

# Generate a new template project from the official starter
cargo generate https://github.com/tari-project/wasm-template

The wasm-template repository offers multiple starting points:

• wasm_templates/empty — A minimal blank template to start from scratch
• examples/guessing_game/template — A complete guessing game template with tests
• examples/guessing_game/cli — A ready-to-use CLI client for the guessing game

When you run cargo generate, select the subfolder that matches your goal. For a blank slate, choose a wasm_templates/ entry. For a working example to learn from, choose from examples/.

A generated template project looks like:

your_template/
├── Cargo.toml
├── src/
│   └── lib.rs      # Template source code
└── tests/
    └── test.rs     # Unit tests

The generated Cargo.toml must include:

[package]
name = "your_template"
version = "0.1.0"
edition = "2024"

[dependencies]
tari_template_lib = "0.31"

[lib]
crate-type = ["cdylib"]

[profile.release]
opt-level = 's'     # Optimize for size.
lto = true          # Enable Link Time Optimization.
codegen-units = 1   # Reduce number of codegen units to increase optimizations.
panic = 'abort'     # Abort on panic.
strip = true          # Strip symbols and debug info.

CRITICAL: The crate-type = ["cdylib"] is required for WASM compilation. Without it, the build will not produce a .wasm file.

CRITICAL: Keep crate-type = ["cdylib"] and nothing else. Adding rlib alongside it — a common attempt to make template code unit-testable — stops Cargo applying lto = true to the target, so the [profile.release] settings below quietly stop shrinking the binary. To unit-test pure logic (tallies, state machines), put it in a separate crate that both the template and the tests depend on.

Tip: The [profile.release] section in Cargo.toml significantly reduces the size of the compiled WASM file, which lowers the fees required for on-chain storage and publishing.

Versions: 0.31 is the current tari_template_lib release on crates.io. Use the minor version (e.g. "0.31" not "0.31.1") to pick up patches automatically, and check crates.io for a newer minor before starting a new template. Keep it in step with tari_template_test_tooling (see Test Setup) — the test harness compiles your template against its own copy of tari_template_lib, and a mismatched pair builds the two halves against different copies of the library types.

Compilation

# Add the WASM target (one-time setup)
rustup target add wasm32-unknown-unknown

# Build the template
cargo build --target wasm32-unknown-unknown --release

Output: target/wasm32-unknown-unknown/release/your_template_name.wasm

Minifying the WASM

Publishing through the wallet daemon — the Web UI, or the transactions.publish_template JSON-RPC method — runs wasm-opt over the binary for you and stores the optimized result, so a Web UI publish is already minified. Run it yourself to check the size before publishing, or when a client publishes a raw binary with TransactionBuilder::publish_template, which stores exactly the bytes it is handed:

wasm-opt -Oz --enable-bulk-memory \
  --strip-debug --strip-producers --strip-target-features \
  target/wasm32-unknown-unknown/release/your_template_name.wasm \
  -o target/wasm32-unknown-unknown/release/your_template_name.min.wasm

--enable-bulk-memory is required: rustc's wasm32-unknown-unknown output uses bulk-memory operations and wasm-opt rejects the module without it.

The strip flags are not optional on a raw publish. The engine accepts no custom section other than tari_tdef, and rustc emits name, producers and target_features sections — a binary that still carries them is rejected at publish time.

Check the size before publishing: ls -la target/wasm32-unknown-unknown/release/*.wasm. Only the first 96 KiB of a template are priced as ordinary storage; every whole kilobyte beyond that is charged quadratically, so keeping the binary near or below 96 KiB is what keeps a publish cheap — see Publishing Templates.

Template Structure

Every template follows this pattern:

use tari_template_lib::prelude::*;

#[template]
mod my_template {
    use super::*;

    // Component state — all fields must be serde-serializable
    pub struct MyComponent {
        my_vault: Vault,
        counter: u64,
    }

    impl MyComponent {
        // CONSTRUCTOR: any function returning Self or Component<Self>
        // creates a new component instance on-chain when called via CallFunction
        pub fn new() -> Component<Self> {
            let token = ResourceBuilder::public_fungible()
                .with_token_symbol("TOK")
                .build(); // → ResourceAddress

            Component::new(Self {
                my_vault: Vault::new_empty(token),
                counter: 0,
            })
            .with_access_rules(
                ComponentAccessRules::new()
                    .method("public_method", rule!(allow_all))
                    .default(rule!(deny_all))
            )
            .create()
        }

        // PUBLIC METHOD: takes &self or &mut self, called via CallMethod
        pub fn public_method(&mut self, value: u64) {
            self.counter += value;
        }

        // READ-ONLY METHOD: takes &self, cannot modify state
        pub fn get_counter(&self) -> u64 {
            self.counter
        }

        // FUNCTION (no &self): called via CallFunction on the template, not a method on a component
        pub fn greet(name: String) -> String {
            format!("Hello, {}!", name)
        }
    }

    // Private helper: no &self, not in impl block's public interface
    fn internal_helper() -> u64 {
        42
    }
}

The #[template] Macro

• Generates the WASM ABI so the validator engine can call your code
• Exposes public functions in impl blocks as callable methods/functions
• Handles state serialization/deserialization automatically
• Structs defined inside the #[template] module automatically get serde derives
• Only one #[template] module per crate

Component State Rules

• All struct fields MUST be serde-serializable (types from tari_template_lib already are)
• Use &mut self methods to modify state
• Use &self methods for read-only access
• A component can also be an enum
• State size affects transaction costs
• Supported field types: all Rust primitives, String, Vec<T>, HashMap<K,V>, BTreeMap<K,V>, Option<T>, Vault, Amount, ResourceAddress, ComponentAddress, RistrettoPublicKeyBytes, NonFungibleId, ComponentManager, and any struct inside the #[template] module

Constructors

// Simple constructor — returning Self creates the component with default rules
pub fn new_simple() -> Self {
    Self { counter: 0 }
}

// Explicit constructor — returns Component<Self> for full control
pub fn new_explicit() -> Component<Self> {
    Component::new(Self { counter: 0 })
        .with_access_rules(ComponentAccessRules::new()
            .method("do_something", rule!(allow_all))
            .default(rule!(deny_all))
        )
        .with_owner_rule(OwnerRule::OwnedBySigner)
        .create()
}

// Constructor with address allocation — allows creating and calling in one transaction
pub fn new_with_allocation(addr: ComponentAddressAllocation) -> Component<Self> {
    Component::new(Self { counter: 0 })
        .with_address_allocation(addr)
        .with_access_rules(ComponentAccessRules::new()
            .method("do_something", rule!(allow_all))
            .default(rule!(deny_all))
        )
        .create()
}

// Constructor with public key address — deterministic address from a public key
pub fn new_with_public_key() -> Component<Self> {
    let pk = CallerContext::transaction_signer_public_key();
    Component::new(Self { counter: 0 })
        .with_public_key_address(pk)
        .with_access_rules(ComponentAccessRules::allow_all())
        .create()
}

Error Handling

Errors in templates are handled by panicking. When a panic occurs, the transaction fails atomically and no state changes are committed:

pub fn do_something(&mut self, value: u64) {
    assert_ne!(value, 0, "Value cannot be zero");
    assert!(value <= 1024, "Value too large");
    if !value.is_power_of_two() {
        panic!("Value must be a power of two");
    }
    self.counter += value;
}

There are no Result-based error flows in templates. Panics are the error mechanism.

Resources

There are 4 resource types:

• Public Fungible — Interchangeable tokens (like ERC-20), amounts visible on-chain
• Public Non-Fungible — Unique tokens (like ERC-721), metadata visible on-chain
• Confidential — Fungible tokens with hidden amounts (Pedersen commitments)
• Stealth — Confidential UTXOs with hidden owners (TARI/tTARI is a stealth resource)

Creating Resources

// ─── Public Fungible Token ───
// Without initial supply → returns ResourceAddress
let token_addr: ResourceAddress = ResourceBuilder::public_fungible()
    .with_token_symbol("TOK")
    .metadata("name", "My Token")
    .build();

// With initial supply → returns Bucket containing tokens
let token_bucket: Bucket = ResourceBuilder::public_fungible()
    .with_token_symbol("TOK")
    .metadata("name", "My Token")
    .initial_supply(Amount::from(1000))
    .build();

// ─── Non-Fungible (NFT) ───
let nft_addr: ResourceAddress = ResourceBuilder::non_fungible()
    .with_token_symbol("NFT")
    .metadata("name", "My NFT Collection")
    .build();    // Mint later via ResourceManager

// With initial supply of NFTs
let nft_bucket: Bucket = ResourceBuilder::non_fungible()
    .with_token_symbol("NFT")
    .initial_supply_with_data(vec![
        (NonFungibleId::from_u64(1), &metadata!["name" => "First"], &()),
        (NonFungibleId::from_u64(2), &metadata!["name" => "Second"], &()),
    ])
    .build();

// ─── Confidential Fungible ───
let conf_addr: ResourceAddress = ResourceBuilder::confidential()
    .with_token_symbol("cTOK")
    .with_view_key(CallerContext::transaction_signer_public_key())
    .build();

// ─── Stealth (like TARI) ───
let stealth_addr: ResourceAddress = ResourceBuilder::stealth()
    .with_token_symbol("sTOK")
    .build();

Resource Builder Options (All Types)

ResourceBuilder::public_fungible()  // or non_fungible(), confidential(), stealth()
    // Metadata
    .with_token_symbol("SYM")              // Token symbol (displayed by explorers)
    .metadata("name", "Token Name")        // Add metadata key-value pair
    .add_metadata("key", "value")          // Same as .metadata()

    // Access rules on the resource itself
    .mintable(rule!(resource(admin_badge))) // Who can mint new tokens (default: deny_all)
    .burnable(rule!(allow_all))            // Who can burn tokens (default: deny_all)
    .recallable(rule!(deny_all))           // Who can forcefully recall from vaults
    .freezable(rule!(deny_all))            // Who can freeze vaults holding this resource
    .withdrawable(rule!(allow_all))        // Who can withdraw (default: allow_all)
    .depositable(rule!(allow_all))         // Who can deposit (default: allow_all)
    .update_non_fungible_data(rule!(...))  // Who can update NFT mutable data
    .update_access_rules(rule!(...))       // Who can change these rules later

    // Ownership
    .with_owner_rule(OwnerRule::OwnedBySigner)  // Owner of the resource definition

    // Advanced
    .with_divisibility(2)                  // Fungible only: decimal places (default: 0)
    .disable_total_supply_tracking()       // Don't track total supply on-chain
    .with_address_allocation(alloc)        // Pre-allocated resource address

    // Finalize
    .build()                               // Create the resource (returns ResourceAddress)
    .initial_supply(Amount::from(1000))    // Mint initial tokens (changes return to Bucket)

Minting NFTs After Creation

let manager: ResourceManager = vault.get_resource_manager();

// Mint a single NFT
let nft_bucket: Bucket = manager.mint_non_fungible(
    NonFungibleId::from_string("unique-id"),   // Unique ID within this resource
    &metadata!["name" => "My NFT"],            // Immutable data (cannot change after mint)
    &(),                                        // Mutable data (can be updated later)
);

// NonFungibleId variants:
NonFungibleId::from_string("my-id")       // String (max 64 chars)
NonFungibleId::from_u64(42)               // u64
NonFungibleId::from_u32(1)                // u32
NonFungibleId::from_u256([0u8; 32])       // 32-byte array
NonFungibleId::random()                   // Random UUID-style

Vaults — Complete API

// ─── Creation ───
let vault = Vault::new_empty(resource_address);      // Empty vault for a resource type
let vault = Vault::from_bucket(bucket);              // Create vault containing bucket's tokens

// ─── Deposits ───
vault.deposit(bucket);                               // Add tokens from bucket into vault

// ─── Withdrawals ───
let bucket = vault.withdraw(amount);                 // Withdraw fungible amount → Bucket
let bucket = vault.withdraw(1u64);                   // Can pass u64 directly
let bucket = vault.withdraw_non_fungible(nft_id);    // Withdraw one NFT by ID → Bucket
let bucket = vault.withdraw_non_fungibles(id_set);   // Withdraw multiple NFTs → Bucket
let bucket = vault.withdraw_all();                   // Withdraw everything → Bucket

// ─── Queries ───
let balance: Amount = vault.balance();               // Current balance
let locked: Amount = vault.locked_balance();         // Locked/frozen balance
let addr: ResourceAddress = vault.resource_address();// Resource type held by this vault
let ids: BTreeSet<NonFungibleId> = vault.get_non_fungible_ids(); // All NFT IDs in vault

// ─── Resource Management ───
let manager: ResourceManager = vault.get_resource_manager(); // For minting etc.

// ─── Fee Payment ───
vault.pay_fee(amount);                               // Pay transaction fee from this vault

// ─── Authorization ───
vault.authorize();                                   // Create auth proof from vault contents (RAII)
let proof = vault.create_proof_by_amount(amount);    // Create proof for a specific amount

CRITICAL: A Vault MUST be stored in a component struct field before the function returns. An orphaned vault (created but not stored) will cause the transaction to fail.

Buckets — Complete API

// ─── Queries ───
let addr: ResourceAddress = bucket.resource_address(); // What resource this holds
let rtype: ResourceType = bucket.resource_type();       // Fungible, NonFungible, etc.
let amt: Amount = bucket.amount();                      // How many tokens
let empty: bool = bucket.is_empty();                    // Whether empty
let ids = bucket.get_non_fungible_ids();                // NFT IDs in bucket
let nfts = bucket.get_non_fungibles();                  // Full NFT data

// ─── Splitting ───
let new_bucket = bucket.take(Amount::from(50));         // Split off some tokens

// ─── Combining ───
let combined = bucket.join(other_bucket);               // Merge two same-resource buckets

// ─── Destruction ───
bucket.burn();                                          // Permanently destroy tokens
bucket.drop_empty();                                    // Assert empty and drop (panics if not)

// ─── Proofs ───
let proof = bucket.create_proof();                      // Create ownership proof

CRITICAL: A Bucket MUST be consumed before the function returns. Consume it by: depositing into a vault, burning, returning from a function, or passing to another component. An orphaned bucket will cause the transaction to fail.

Access Rules and Authorization

Component Access Rules

Component::new(Self { ... })
    .with_access_rules(
        ComponentAccessRules::new()      // Default: deny_all for unlisted methods
            .method("guess", rule!(allow_all))
            .method("admin_action", rule!(resource(admin_badge)))
            .default(rule!(deny_all))
    )
    // OR use the convenience constructor:
    // ComponentAccessRules::allow_all()  // Default: allow_all for unlisted methods
    .with_owner_rule(OwnerRule::OwnedBySigner)
    .create()

The rule! Macro — Complete Reference

// ─── Basic Rules ───
rule!(allow_all)                                        // No restrictions
rule!(deny_all)                                         // Nobody can call

// ─── Resource-Based Rules ───
rule!(resource(resource_address))                       // Must hold this resource in a proof
rule!(non_fungible(NonFungibleAddress::new(res, id)))   // Must hold specific NFT
rule!(public_key(ristretto_public_key_bytes))           // Must be signed by this key

// ─── Scope Rules ───
rule!(component(component_address))                     // Only callable from this component
rule!(template(template_address))                       // Only callable from this template

// ─── Composite Rules ───
rule!(any_of(resource(a), resource(b)))                 // Any one condition met (OR)
rule!(all_of(resource(a), resource(b)))                 // All conditions met (AND)
rule!(m_of_n(2, resource(a), resource(b), resource(c)))// M of N conditions met

Owner Rules

OwnerRule::OwnedBySigner           // Default: transaction signer is owner
OwnerRule::None                    // No owner (nobody can update access rules)
OwnerRule::ByAccessRule(rule)      // Custom rule determines ownership
OwnerRule::ByPublicKey(pk)         // Specific public key is owner

Caller Context — Complete API

// Get the authenticated signer (ALWAYS use this for identity, never accept as argument)
let signer: RistrettoPublicKeyBytes = CallerContext::transaction_signer_public_key();

// Get current component address (only in CallMethod context)
let addr: ComponentAddress = CallerContext::current_component_address();

// Get signer proof (for passing as authorization)
let proof: Proof = CallerContext::get_main_signer_proof();
let proof: Proof = CallerContext::get_signer_proof_for_public_key(pk);

// Address allocation (for creating components/resources with deterministic addresses)
let alloc: ComponentAddressAllocation = CallerContext::allocate_component_address(None);
let alloc: ComponentAddressAllocation = CallerContext::allocate_component_address(Some(pk));
let alloc: ResourceAddressAllocation = CallerContext::allocate_resource_address();

NEVER accept a public key as a method argument for authentication. Always use CallerContext::transaction_signer_public_key() — it cannot be spoofed.

Cross-Component Calls

// Get a reference to another component
let other: ComponentManager = ComponentManager::get(component_address);

// Call a method that returns a value
let value: u64 = other.call("method_name", args![arg1, arg2]);

// Call a method that returns unit (fire-and-forget)
other.invoke("method_name", args![arg1, arg2]);

// Common pattern: deposit a bucket into another component (e.g., Account)
other.invoke("deposit", args![prize_bucket]);

// Get template address of a component
let tmpl: TemplateAddress = other.get_template_address();

// Get the address
let addr: ComponentAddress = other.component_address();

Events

// Emit an event (permanently recorded in the transaction receipt)
emit_event("GameEnded", metadata![
    "winner_account" => winner_address.to_string(),
    "number" => winning_number.to_string(),
    "round" => round.to_string(),
]);

Events are indexed by the Indexer and can be queried by explorers and dApps. The topic is formatted as "TemplateName.EventTopic" in receipts.

Randomness

use tari_template_lib::rand::random_bytes;

// Get N pseudorandom bytes
let bytes: Vec<u8> = random_bytes(4);

// Convenience: get a random u32
use tari_template_lib::rand::random_u32;
let n: u32 = random_u32();

// Common pattern: random number in range
fn generate_number() -> u8 {
    random_bytes(1)[0] % 11  // 0..=10
}

WARNING: random_bytes is deterministic — entropy comes from the transaction itself to ensure all validators produce the same result. Do NOT use for cryptographic security. You cannot use the rand crate in templates (no entropy source on wasm32-unknown-unknown).

Publishing Templates

Publishing is the most expensive operation you will run — every validator stores the WASM permanently, so the fee is dominated by binary size. Fees are denominated in microtari (µT); 1 tTARI = 1,000,000 µT. At current testnet rates a publish costs a flat 250,000 µT, plus the first 96 KiB of the binary at the per-byte storage rate, plus a quadratic premium of 100 µT × units² where units is the number of whole kilobytes beyond 96 KiB. On top of that every validator Cranelift-compiles the binary, charged as execution points at 140,000 µT plus 2.1 µT per binary byte — for a 96 KiB template that roughly doubles the bill, to ~0.72 tTARI all-in. Binaries over 1 MiB are rejected outright. Shrink the binary first (see Minifying the WASM) and take the fee from a dry run (see Fee Estimation (Dry-Run)) or the Web UI's "Estimate Fee" button — never a hardcoded number. Do NOT add a publish subcommand to an example CLI; a client app may publish programmatically when the fee comes from a dry-run estimate.

Publish via Wallet Web UI (Easiest for a One-Off Publish)

1. Open the Tari Ootle Wallet web UI (default: http://127.0.0.1:5100)
2. Click "Publish Template" on the Home page
3. Select fee account with tTARI (testnet Tari)
4. Upload the .wasm file from target/wasm32-unknown-unknown/release/
5. Click "Estimate Fee" then "Publish Template"
6. Find the template address under "Templates" in the sidebar
7. Paste the template address into the CLI's state file or --template-address flag

Publish Programmatically (ootle-rs)

Programmatic publishing is fine for a client app when the fee comes from a dry-run estimate, never from a hardcoded value — see Fee Estimation (Dry-Run).

use tari_ootle_transaction::TransactionBuilder;
use ootle_rs::TransactionRequest;

let wasm_binary: Vec<u8> = std::fs::read("target/wasm32-unknown-unknown/release/your_template.wasm")?;
let build = |max_fee: u64| TransactionBuilder::new(provider.network())
    .with_auto_fill_inputs()
    .pay_fee_from_component(account_addr, max_fee)
    .publish_template(wasm_binary.clone().try_into().unwrap())
    .build_unsigned();

// Dry-run with a generous max_fee, then submit with the fee it reports.
let required = provider
    .sign_and_send_dry_run(build(50_000_000))
    .await?
    .finalize
    .required_fees();

let tx = TransactionRequest::default()
    .with_transaction(build(required))
    .build(provider.wallet())
    .await?;
let receipt = provider.send_transaction(tx).await?.watch().await?;

// Get the new template address from the receipt
let template_addr = receipt.diff_summary.upped
    .iter()
    .find_map(|s| s.substate_id.as_template())
    .expect("template address in receipt");

Fee guidance for publishing: the cost follows the binary size, not what the template does — see the rates above. Take the exact figure from a dry run: from ootle-rs, IndexerProvider::sign_and_send_dry_run returns an ExecuteResult whose finalize.required_fees() is the minimum a real submission may carry; over JSON-RPC, transactions.publish_template with dry_run: true returns dry_run_fee.

Fee Estimation (Dry-Run)

Never hardcode a fee in client code. Dry-run the transaction, then submit with what the dry run reports:

// 1. Build with a generous max_fee. A dry run is metered at whatever max_fee it
//    carries, so one that is too small aborts on fee exhaustion instead of
//    reporting the real cost.
let unsigned = TransactionBuilder::new(provider.network())
    .with_auto_fill_inputs()
    .pay_fee_from_component(account_addr, 10_000_000u64)
    .call_method(component_addr, "some_method", args![])
    .build_unsigned();

// 2. Dry-run it. The result is the ExecuteResult a real run would produce.
let result = provider.sign_and_send_dry_run(unsigned).await?;
let required = result.finalize.required_fees();

// 3. Rebuild with the reported fee and submit that.
let unsigned = TransactionBuilder::new(provider.network())
    .with_auto_fill_inputs()
    .pay_fee_from_component(account_addr, required)
    .call_method(component_addr, "some_method", args![])
    .build_unsigned();

The max_fee the dry run carries has to be generous but still within the paying account's balance — it is withdrawn from the fee vault for the run, so a dry run cannot be metered above what the account holds.

required_fees() is a floor, not an estimate to pad: it is what the dry run was charged plus a small allowance for the metering drift that carrying a different max_fee causes, so no margin multiplier is needed. Submitting above the floor is harmless when an account vault pays — the unspent remainder is returned to that vault — but a fee paid purely by a stealth reveal keeps no change, so pay close to the floor there.

The wallet daemon exposes the same estimate over JSON-RPC: transactions.publish_template with dry_run: true returns dry_run_fee, and transactions.submit_dry_run returns the full result. The Web UI's "Estimate Fee" button runs a dry run too.

The pattern applies to every transaction type: faucet claims, publishes, method calls and stealth spends.

Interacting with Deployed Components (Client-Side Rust)

Setup Wallet and Provider

use ootle_rs::{
    key_provider::PrivateKeyProvider,
    provider::ProviderBuilder,
    wallet::OotleWallet,
    default_indexer_url,
};
use tari_ootle_common_types::Network;

const NETWORK: Network = Network::Esmeralda;  // Testnet (default_indexer_url is configured)

// Create a random wallet (for testing) or load from seed
let secret = PrivateKeyProvider::random(NETWORK);
let wallet = OotleWallet::from(secret);

let mut provider = ProviderBuilder::new()
    .wallet(wallet)
    .connect(default_indexer_url(NETWORK))
    .await?;

// With custom transaction timeout (default is 32 seconds — too short for testnet):
use std::time::Duration;
let mut provider = ProviderBuilder::new()
    .wallet(wallet)
    .connect_with_transaction_timeout(default_indexer_url(NETWORK), Duration::from_secs(120))
    .await?;

Timeout guidance: The default transaction timeout is 32 seconds, which is often too short for the Esmeralda testnet. Use connect_with_transaction_timeout() with 120 seconds for testnet usage. LocalNet is faster and the default is usually fine.

Available networks:

• Network::Esmeralda — Public testnet (indexer: http://217.182.93.35:50124)
• Network::LocalNet — Local development (indexer: http://localhost:12500)
• Other networks (MainNet, StageNet, NextNet, Igor) are not yet configured with default indexer URLs

Fund Account (Testnet Faucet)

use ootle_rs::{
    TransactionRequest,
    builtin_templates::{UnsignedTransactionBuilder, faucet::IFaucet},
};
use tari_template_lib_types::constants::TARI;

let unsigned_tx = IFaucet::new(&provider)
    .take_faucet_funds(10 * TARI)    // Request 10 TARI
    .pay_fee(500u64)                     // Fee for the transaction
    .prepare()
    .await?;

let tx = TransactionRequest::default()
    .with_transaction(unsigned_tx)
    .build(provider.wallet())
    .await?;

let pending = provider.send_transaction(tx).await?;
let outcome = pending.watch().await?;

Transaction Pattern (Sign → Send → Watch)

Every on-chain interaction follows this pattern:

use tari_ootle_transaction::{TransactionBuilder, args};
use ootle_rs::TransactionRequest;

// 1. Build an unsigned transaction
let unsigned_tx = TransactionBuilder::new(provider.network())
    .with_auto_fill_inputs()                              // Auto-detect input substates
    .pay_fee_from_component(account_addr, 2000u64)        // Placeholder — take this from a dry run
    .call_function(template_addr, "new", args![])         // Or call_method(...)
    .build_unsigned();

// 2. Sign it
let tx = TransactionRequest::default()
    .with_transaction(unsigned_tx)
    .build(provider.wallet())
    .await?;

// 3. Send and wait for finalization
let pending = provider.send_transaction(tx).await?;
let receipt = pending.watch().await?;

The fees in these examples are placeholders. Real client code takes the fee from a dry run (see Fee Estimation (Dry-Run)); a hardcoded amount is either wasteful or fails the transaction.

Call a Template Function (Create Component)

let unsigned_tx = TransactionBuilder::new(provider.network())
    .with_auto_fill_inputs()
    .pay_fee_from_component(account_addr, 2000u64)
    .call_function(template_addr, "new", args![])
    .build_unsigned();

Call a Component Method

let unsigned_tx = TransactionBuilder::new(provider.network())
    .with_auto_fill_inputs()
    .pay_fee_from_component(account_addr, 2000u64)
    .call_method(component_addr, "start_game", args![nft_id])
    .build_unsigned();

TransactionBuilder — Key Methods

TransactionBuilder::new(network)
    // Input handling
    .with_auto_fill_inputs()                    // Auto-detect required substates
    .add_input(substate_address)                // Add specific input
    .with_inputs(iter_of_inputs)                // Add multiple inputs
    .with_unversioned_inputs(iter)              // Add unversioned inputs

    // Fee payment
    .pay_fee_from_component(account, amount)    // Pay fee from an account component
    .pay_fee_from_bucket(bucket_label, amount)  // Pay fee from a workspace bucket

    // Instructions
    .call_function(template, "fn_name", args![...])   // Call template function
    .call_method(component, "method", args![...])     // Call component method
    .create_account(public_key)                        // Create an account component
    .create_account_with_bucket(pk, bucket_label)      // Create account with initial funds
    .publish_template(wasm_binary)                     // Deploy a template

    // Workspace (chain instruction outputs)
    .put_last_instruction_output_on_workspace("label") // Store output for later use
    .take_from_bucket("label", amount)                 // Take from workspace bucket

    // Address allocation
    .allocate_component_address()                      // Pre-allocate component address
    .allocate_resource_address()                       // Pre-allocate resource address

    // Build
    .build_unsigned()                                  // Produce unsigned transaction

Read Addresses from Receipts

// Find the new component address
let component_addr = receipt.diff_summary.upped
    .iter()
    .find_map(|s| s.substate_id.as_component_address())
    .expect("component address in receipt");

// Find a resource address (excluding native TARI)
use tari_template_lib_types::constants::TARI_TOKEN;
let resource_addr = receipt.diff_summary.upped
    .iter()
    .find_map(|s| s.substate_id.as_resource_address().filter(|a| *a != TARI_TOKEN))
    .expect("resource address in receipt");

// Find a template address (returns PublishedTemplateAddress)
// IMPORTANT: Use as_template() on SubstateId — NOT as_template_address()
// as_template_address() does NOT exist on SubstateId
let template_addr = receipt.diff_summary.upped
    .iter()
    .find_map(|s| s.substate_id.as_template())
    .expect("template address in receipt");

IMPORTANT API note: On SubstateId, the method is as_template() — it returns Option<PublishedTemplateAddress>. There is NO as_template_address() method on SubstateId. If you need the underlying TemplateAddress (a Hash32), call .as_template_address() on the PublishedTemplateAddress result, not on the SubstateId.

Read Events from Receipts

let event = receipt.events
    .iter()
    .find(|e| e.topic() == "GuessingGame.GameEnded")
    .expect("event in receipt");

let value = event.get_payload("field_name");

Manual Inputs

When the transaction touches vaults/components that auto-fill can't detect, add them manually:

let unsigned_tx = TransactionBuilder::new(provider.network())
    .with_auto_fill_inputs()
    .add_input(specific_substate_address)
    .with_inputs(addresses.iter().copied().map(Into::into))
    .pay_fee_from_component(account_addr, 2000u64)
    .call_method(component_addr, "end_game", args![])
    .build_unsigned();

Testing Templates

Use tari_template_test_tooling as a dev-dependency. It compiles your template to WASM and runs transactions against it locally using the same execution engine as the network.

Test Setup

Add to your test crate's Cargo.toml:

[dev-dependencies]
tari_template_test_tooling = "0.40"

tari_template_test_tooling re-exports the transaction crate, so use tari_template_test_tooling::transaction::args (and the other transaction items) in tests rather than adding `t

…(truncated)