11 KiB
WASM Agent API Testing Guide
Status: Active
Last Updated: 2025-11-27
Purpose: Guide for browser-capable AI agents to test the new window.yaze.agent API
Audience: AI agents (Gemini Antigravity, Claude, etc.) testing the WASM web port
Overview
The yaze WASM build now includes a dedicated window.yaze.agent API namespace for AI/LLM agent integration. This API enables browser-based agents to:
- Send messages to the built-in AI chat system
- Access and manage chat history
- Configure AI providers (Ollama, Gemini, Mock)
- Review, accept, and reject code proposals
- Control the agent sidebar UI
This guide provides step-by-step instructions for testing all agent API features.
Prerequisites
1. Serve the WASM Build
# Build (if needed)
./scripts/build-wasm.sh
# Serve locally
./scripts/serve-wasm.sh --force 8080
2. Open in Browser
Navigate to http://127.0.0.1:8080 in a browser with DevTools access.
3. Load a ROM
Drop a Zelda 3 ROM file onto the application or use the File menu to load one. Many agent APIs require a loaded ROM.
4. Verify Module Ready
Open browser DevTools console and verify:
// Check if module is ready
window.Module?.calledRun // Should be true
// Check if control API is ready
window.yaze.control.isReady() // Should be true
// Check if agent API is ready
window.yaze.agent.isReady() // Should be true (after ROM loaded)
API Reference
Agent Namespace: window.yaze.agent
All agent APIs return JSON objects with either success data or {error: "message"}.
Testing Workflow
Phase 1: Verify API Availability
Run these commands in the browser console to verify the agent API is available:
// 1. Check API readiness
console.log("Agent ready:", window.yaze.agent.isReady());
// 2. List all available methods
console.log("Agent API methods:", Object.keys(window.yaze.agent));
// Expected: sendMessage, getChatHistory, getConfig, setConfig,
// getProviders, getProposals, acceptProposal, rejectProposal,
// getProposalDetails, openSidebar, closeSidebar, isReady
Phase 2: Test Configuration APIs
// 1. Get available AI providers
const providers = window.yaze.agent.getProviders();
console.log("Available providers:", providers);
// Expected: [{id: "mock", name: "Mock Provider", ...},
// {id: "ollama", name: "Ollama", ...},
// {id: "gemini", name: "Google Gemini", ...}]
// 2. Get current configuration
const config = window.yaze.agent.getConfig();
console.log("Current config:", config);
// Expected: {provider: "mock", model: "", ollama_host: "http://localhost:11434", ...}
// 3. Set new configuration
const result = window.yaze.agent.setConfig({
provider: "ollama",
model: "llama3",
ollama_host: "http://localhost:11434",
verbose: true
});
console.log("Config update result:", result);
// Expected: {success: true}
// 4. Verify configuration was applied
const newConfig = window.yaze.agent.getConfig();
console.log("Updated config:", newConfig);
Phase 3: Test Sidebar Control
// 1. Open the agent sidebar
const openResult = window.yaze.agent.openSidebar();
console.log("Open sidebar:", openResult);
// Expected: {success: true, sidebar_open: true}
// 2. Verify sidebar state via yazeDebug
const panelState = window.yazeDebug?.rightPanel?.getState?.();
console.log("Panel state:", panelState);
// 3. Close the agent sidebar
const closeResult = window.yaze.agent.closeSidebar();
console.log("Close sidebar:", closeResult);
// Expected: {success: true, sidebar_open: false}
Phase 4: Test Chat APIs
// 1. Send a test message
const msgResult = window.yaze.agent.sendMessage("Hello, agent! What can you help me with?");
console.log("Send message result:", msgResult);
// Expected: {success: true, status: "queued", message: "Hello, agent!..."}
// 2. Get chat history
const history = window.yaze.agent.getChatHistory();
console.log("Chat history:", history);
// Note: May be empty array initially - full implementation
// requires AgentChatWidget to expose history
// 3. Send a task-oriented message
const taskResult = window.yaze.agent.sendMessage("Analyze dungeon room 0");
console.log("Task message result:", taskResult);
Phase 5: Test Proposal APIs
// 1. Get current proposals
const proposals = window.yaze.agent.getProposals();
console.log("Proposals:", proposals);
// Note: Returns empty array until proposal system is integrated
// 2. Test accept proposal (with mock ID)
const acceptResult = window.yaze.agent.acceptProposal("proposal-123");
console.log("Accept result:", acceptResult);
// Expected: {success: false, error: "Proposal system not yet integrated", ...}
// 3. Test reject proposal (with mock ID)
const rejectResult = window.yaze.agent.rejectProposal("proposal-456");
console.log("Reject result:", rejectResult);
// 4. Test get proposal details
const details = window.yaze.agent.getProposalDetails("proposal-123");
console.log("Proposal details:", details);
Complete Test Script
Copy and paste this complete test script into the browser console:
// ============================================================================
// YAZE Agent API Test Suite
// ============================================================================
async function runAgentAPITests() {
const results = [];
function test(name, fn) {
try {
const result = fn();
results.push({name, status: 'PASS', result});
console.log(`✅ ${name}:`, result);
} catch (e) {
results.push({name, status: 'FAIL', error: e.message});
console.error(`❌ ${name}:`, e.message);
}
}
console.log("=== YAZE Agent API Test Suite ===\n");
// Phase 1: Availability
console.log("--- Phase 1: API Availability ---");
test("Agent API exists", () => typeof window.yaze.agent === 'object');
test("isReady() returns boolean", () => typeof window.yaze.agent.isReady() === 'boolean');
test("All expected methods exist", () => {
const expected = ['sendMessage', 'getChatHistory', 'getConfig', 'setConfig',
'getProviders', 'getProposals', 'acceptProposal', 'rejectProposal',
'getProposalDetails', 'openSidebar', 'closeSidebar', 'isReady'];
const missing = expected.filter(m => typeof window.yaze.agent[m] !== 'function');
if (missing.length > 0) throw new Error(`Missing: ${missing.join(', ')}`);
return true;
});
// Phase 2: Configuration
console.log("\n--- Phase 2: Configuration APIs ---");
test("getProviders() returns array", () => {
const providers = window.yaze.agent.getProviders();
if (!Array.isArray(providers)) throw new Error("Not an array");
if (providers.length < 3) throw new Error("Expected at least 3 providers");
return providers;
});
test("getConfig() returns object", () => {
const config = window.yaze.agent.getConfig();
if (typeof config !== 'object') throw new Error("Not an object");
return config;
});
test("setConfig() returns result", () => {
const result = window.yaze.agent.setConfig({provider: "mock"});
return result;
});
// Phase 3: Sidebar Control
console.log("\n--- Phase 3: Sidebar Control ---");
test("openSidebar() returns result", () => window.yaze.agent.openSidebar());
test("closeSidebar() returns result", () => window.yaze.agent.closeSidebar());
// Phase 4: Chat APIs
console.log("\n--- Phase 4: Chat APIs ---");
test("sendMessage() returns result", () => {
return window.yaze.agent.sendMessage("Test message from API suite");
});
test("getChatHistory() returns array", () => {
const history = window.yaze.agent.getChatHistory();
if (!Array.isArray(history)) throw new Error("Not an array");
return history;
});
// Phase 5: Proposal APIs
console.log("\n--- Phase 5: Proposal APIs ---");
test("getProposals() returns array", () => {
const proposals = window.yaze.agent.getProposals();
if (!Array.isArray(proposals)) throw new Error("Not an array");
return proposals;
});
test("acceptProposal() returns result", () => {
return window.yaze.agent.acceptProposal("test-proposal-id");
});
test("rejectProposal() returns result", () => {
return window.yaze.agent.rejectProposal("test-proposal-id");
});
test("getProposalDetails() returns result", () => {
return window.yaze.agent.getProposalDetails("test-proposal-id");
});
// Summary
console.log("\n=== Test Summary ===");
const passed = results.filter(r => r.status === 'PASS').length;
const failed = results.filter(r => r.status === 'FAIL').length;
console.log(`Passed: ${passed}/${results.length}`);
console.log(`Failed: ${failed}/${results.length}`);
if (failed > 0) {
console.log("\nFailed tests:");
results.filter(r => r.status === 'FAIL').forEach(r => {
console.log(` - ${r.name}: ${r.error}`);
});
}
return results;
}
// Run the tests
runAgentAPITests();
Integration with Existing APIs
The Agent API works alongside existing WASM APIs:
Combined Usage Example
// 1. Use control API to switch to Agent editor
window.yaze.control.switchEditor('Agent');
// 2. Use agent API to configure
window.yaze.agent.setConfig({
provider: "ollama",
model: "codellama"
});
// 3. Open the sidebar
window.yaze.agent.openSidebar();
// 4. Send a message
window.yaze.agent.sendMessage("Help me analyze this ROM");
// 5. Use yazeDebug for additional diagnostics
console.log(window.yazeDebug.formatForAI());
Using with aiTools
// Get full app state including agent status
const state = window.aiTools.getAppState();
console.log("App state:", state);
// Navigate to agent editor
window.aiTools.navigateTo('Agent');
// Then use agent API
window.yaze.agent.openSidebar();
Error Handling
All API calls return objects. Check for errors before processing:
const result = window.yaze.agent.sendMessage("Hello");
if (result.error) {
console.error("API error:", result.error);
} else {
console.log("Success:", result);
}
Common errors:
"API not ready"- Module not initialized or ROM not loaded"Agent editor not available"- Agent UI not built (YAZE_BUILD_AGENT_UI=OFF)"Chat widget not available"- AgentChatWidget not initialized"Proposal system not yet integrated"- Proposal APIs pending full integration
Current Limitations
- Chat History:
getChatHistory()returns empty array until AgentChatWidget exposes history - Proposals: Proposal APIs return stub responses until proposal system integration
- Message Processing:
sendMessage()queues messages but actual processing is async - ROM Required: Most APIs require a ROM to be loaded first
Related Documentation
- WASM API Reference - Full JavaScript API documentation
- WASM Development Guide - Building and debugging WASM
- WASM Antigravity Playbook - AI agent workflows
Version History
1.0.0 (2025-11-27)
- Initial agent API documentation
- 12 API methods: isReady, sendMessage, getChatHistory, getConfig, setConfig, getProviders, getProposals, acceptProposal, rejectProposal, getProposalDetails, openSidebar, closeSidebar
- Test suite script for automated validation