@salimassili/ai-costguard 2.1.2 → 2.1.4

package/examples/integrations/slack-webhook-block-demo.mjs

@@ -0,0 +1,66 @@
+import { guard, GuardError, registerTokenizer } from '@salimassili/ai-costguard';
+
+const slackWebhook = process.env.COSTGUARD_SLACK_WEBHOOK;
+let providerCalls = 0;
+
+// Keeps this demo output quiet and deterministic. Production apps can keep the
+// default estimator or register a provider-specific tokenizer.
+registerTokenizer('gpt-4o-mini', (text) => Math.ceil(text.length / 4));
+
+const mockedOpenAI = {
+  chat: {
+    completions: {
+      create: async () => {
+        providerCalls += 1;
+        return {
+          id: 'mocked-provider-response',
+          choices: [{ message: { content: 'This should never execute in the demo.' } }],
+        };
+      },
+    },
+  },
+};
+
+const openai = guard(mockedOpenAI, {
+  budget: 0.0001,
+  scope: { projectId: 'sales-demo', sessionId: 'slack-webhook-block-demo' },
+  webhooks: {
+    slack: slackWebhook,
+    retries: 0,
+    timeoutMs: 1000,
+  },
+});
+
+console.log('AI CostGuard Slack webhook block demo');
+console.log('Scenario: an agent tries to start an expensive model call with almost no budget left.');
+console.log(`Slack alert: ${slackWebhook ? 'enabled from COSTGUARD_SLACK_WEBHOOK' : 'off by default'}`);
+
+try {
+  await openai.chat.completions.create({
+    model: 'gpt-4o-mini',
+    messages: [
+      {
+        role: 'user',
+        content:
+          'Run a large analysis job, summarize every retry, and reserve enough output for a long report.',
+      },
+    ],
+    max_tokens: 800,
+  });
+
+  throw new Error('Expected AI CostGuard to block before the mocked provider ran.');
+} catch (error) {
+  if (!(error instanceof GuardError)) {
+    throw error;
+  }
+
+  console.log('Result: BLOCKED before provider execution');
+  console.log(`Reason: ${error.code}`);
+  console.log(`Estimated avoided spend: $${error.context.estimatedCost.toFixed(6)}`);
+  console.log(`Provider calls executed: ${providerCalls}`);
+  console.log('Webhook URL printed: no');
+}
+
+if (providerCalls !== 0) {
+  throw new Error(`Provider executed ${providerCalls} time(s); expected 0.`);
+}

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@salimassili/ai-costguard",
-  "version": "2.1.2",
+  "version": "2.1.4",
   "description": "Local-first runtime safety layer for AI agents that blocks runaway costs, loops, retries, and budget overruns before API calls execute.",
   "type": "module",
   "main": "./dist/index.js",