@outputai/cli 0.12.0 → 0.12.1-next.69255d7.0

package/dist/assets/docker/docker-compose-dev.yml

@@ -80,7 +80,7 @@ services:
         condition: service_healthy
       worker:
         condition: service_healthy
-    image: outputai/api:${OUTPUT_API_VERSION:-0.12.0}
+    image: outputai/api:${OUTPUT_API_VERSION:-0.12.1-next.69255d7.0}
     init: true
     networks:
       - main

package/dist/generated/framework_version.json

@@ -1,3 +1,3 @@
 {
-    "framework": "0.12.0"
+    "framework": "0.12.1-next.69255d7.0"
 }

package/dist/services/cost_calculator.js

@@ -84,7 +84,8 @@ function parseLegacyLLMUsageEvent(node, stepName, event) {
         model: event.modelId || 'unknown',
         usage: eventTokenUsage(event.usage ?? []),
         originalCost: event.total,
-        lines: event.usage ?? []
+        lines: event.usage ?? [],
+        incomplete: false
     };
 }
 function parseNormalizedLLMUsage(node, stepName, event) {
@@ -100,7 +101,11 @@ function parseNormalizedLLMUsage(node, stepName, event) {
         model: event.modelId || 'unknown',
         usage: normalizedItemTokenUsage(event.items),
         originalCost: 0,
-        lines
+        lines,
+        // This path runs whenever cost computation produced no llm:generation:cost
+        // attribute at all (e.g. pricing config missing, not just unrated grounding),
+        // so every line here is unpriced by construction, not only grounding calls.
+        incomplete: true
     };
 }
 function parseLLMCostEvent(node, stepName, event) {
@@ -116,7 +121,10 @@ function parseLLMCostEvent(node, stepName, event) {
         model: event.modelId || 'unknown',
         usage: normalizedItemTokenUsage(event.items),
         originalCost: event.total ?? 0,
-        lines
+        lines,
+        // An INCOMPLETE event carries a real but unpriced charge (e.g. unrated
+        // grounding). Flag it so the report can mark the understated total.
+        incomplete: event.status === 'incomplete'
     };
 }
 export function extractValue(obj, path) {
@@ -461,7 +469,8 @@ function aggregateLLMCosts(llmCalls, config) {
             cached: call.usage.cachedInputTokens ?? 0,
             reasoning: call.usage.reasoningTokens ?? 0,
             originalCost,
-            adjustedCost
+            adjustedCost,
+            incomplete: call.incomplete
         });
         totals.inputTokens += call.usage.inputTokens ?? 0;
         totals.outputTokens += call.usage.outputTokens ?? 0;

package/dist/services/cost_calculator.spec.js

@@ -37,10 +37,11 @@ function llmEventNode(id, model, lines) {
         }
     };
 }
-function llmCostNode(id, model, items) {
+function llmCostNode(id, model, items, status = 'precise') {
     const totalOf = (group) => items.filter(item => item.group === group).reduce((sum, item) => sum + (item.total ?? 0), 0);
     const input = totalOf('input');
     const output = totalOf('output');
+    const request = totalOf('request');
     return {
         id,
         kind: 'llm',
@@ -52,8 +53,9 @@ function llmCostNode(id, model, items) {
                 modelId: model,
                 input,
                 output,
-                total: input + output,
-                status: 'precise',
+                request,
+                total: input + output + request,
+                status,
                 items
             }
         }
@@ -280,6 +282,66 @@ describe('findLLMCalls', () => {
         ]);
         expect(calls[0].originalCost).toBeCloseTo(0.001965, 8);
     });
+    it('reads a priced grounding request item without counting it as tokens', () => {
+        const items = [
+            { group: 'input', label: 'no_cache', amount: 1032, ppm: 0.3, total: 0.0003096, status: 'ok' },
+            { group: 'output', label: 'text', amount: 793, ppm: 2.5, total: 0.0019825, status: 'ok' },
+            { group: 'request', label: 'grounding_prompt', amount: 1, ppm: 35_000, total: 0.035, status: 'ok' }
+        ];
+        const calls = findLLMCalls({
+            kind: 'workflow',
+            children: [llmCostNode('grounded', 'gemini-2.5-flash', items)]
+        });
+        expect(calls[0].incomplete).toBe(false);
+        expect(calls[0].originalCost).toBeCloseTo(0.0372921, 8);
+        // Grounding bills per request, so it must not inflate token usage.
+        expect(calls[0].usage).toEqual({
+            inputTokens: 1032,
+            cachedInputTokens: 0,
+            outputTokens: 793,
+            reasoningTokens: 0
+        });
+        expect(calls[0].lines.find(l => l.type === 'request_grounding_prompt')).toEqual({
+            type: 'request_grounding_prompt',
+            ppm: 35_000,
+            amount: 1,
+            total: 0.035
+        });
+    });
+    it('preserves a priced grounding charge as-charged when re-pricing at costs.yml rates', () => {
+        const items = [
+            { group: 'input', label: 'no_cache', amount: 1032, ppm: 0.3, total: 0.0003096, status: 'ok' },
+            { group: 'output', label: 'text', amount: 793, ppm: 2.5, total: 0.0019825, status: 'ok' },
+            { group: 'request', label: 'grounding_prompt', amount: 1, ppm: 35_000, total: 0.035, status: 'ok' }
+        ];
+        const config = {
+            models: { 'gemini-2.5-flash': { provider: 'google-vertex', input: 0.3, output: 2.5 } },
+            services: {}
+        };
+        const report = calculateCost({ kind: 'workflow', name: 'w', children: [llmCostNode('grounded', 'gemini-2.5-flash', items)] }, config);
+        // grounding has no costs.yml line rate, so it degrades to its as-charged total, not $0.
+        expect(report.llmAdjustedCost).toBeCloseTo(0.0372921, 8);
+    });
+    it('flags an unrated grounding charge as incomplete instead of a silent $0', () => {
+        const items = [
+            { group: 'input', label: 'no_cache', amount: 1000, ppm: 1, total: 0.001, status: 'ok' },
+            { group: 'output', label: null, amount: 500, ppm: 5, total: 0.0025, status: 'ok' },
+            { group: 'request', label: 'grounding', amount: 3, ppm: null, total: null, status: 'missing' }
+        ];
+        const calls = findLLMCalls({
+            kind: 'workflow',
+            children: [llmCostNode('unrated', 'gemini-9-ultra', items, 'incomplete')]
+        });
+        expect(calls[0].incomplete).toBe(true);
+        // The unrated line still shows up (as $0), but the call carries the signal.
+        expect(calls[0].lines.find(l => l.type === 'request_grounding')).toEqual({
+            type: 'request_grounding',
+            ppm: 0,
+            amount: 3,
+            total: 0
+        });
+        expect(calls[0].usage).toMatchObject({ inputTokens: 1000, outputTokens: 500 });
+    });
     it('reads normalized usage without interpreting token totals as cost', () => {
         const items = [
             { group: 'input', label: null, amount: 100 },
@@ -303,7 +365,9 @@ describe('findLLMCalls', () => {
                 outputTokens: 110,
                 reasoningTokens: 70
             },
-            originalCost: 0
+            originalCost: 0,
+            // Usage-only means cost computation produced no priced result at all.
+            incomplete: true
         });
         expect(calls[0].lines).toEqual([
             { type: 'input', ppm: 0, amount: 100, total: 0 },

package/dist/types/cost.d.ts

@@ -51,6 +51,7 @@ export interface LLMCall {
     usage: TokenUsage;
     originalCost: number;
     lines: LLMUsageLine[];
+    incomplete: boolean;
 }
 export interface HTTPCall {
     stepName: string;
@@ -109,6 +110,7 @@ export interface LLMCostResult {
     reasoning: number;
     originalCost: number;
     adjustedCost: number;
+    incomplete: boolean;
 }
 export interface ServiceCostResult {
     step: string;
@@ -155,6 +157,7 @@ export interface LLMModelSummary {
     count: number;
     originalCost: number;
     adjustedCost: number;
+    incomplete: boolean;
 }
 export interface HostSummary {
     host: string;

package/dist/utils/cost_formatter.js

@@ -25,15 +25,23 @@ function formatCurrency(amount) {
 function pluralize(count, singular) {
     return count === 1 ? `1 ${singular}` : `${count} ${singular}s`;
 }
+// Marks a figure whose call had a real but unpriced charge (e.g. unrated
+// grounding), so a $0.00 line reads as "unpriced" rather than a clean zero.
+const INCOMPLETE_MARKER = '*';
+const INCOMPLETE_FOOTNOTE = '* cost incomplete (unpriced charge such as grounding, or missing usage); the total understates the bill';
+function markIncomplete(text, incomplete) {
+    return incomplete ? `${text} ${INCOMPLETE_MARKER}` : text;
+}
 export function parseCostData(report) {
     const byModel = {};
     for (const r of report.llmCalls) {
         if (!byModel[r.model]) {
-            byModel[r.model] = { count: 0, originalCost: 0, adjustedCost: 0 };
+            byModel[r.model] = { count: 0, originalCost: 0, adjustedCost: 0, incomplete: false };
         }
         byModel[r.model].count++;
         byModel[r.model].originalCost += r.originalCost;
         byModel[r.model].adjustedCost += r.adjustedCost;
+        byModel[r.model].incomplete ||= r.incomplete;
     }
     const llmModels = Object.entries(byModel)
         .sort((a, b) => b[1].adjustedCost - a[1].adjustedCost)
@@ -82,12 +90,13 @@ function formatSummary(data) {
             style: { head: ['cyan'] },
             colAligns: ['left', 'right', 'right', 'right']
         });
+        const anyIncomplete = data.llmModels.some(m => m.incomplete);
         for (const m of data.llmModels) {
             table.push([
                 m.model,
                 pluralize(m.count, 'call'),
-                formatCurrency(m.originalCost),
-                formatCurrency(m.adjustedCost)
+                markIncomplete(formatCurrency(m.originalCost), m.incomplete),
+                markIncomplete(formatCurrency(m.adjustedCost), m.incomplete)
             ]);
         }
         table.push([
@@ -98,6 +107,9 @@ function formatSummary(data) {
         ]);
         lines.push('LLM Costs:');
         lines.push(table.toString());
+        if (anyIncomplete) {
+            lines.push(INCOMPLETE_FOOTNOTE);
+        }
         lines.push('');
     }
     if (data.hosts.length > 0) {
@@ -146,6 +158,7 @@ function formatVerbose(data) {
             style: { head: ['cyan'] },
             colAligns
         });
+        const anyIncomplete = data.llmCalls.some(r => r.incomplete);
         for (const r of data.llmCalls) {
             const row = [
                 r.step,
@@ -159,7 +172,7 @@ function formatVerbose(data) {
             if (data.verbose.hasReasoning) {
                 row.push(formatNumber(r.reasoning));
             }
-            row.push(formatCurrency(r.originalCost), formatCurrency(r.adjustedCost));
+            row.push(markIncomplete(formatCurrency(r.originalCost), r.incomplete), markIncomplete(formatCurrency(r.adjustedCost), r.incomplete));
             table.push(row);
         }
         const totalRow = [
@@ -178,6 +191,9 @@ function formatVerbose(data) {
         table.push(totalRow);
         lines.push('LLM Calls:');
         lines.push(table.toString());
+        if (anyIncomplete) {
+            lines.push(INCOMPLETE_FOOTNOTE);
+        }
         lines.push('');
     }
     if (data.httpDetails.length > 0) {

package/dist/utils/cost_formatter.spec.d.ts

@@ -0,0 +1 @@
+export {};

package/dist/utils/cost_formatter.spec.js

@@ -0,0 +1,60 @@
+import { describe, it, expect } from 'vitest';
+import { formatCostReport } from '#utils/cost_formatter.js';
+function llmResult(overrides = {}) {
+    return {
+        step: 'gen',
+        model: 'gemini-2.5-flash',
+        input: 1032,
+        output: 793,
+        cached: 0,
+        reasoning: 0,
+        originalCost: 0.0037,
+        adjustedCost: 0.0037,
+        incomplete: false,
+        ...overrides
+    };
+}
+function report(llmCalls) {
+    const originalCost = llmCalls.reduce((s, c) => s + c.originalCost, 0);
+    const adjustedCost = llmCalls.reduce((s, c) => s + c.adjustedCost, 0);
+    return {
+        traceFile: 'trace.json',
+        workflowName: 'w',
+        durationMs: 1000,
+        llmCalls,
+        llmOriginalCost: originalCost,
+        llmAdjustedCost: adjustedCost,
+        totalInputTokens: 1032,
+        totalOutputTokens: 793,
+        totalCachedTokens: 0,
+        totalReasoningTokens: 0,
+        httpCosts: [],
+        httpOriginalCost: 0,
+        httpAdjustedCost: 0,
+        originalTotalCost: originalCost,
+        totalCost: adjustedCost
+    };
+}
+describe('formatCostReport incomplete marker', () => {
+    it('marks the model row with an unpriced charge and prints the footnote', () => {
+        const out = formatCostReport(report([llmResult({ model: 'gemini-2.5-flash', incomplete: true })]));
+        const modelRow = out.split('\n').find(line => line.includes('gemini-2.5-flash'));
+        expect(modelRow).toContain('$0.0037 *');
+        expect(out).toContain('cost incomplete');
+    });
+    it('omits the marker and footnote when every call is fully priced', () => {
+        const out = formatCostReport(report([llmResult()]));
+        const modelRow = out.split('\n').find(line => line.includes('gemini-2.5-flash'));
+        expect(modelRow).not.toContain('*');
+        expect(out).not.toContain('cost incomplete');
+    });
+    it('marks the flagged call row in the verbose table', () => {
+        const out = formatCostReport(report([llmResult({ step: 'grounded', incomplete: true }), llmResult({ step: 'plain' })]), { verbose: true });
+        const lines = out.split('\n');
+        const groundedRow = lines.find(line => line.includes('grounded'));
+        const plainRow = lines.find(line => line.includes('plain'));
+        expect(groundedRow).toContain('$0.0037 *');
+        expect(plainRow).not.toContain('*');
+        expect(out).toContain('cost incomplete');
+    });
+});

package/oclif.manifest.json

@@ -1752,5 +1752,5 @@
       ]
     }
   },
-  "version": "0.12.0"
+  "version": "0.12.1-next.69255d7.0"
 }

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@outputai/cli",
-  "version": "0.12.0",
+  "version": "0.12.1-next.69255d7.0",
   "description": "CLI for Output.ai workflow generation",
   "type": "module",
   "main": "dist/index.js",
@@ -36,9 +36,9 @@
     "react": "19.2.5",
     "semver": "7.7.4",
     "undici": "8.9.0",
-    "@outputai/credentials": "0.12.0",
-    "@outputai/llm": "0.12.0",
-    "@outputai/evals": "0.12.0"
+    "@outputai/credentials": "0.12.1-next.69255d7.0",
+    "@outputai/evals": "0.12.1-next.69255d7.0",
+    "@outputai/llm": "0.12.1-next.69255d7.0"
   },
   "devDependencies": {
     "@types/cli-progress": "3.11.6",