sentinelayer-cli 0.12.2 → 0.12.4

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "sentinelayer-cli",
-  "version": "0.12.2",
+  "version": "0.12.4",
   "description": "Scaffold Sentinelayer spec/prompt/guide artifacts with secure browser auth and token bootstrap.",
   "type": "module",
   "scripts": {

package/src/billing/ledger-entry.js

@@ -7,7 +7,10 @@ export const PRICED_ACTIONS = Object.freeze([
   "audit_run",
   "audit_security",
   "audit_frontend",
+  "chat_ask",
   "omargate_deep",
+  "scan_precheck",
+  "spec_generate_ai",
 ]);
 
 const OMIT_METADATA_KEYS = new Set([

package/src/billing/llm-session-usage.js

@@ -0,0 +1,91 @@
+import { buildBillingRunId, buildCallIdempotencyKey, stableHash } from "./ledger-entry.js";
+import { recordSessionUsage } from "./session-usage.js";
+
+function normalizeString(value) {
+  return String(value || "").trim();
+}
+
+function nonNegativeInteger(value) {
+  const parsed = Number(value || 0);
+  if (!Number.isFinite(parsed) || parsed < 0) return 0;
+  return Math.floor(parsed);
+}
+
+export function usageNumber(value, fallback = 0) {
+  const parsed = Number(value);
+  if (Number.isFinite(parsed) && parsed >= 0) return parsed;
+  const parsedFallback = Number(fallback || 0);
+  return Number.isFinite(parsedFallback) && parsedFallback >= 0 ? parsedFallback : 0;
+}
+
+export async function recordCliLlmSessionUsage({
+  sessionId,
+  agentId,
+  action,
+  model,
+  inputTokens = 0,
+  outputTokens = 0,
+  startedAtIso = "",
+  targetPath,
+  billingTier = "internal",
+  sourceCommand = "",
+  provider = "",
+  metadata = {},
+} = {}) {
+  const normalizedSessionId = normalizeString(sessionId);
+  const normalizedAgentId = normalizeString(agentId);
+  const normalizedAction = normalizeString(action);
+  const normalizedModel = normalizeString(model);
+  const createdAt = normalizeString(startedAtIso) || new Date().toISOString();
+  const safeInputTokens = nonNegativeInteger(inputTokens);
+  const safeOutputTokens = nonNegativeInteger(outputTokens);
+
+  if (!normalizedSessionId || !normalizedAgentId || !normalizedAction || !normalizedModel) {
+    return { ok: false, reason: "missing_session_usage_fields" };
+  }
+  if (safeInputTokens + safeOutputTokens <= 0) {
+    return { ok: false, reason: "zero_tokens" };
+  }
+
+  try {
+    const configHash = stableHash(
+      JSON.stringify({
+        action: normalizedAction,
+        agentId: normalizedAgentId,
+        model: normalizedModel,
+        provider: normalizeString(provider),
+        sourceCommand: normalizeString(sourceCommand),
+        metadata,
+      }),
+    );
+    const billingRunId = buildBillingRunId({
+      sessionId: normalizedSessionId,
+      invocationTimestamp: createdAt,
+      configHash,
+    });
+    return await recordSessionUsage(
+      normalizedSessionId,
+      {
+        agentId: normalizedAgentId,
+        action: normalizedAction,
+        model: normalizedModel,
+        inputTokens: safeInputTokens,
+        outputTokens: safeOutputTokens,
+        idempotencyKey: buildCallIdempotencyKey({ runId: billingRunId, callIndex: 0 }),
+        billingTier,
+        createdAt,
+        metadata: {
+          sourceCommand,
+          provider,
+          ...metadata,
+        },
+      },
+      { targetPath },
+    );
+  } catch (error) {
+    return {
+      ok: false,
+      reason: error instanceof Error ? error.message : String(error || "session_usage_failed"),
+    };
+  }
+}

package/src/commands/chat.js

@@ -9,7 +9,9 @@ import {
   resolveModel,
   resolveProvider,
 } from "../ai/client.js";
+import { recordCliLlmSessionUsage, usageNumber } from "../billing/llm-session-usage.js";
 import { resolveOutputRoot } from "../config/service.js";
+import { estimateModelCost } from "../cost/tracker.js";
 import { estimateTokens } from "../cost/tokenizer.js";
 
 function shouldEmitJson(options, command) {
@@ -92,7 +94,9 @@ export function registerChatCommand(program) {
       });
 
       const startedAt = Date.now();
+      const startedAtIso = new Date(startedAt).toISOString();
       let responseText = "";
+      let invocation = null;
 
       if (options.dryRun) {
         responseText = `DRY_RUN_RESPONSE: ${prompt.slice(0, 240)}`;
@@ -100,7 +104,7 @@ export function registerChatCommand(program) {
         const streamEnabled = Boolean(options.stream);
         let streamedText = "";
         const client = createMultiProviderApiClient();
-        const invocation = await client.invoke({
+        invocation = await client.invoke({
           provider,
           model,
           prompt,
@@ -125,8 +129,31 @@ export function registerChatCommand(program) {
 
       const durationMs = Date.now() - startedAt;
       const generatedAt = new Date().toISOString();
-      const inputTokens = estimateTokens(prompt, { model });
-      const outputTokens = estimateTokens(responseText, { model });
+      const estimatedInputTokens = estimateTokens(prompt, { model });
+      const estimatedOutputTokens = estimateTokens(responseText, { model });
+      const inputTokens = usageNumber(invocation?.usage?.inputTokens, estimatedInputTokens);
+      const outputTokens = usageNumber(invocation?.usage?.outputTokens, estimatedOutputTokens);
+      const modelCost = estimateModelCost({ modelId: invocation?.model || model, inputTokens, outputTokens });
+      const costUsd = usageNumber(invocation?.usage?.costUsd, modelCost.costUsd);
+
+      const sessionUsageLedger = options.dryRun
+        ? { ok: false, reason: "dry_run" }
+        : await recordCliLlmSessionUsage({
+            sessionId,
+            agentId: "chat-cli",
+            action: "chat_ask",
+            model: invocation?.model || model,
+            inputTokens,
+            outputTokens,
+            startedAtIso,
+            targetPath,
+            sourceCommand: "chat ask",
+            provider: invocation?.provider || provider,
+            metadata: {
+              streamed: Boolean(options.stream),
+              pricingFound: modelCost.pricingFound,
+            },
+          });
 
       await appendTranscriptEntries({
         transcriptPath,
@@ -154,8 +181,8 @@ export function registerChatCommand(program) {
       const payload = {
         command: "chat ask",
         sessionId,
-        provider,
-        model,
+        provider: invocation?.provider || provider,
+        model: invocation?.model || model,
         dryRun: Boolean(options.dryRun),
         streamed: Boolean(options.stream),
         transcriptPath,
@@ -165,8 +192,10 @@ export function registerChatCommand(program) {
           inputTokens,
           outputTokens,
           totalTokens: inputTokens + outputTokens,
+          costUsd,
           durationMs,
         },
+        billing: sessionUsageLedger,
       };
 
       if (emitJson) {
@@ -179,6 +208,6 @@ export function registerChatCommand(program) {
       }
       console.log(pc.gray(`session: ${sessionId}`));
       console.log(pc.gray(`transcript: ${transcriptPath}`));
-      console.log(pc.gray(`usage: input=${inputTokens} output=${outputTokens} duration_ms=${durationMs}`));
+      console.log(pc.gray(`usage: input=${inputTokens} output=${outputTokens} cost_usd=${costUsd.toFixed(6)} duration_ms=${durationMs}`));
     });
 }

package/src/commands/scan.js

@@ -11,6 +11,7 @@ import {
   resolveModel,
   resolveProvider,
 } from "../ai/client.js";
+import { recordCliLlmSessionUsage, usageNumber } from "../billing/llm-session-usage.js";
 import { loadConfig, resolveOutputRoot } from "../config/service.js";
 import { evaluateBudget } from "../cost/budget.js";
 import { appendCostEntry, summarizeCostHistory } from "../cost/history.js";
@@ -585,6 +586,7 @@ export function registerScanCommand(program) {
         profile,
       });
 
+      const startedAtIso = new Date().toISOString();
       const startedAtMs = Date.now();
       const client = createMultiProviderApiClient();
       const response = await client.invoke({
@@ -615,13 +617,16 @@ export function registerScanCommand(program) {
       await fsp.mkdir(path.dirname(reportPath), { recursive: true });
       await fsp.writeFile(reportPath, reportMarkdown, "utf-8");
 
-      const inputTokens = estimateTokens(prompt, { model: response.model });
-      const outputTokens = estimateTokens(aiMarkdown, { model: response.model });
+      const estimatedInputTokens = estimateTokens(prompt, { model: response.model });
+      const estimatedOutputTokens = estimateTokens(aiMarkdown, { model: response.model });
+      const inputTokens = usageNumber(response.usage?.inputTokens, estimatedInputTokens);
+      const outputTokens = usageNumber(response.usage?.outputTokens, estimatedOutputTokens);
       const modelCost = maybeEstimateModelCost({
         modelId: response.model,
         inputTokens,
         outputTokens,
       });
+      const costUsd = usageNumber(response.usage?.costUsd, modelCost.costUsd);
       const sessionId = String(options.sessionId || "scan-ai-precheck").trim() || "scan-ai-precheck";
 
       const appendedCost = await appendCostEntry(
@@ -639,7 +644,7 @@ export function registerScanCommand(program) {
           cacheWriteTokens: 0,
           durationMs,
           toolCalls: 1,
-          costUsd: modelCost.costUsd,
+          costUsd,
           progressScore: aiMarkdown ? 1 : 0,
         }
       );
@@ -682,7 +687,7 @@ export function registerScanCommand(program) {
             outputTokens,
             cacheReadTokens: 0,
             cacheWriteTokens: 0,
-            costUsd: modelCost.costUsd,
+            costUsd,
             durationMs,
             toolCalls: 1,
           },
@@ -727,9 +732,28 @@ export function registerScanCommand(program) {
               invocationId: appendedCost.entry.invocationId,
             },
           }
-        );
+          );
       }
 
+      const sessionUsageLedger = await recordCliLlmSessionUsage({
+        sessionId,
+        agentId: "scan-precheck",
+        action: "scan_precheck",
+        model: response.model,
+        inputTokens,
+        outputTokens,
+        startedAtIso,
+        targetPath,
+        sourceCommand: "scan precheck",
+        provider: response.provider,
+        metadata: {
+          specPath,
+          profile,
+          policyPackId: activePolicy.selected?.id || "",
+          pricingFound: modelCost.pricingFound,
+        },
+      });
+
       const payload = {
         command: "scan precheck",
         targetPath,
@@ -749,10 +773,11 @@ export function registerScanCommand(program) {
           usage: {
             inputTokens,
             outputTokens,
-            costUsd: modelCost.costUsd,
+            costUsd,
             durationMs,
             toolCalls: 1,
           },
+          billing: sessionUsageLedger,
           budget,
           cost: {
             filePath: appendedCost.filePath,

package/src/commands/spec.js

@@ -9,6 +9,7 @@ import {
   resolveModel,
   resolveProvider,
 } from "../ai/client.js";
+import { recordCliLlmSessionUsage, usageNumber } from "../billing/llm-session-usage.js";
 import { loadConfig } from "../config/service.js";
 import { evaluateBudget } from "../cost/budget.js";
 import { appendCostEntry, summarizeCostHistory } from "../cost/history.js";
@@ -273,6 +274,7 @@ async function maybeEnhanceSpecWithAi({
     ingest,
   });
 
+  const startedAtIso = new Date().toISOString();
   const startedAtMs = Date.now();
   const client = createMultiProviderApiClient();
   const result = await client.invoke({
@@ -288,13 +290,16 @@ async function maybeEnhanceSpecWithAi({
   const normalizedText = String(result.text || "").trim();
   const enhancedMarkdown = normalizedText || baseSpecMarkdown;
 
-  const inputTokens = estimateTokens(prompt, { model: result.model });
-  const outputTokens = estimateTokens(enhancedMarkdown, { model: result.model });
+  const estimatedInputTokens = estimateTokens(prompt, { model: result.model });
+  const estimatedOutputTokens = estimateTokens(enhancedMarkdown, { model: result.model });
+  const inputTokens = usageNumber(result.usage?.inputTokens, estimatedInputTokens);
+  const outputTokens = usageNumber(result.usage?.outputTokens, estimatedOutputTokens);
   const modelCost = maybeEstimateModelCost({
     modelId: result.model,
     inputTokens,
     outputTokens,
   });
+  const costUsd = usageNumber(result.usage?.costUsd, modelCost.costUsd);
 
   const sessionId = String(options.sessionId || "spec-generate-ai").trim() || "spec-generate-ai";
   const appendedCost = await appendCostEntry(
@@ -312,7 +317,7 @@ async function maybeEnhanceSpecWithAi({
       cacheWriteTokens: 0,
       durationMs,
       toolCalls: 1,
-      costUsd: modelCost.costUsd,
+      costUsd,
       progressScore: normalizedText ? 1 : 0,
     }
   );
@@ -355,7 +360,7 @@ async function maybeEnhanceSpecWithAi({
         outputTokens,
         cacheReadTokens: 0,
         cacheWriteTokens: 0,
-        costUsd: modelCost.costUsd,
+        costUsd,
         durationMs,
         toolCalls: 1,
       },
@@ -403,6 +408,24 @@ async function maybeEnhanceSpecWithAi({
     );
   }
 
+  const sessionUsageLedger = await recordCliLlmSessionUsage({
+    sessionId,
+    agentId: "spec-generator",
+    action: "spec_generate_ai",
+    model: result.model,
+    inputTokens,
+    outputTokens,
+    startedAtIso,
+    targetPath,
+    sourceCommand: "spec generate --ai",
+    provider: result.provider,
+    metadata: {
+      template: template?.id || template?.name || "",
+      projectType: template?.projectType || "",
+      pricingFound: modelCost.pricingFound,
+    },
+  });
+
   return {
     markdown: enhancedMarkdown,
     ai: {
@@ -413,10 +436,11 @@ async function maybeEnhanceSpecWithAi({
       usage: {
         inputTokens,
         outputTokens,
-        costUsd: modelCost.costUsd,
+        costUsd,
         durationMs,
         toolCalls: 1,
       },
+      billing: sessionUsageLedger,
       budget,
       cost: {
         filePath: appendedCost.filePath,