feat: add usage cost reporting

src/auto-reply/reply/agent-runner.ts

@@ -2,12 +2,13 @@ import crypto from "node:crypto";
 import fs from "node:fs";
 import { lookupContextTokens } from "../../agents/context.js";
 import { DEFAULT_CONTEXT_TOKENS } from "../../agents/defaults.js";
+import { resolveModelAuthMode } from "../../agents/model-auth.js";
 import { runWithModelFallback } from "../../agents/model-fallback.js";
 import {
   queueEmbeddedPiMessage,
   runEmbeddedPiAgent,
 } from "../../agents/pi-embedded.js";
-import { hasNonzeroUsage } from "../../agents/usage.js";
+import { hasNonzeroUsage, type NormalizedUsage } from "../../agents/usage.js";
 import {
   loadSessionStore,
   resolveSessionTranscriptPath,
@@ -18,6 +19,12 @@ import type { TypingMode } from "../../config/types.js";
 import { logVerbose } from "../../globals.js";
 import { registerAgentRunContext } from "../../infra/agent-events.js";
 import { defaultRuntime } from "../../runtime.js";
+import {
+  estimateUsageCost,
+  formatTokenCount,
+  formatUsd,
+  resolveModelCostConfig,
+} from "../../utils/usage-format.js";
 import { stripHeartbeatToken } from "../heartbeat.js";
 import type { OriginatingChannelType, TemplateContext } from "../templating.js";
 import { normalizeVerboseLevel, type VerboseLevel } from "../thinking.js";
@@ -62,6 +69,65 @@ const formatBunFetchSocketError = (message: string) => {
   ].join("\n");
 };
 
+const formatResponseUsageLine = (params: {
+  usage?: NormalizedUsage;
+  showCost: boolean;
+  costConfig?: {
+    input: number;
+    output: number;
+    cacheRead: number;
+    cacheWrite: number;
+  };
+}): string | null => {
+  const usage = params.usage;
+  if (!usage) return null;
+  const input = usage.input;
+  const output = usage.output;
+  if (typeof input !== "number" && typeof output !== "number") return null;
+  const inputLabel = typeof input === "number" ? formatTokenCount(input) : "?";
+  const outputLabel =
+    typeof output === "number" ? formatTokenCount(output) : "?";
+  const cost =
+    params.showCost && typeof input === "number" && typeof output === "number"
+      ? estimateUsageCost({
+          usage: {
+            input,
+            output,
+            cacheRead: usage.cacheRead,
+            cacheWrite: usage.cacheWrite,
+          },
+          cost: params.costConfig,
+        })
+      : undefined;
+  const costLabel = params.showCost ? formatUsd(cost) : undefined;
+  const suffix = costLabel ? ` · est ${costLabel}` : "";
+  return `Usage: ${inputLabel} in / ${outputLabel} out${suffix}`;
+};
+
+const appendUsageLine = (
+  payloads: ReplyPayload[],
+  line: string,
+): ReplyPayload[] => {
+  let index = -1;
+  for (let i = payloads.length - 1; i >= 0; i -= 1) {
+    if (payloads[i]?.text) {
+      index = i;
+      break;
+    }
+  }
+  if (index === -1) return [...payloads, { text: line }];
+  const existing = payloads[index];
+  const existingText = existing.text ?? "";
+  const separator = existingText.endsWith("\n") ? "" : "\n";
+  const next = {
+    ...existing,
+    text: `${existingText}${separator}${line}`,
+  };
+  const updated = payloads.slice();
+  updated[index] = next;
+  return updated;
+};
+
 const withTimeout = async <T>(
   promise: Promise<T>,
   timeoutMs: number,
@@ -191,6 +257,7 @@ export async function runReplyAgent(params: {
     replyToChannel,
   );
   const applyReplyToMode = createReplyToModeFilter(replyToMode);
+  const cfg = followupRun.run.config;
 
   if (shouldSteer && isStreaming) {
     const steered = queueEmbeddedPiMessage(
@@ -242,6 +309,7 @@ export async function runReplyAgent(params: {
 
   let didLogHeartbeatStrip = false;
   let autoCompactionCompleted = false;
+  let responseUsageLine: string | undefined;
   try {
     const runId = crypto.randomUUID();
     if (sessionKey) {
@@ -641,20 +709,20 @@ export async function runReplyAgent(params: {
       await typingSignals.signalRunStart();
     }
 
-    if (sessionStore && sessionKey) {
-      const usage = runResult.meta.agentMeta?.usage;
-      const modelUsed =
-        runResult.meta.agentMeta?.model ?? fallbackModel ?? defaultModel;
-      const providerUsed =
-        runResult.meta.agentMeta?.provider ??
-        fallbackProvider ??
-        followupRun.run.provider;
-      const contextTokensUsed =
-        agentCfgContextTokens ??
-        lookupContextTokens(modelUsed) ??
-        sessionEntry?.contextTokens ??
-        DEFAULT_CONTEXT_TOKENS;
+    const usage = runResult.meta.agentMeta?.usage;
+    const modelUsed =
+      runResult.meta.agentMeta?.model ?? fallbackModel ?? defaultModel;
+    const providerUsed =
+      runResult.meta.agentMeta?.provider ??
+      fallbackProvider ??
+      followupRun.run.provider;
+    const contextTokensUsed =
+      agentCfgContextTokens ??
+      lookupContextTokens(modelUsed) ??
+      sessionEntry?.contextTokens ??
+      DEFAULT_CONTEXT_TOKENS;
 
+    if (sessionStore && sessionKey) {
       if (hasNonzeroUsage(usage)) {
         const entry = sessionEntry ?? sessionStore[sessionKey];
         if (entry) {
@@ -694,6 +762,29 @@ export async function runReplyAgent(params: {
       }
     }
 
+    const responseUsageEnabled =
+      (sessionEntry?.responseUsage ??
+        (sessionKey
+          ? sessionStore?.[sessionKey]?.responseUsage
+          : undefined)) === "on";
+    if (responseUsageEnabled && hasNonzeroUsage(usage)) {
+      const authMode = resolveModelAuthMode(providerUsed, cfg);
+      const showCost = authMode === "api-key";
+      const costConfig = showCost
+        ? resolveModelCostConfig({
+            provider: providerUsed,
+            model: modelUsed,
+            config: cfg,
+          })
+        : undefined;
+      const formatted = formatResponseUsageLine({
+        usage,
+        showCost,
+        costConfig,
+      });
+      if (formatted) responseUsageLine = formatted;
+    }
+
     // If verbose is enabled and this is a new session, prepend a session hint.
     let finalPayloads = replyPayloads;
     if (autoCompactionCompleted) {
@@ -717,6 +808,9 @@ export async function runReplyAgent(params: {
         ...finalPayloads,
       ];
     }
+    if (responseUsageLine) {
+      finalPayloads = appendUsageLine(finalPayloads, responseUsageLine);
+    }
 
     return finalizeWithFollowup(
       finalPayloads.length === 1 ? finalPayloads[0] : finalPayloads,

src/auto-reply/reply/commands.ts

@@ -1,11 +1,5 @@
-import {
-  ensureAuthProfileStore,
-  listProfilesForProvider,
-} from "../../agents/auth-profiles.js";
-import {
-  getCustomProviderApiKey,
-  resolveEnvApiKey,
-} from "../../agents/model-auth.js";
+import crypto from "node:crypto";
+import { resolveModelAuthMode } from "../../agents/model-auth.js";
 import { normalizeProviderId } from "../../agents/model-selection.js";
 import {
   abortEmbeddedPiRun,
@@ -55,8 +49,10 @@ import type {
   ElevatedLevel,
   ReasoningLevel,
   ThinkLevel,
+  UsageDisplayLevel,
   VerboseLevel,
 } from "../thinking.js";
+import { normalizeUsageDisplay } from "../thinking.js";
 import type { ReplyPayload } from "../types.js";
 import { isAbortTrigger, setAbortMemory } from "./abort.js";
 import type { InlineDirectives } from "./directive-handling.js";
@@ -109,36 +105,6 @@ export type CommandContext = {
   to?: string;
 };
 
-function resolveModelAuthLabel(
-  provider?: string,
-  cfg?: ClawdbotConfig,
-): string | undefined {
-  const resolved = provider?.trim();
-  if (!resolved) return undefined;
-
-  const store = ensureAuthProfileStore();
-  const profiles = listProfilesForProvider(store, resolved);
-  if (profiles.length > 0) {
-    const modes = new Set(
-      profiles
-        .map((id) => store.profiles[id]?.type)
-        .filter((mode): mode is "api_key" | "oauth" => Boolean(mode)),
-    );
-    if (modes.has("oauth") && modes.has("api_key")) return "mixed";
-    if (modes.has("oauth")) return "oauth";
-    if (modes.has("api_key")) return "api-key";
-  }
-
-  const envKey = resolveEnvApiKey(resolved);
-  if (envKey?.apiKey) {
-    return envKey.source.includes("OAUTH_TOKEN") ? "oauth" : "api-key";
-  }
-
-  if (getCustomProviderApiKey(cfg, resolved)) return "api-key";
-
-  return "unknown";
-}
-
 function extractCompactInstructions(params: {
   rawBody?: string;
   ctx: MsgContext;
@@ -468,6 +434,7 @@ export async function handleCommands(params: {
         defaultGroupActivation())
       : undefined;
     const statusText = buildStatusMessage({
+      config: cfg,
       agent: {
         ...cfg.agent,
         model: {
@@ -488,7 +455,7 @@ export async function handleCommands(params: {
       resolvedVerbose: resolvedVerboseLevel,
       resolvedReasoning: resolvedReasoningLevel,
       resolvedElevated: resolvedElevatedLevel,
-      modelAuth: resolveModelAuthLabel(provider, cfg),
+      modelAuth: resolveModelAuthMode(provider, cfg),
       usageLine: usageLine ?? undefined,
       queue: {
         mode: queueSettings.mode,
@@ -503,6 +470,51 @@ export async function handleCommands(params: {
     return { shouldContinue: false, reply: { text: statusText } };
   }
 
+  const costRequested =
+    command.commandBodyNormalized === "/cost" ||
+    command.commandBodyNormalized.startsWith("/cost ");
+  if (allowTextCommands && costRequested) {
+    if (!command.isAuthorizedSender) {
+      logVerbose(
+        `Ignoring /cost from unauthorized sender: ${command.senderE164 || "<unknown>"}`,
+      );
+      return { shouldContinue: false };
+    }
+    const rawArgs = command.commandBodyNormalized.slice("/cost".length).trim();
+    const normalized =
+      rawArgs.length > 0 ? normalizeUsageDisplay(rawArgs) : undefined;
+    if (rawArgs.length > 0 && !normalized) {
+      return {
+        shouldContinue: false,
+        reply: { text: "⚙️ Usage: /cost on|off" },
+      };
+    }
+    const current: UsageDisplayLevel =
+      sessionEntry?.responseUsage === "on" ? "on" : "off";
+    const next = normalized ?? (current === "on" ? "off" : "on");
+    if (sessionStore && sessionKey) {
+      const entry = sessionEntry ??
+        sessionStore[sessionKey] ?? {
+          sessionId: crypto.randomUUID(),
+          updatedAt: Date.now(),
+        };
+      if (next === "off") delete entry.responseUsage;
+      else entry.responseUsage = next;
+      entry.updatedAt = Date.now();
+      sessionStore[sessionKey] = entry;
+      if (storePath) {
+        await saveSessionStore(storePath, sessionStore);
+      }
+    }
+    return {
+      shouldContinue: false,
+      reply: {
+        text:
+          next === "on" ? "⚙️ Usage line enabled." : "⚙️ Usage line disabled.",
+      },
+    };
+  }
+
   const stopRequested = command.commandBodyNormalized === "/stop";
   if (allowTextCommands && stopRequested) {
     if (!command.isAuthorizedSender) {

src/auto-reply/reply/session.ts

@@ -194,6 +194,7 @@ export async function initSessionState(params: {
     // Persist previously stored thinking/verbose levels when present.
     thinkingLevel: persistedThinking ?? baseEntry?.thinkingLevel,
     verboseLevel: persistedVerbose ?? baseEntry?.verboseLevel,
+    responseUsage: baseEntry?.responseUsage,
     modelOverride: persistedModelOverride ?? baseEntry?.modelOverride,
     providerOverride: persistedProviderOverride ?? baseEntry?.providerOverride,
     sendPolicy: baseEntry?.sendPolicy,