glotfile 0.7.1 → 0.7.2

package/dist/server/cli.js

@@ -2216,6 +2216,73 @@ var init_batch = __esm({
   }
 });
 
+// src/server/ai/pricing.ts
+function addUsage(into, add) {
+  into.inputTokens += add.inputTokens;
+  into.outputTokens += add.outputTokens;
+  into.cacheCreationInputTokens += add.cacheCreationInputTokens;
+  into.cacheReadInputTokens += add.cacheReadInputTokens;
+}
+function usageCostUsd(usage, ai, multiplier = 1) {
+  if (!usage) return void 0;
+  const pricing = resolvePricing(ai);
+  return pricing ? estimateUsageCostUsd(usage, pricing, multiplier) : void 0;
+}
+function estimateUsageCostUsd(usage, pricing, multiplier = 1) {
+  const inputCost = (usage.inputTokens + usage.cacheCreationInputTokens * CACHE_WRITE_MULTIPLIER + usage.cacheReadInputTokens * CACHE_READ_MULTIPLIER) * pricing.inputPerMTok;
+  return (inputCost + usage.outputTokens * pricing.outputPerMTok) / 1e6 * multiplier;
+}
+function bareModelId(model) {
+  let id = model.trim().toLowerCase();
+  const slash = id.lastIndexOf("/");
+  if (slash !== -1) id = id.slice(slash + 1);
+  const anth = id.lastIndexOf("anthropic.");
+  if (anth !== -1) id = id.slice(anth + "anthropic.".length);
+  return id;
+}
+function resolvePricing(ai) {
+  if (ai.inputPricePerMTok !== void 0 && ai.outputPricePerMTok !== void 0) {
+    return { source: "profile", inputPerMTok: ai.inputPricePerMTok, outputPerMTok: ai.outputPricePerMTok };
+  }
+  if (FREE_PROVIDERS.has(ai.provider)) return { source: "builtin", inputPerMTok: 0, outputPerMTok: 0 };
+  const id = bareModelId(ai.model);
+  let best;
+  for (const row of PRICE_TABLE) {
+    if (id.startsWith(row[0]) && (!best || row[0].length > best[0].length)) best = row;
+  }
+  return best ? { source: "builtin", inputPerMTok: best[1], outputPerMTok: best[2] } : null;
+}
+var BATCH_PRICE_MULTIPLIER, CACHE_WRITE_MULTIPLIER, CACHE_READ_MULTIPLIER, PRICE_TABLE, FREE_PROVIDERS;
+var init_pricing = __esm({
+  "src/server/ai/pricing.ts"() {
+    "use strict";
+    BATCH_PRICE_MULTIPLIER = 0.5;
+    CACHE_WRITE_MULTIPLIER = 1.25;
+    CACHE_READ_MULTIPLIER = 0.1;
+    PRICE_TABLE = [
+      ["claude-fable-5", 10, 50],
+      ["claude-mythos-5", 10, 50],
+      // Deprecated Opus 4.1 / 4.0 cost 3x the 4.5+ generation — they must outrank
+      // the shorter "claude-opus-4" prefix (4-2025 covers the dated Opus 4 full IDs).
+      ["claude-opus-4-1", 15, 75],
+      ["claude-opus-4-0", 15, 75],
+      ["claude-opus-4-2025", 15, 75],
+      ["claude-opus-4", 5, 25],
+      ["claude-sonnet-4", 3, 15],
+      ["claude-haiku-4", 1, 5],
+      ["claude-3-5-haiku", 0.8, 4],
+      ["gpt-5.5-pro", 30, 180],
+      ["gpt-5.5", 5, 30],
+      ["gpt-5.4-pro", 30, 180],
+      ["gpt-5.4-mini", 0.75, 4.5],
+      ["gpt-5.4-nano", 0.2, 1.25],
+      ["gpt-5.4", 2.5, 15],
+      ["gpt-5.3-codex", 1.75, 14]
+    ];
+    FREE_PROVIDERS = /* @__PURE__ */ new Set(["ollama", "claude-code"]);
+  }
+});
+
 // src/server/ai/anthropic.ts
 import Anthropic from "@anthropic-ai/sdk";
 var AnthropicProvider;
@@ -2224,6 +2291,7 @@ var init_anthropic = __esm({
     "use strict";
     init_provider();
     init_batch();
+    init_pricing();
     AnthropicProvider = class {
       constructor(config, client) {
         this.config = config;
@@ -2238,9 +2306,25 @@ var init_anthropic = __esm({
       }
       config;
       client;
+      usage = { inputTokens: 0, outputTokens: 0, cacheCreationInputTokens: 0, cacheReadInputTokens: 0 };
       supportsVision() {
         return true;
       }
+      recordUsage(usage) {
+        if (!usage) return;
+        addUsage(this.usage, {
+          inputTokens: usage.input_tokens ?? 0,
+          outputTokens: usage.output_tokens ?? 0,
+          cacheCreationInputTokens: usage.cache_creation_input_tokens ?? 0,
+          cacheReadInputTokens: usage.cache_read_input_tokens ?? 0
+        });
+      }
+      takeUsage() {
+        const taken = this.usage;
+        this.usage = { inputTokens: 0, outputTokens: 0, cacheCreationInputTokens: 0, cacheReadInputTokens: 0 };
+        const any = taken.inputTokens || taken.outputTokens || taken.cacheCreationInputTokens || taken.cacheReadInputTokens;
+        return any ? taken : void 0;
+      }
       translate(reqs, onBatchComplete, signal, onMalformedReply) {
         return runBatched(reqs, this.config.batchSize, (batch, sig) => this.callBatch(batch, sig), onBatchComplete, signal, onMalformedReply);
       }
@@ -2273,6 +2357,7 @@ var init_anthropic = __esm({
           output_config: { format: { type: "json_schema", schema: req.schema } },
           messages: [{ role: "user", content }]
         });
+        this.recordUsage(res.usage);
         const text = res.content.find((b) => b.type === "text")?.text ?? "{}";
         try {
           return JSON.parse(text);
@@ -2314,6 +2399,7 @@ var init_anthropic = __esm({
             out.set(entry.custom_id, { type: "failed", error: entry.result.error?.message ?? entry.result.type });
             continue;
           }
+          this.recordUsage(entry.result.message?.usage);
           const text = entry.result.message?.content.find((b) => b.type === "text")?.text ?? "";
           try {
             out.set(entry.custom_id, { type: "items", items: parseReplyItems(text) });
@@ -2336,6 +2422,7 @@ var init_anthropic = __esm({
           output_config: { format: { type: "json_schema", schema: BATCH_SCHEMA } },
           messages: [{ role: "user", content }]
         }, { signal });
+        this.recordUsage(res.usage);
         const text = res.content.find((b) => b.type === "text")?.text ?? "";
         return parseReplyItems(text);
       }
@@ -3133,6 +3220,30 @@ var init_pending_batch = __esm({
   }
 });
 
+// src/server/log.ts
+import { appendFileSync, readFileSync as readFileSync7, existsSync as existsSync7 } from "fs";
+import { resolve as resolve5 } from "path";
+function logPath(projectRoot) {
+  return resolve5(projectRoot, ".glotfile", "log.jsonl");
+}
+function appendLog(projectRoot, entry) {
+  ensureGlotfileDir(projectRoot);
+  appendFileSync(logPath(projectRoot), JSON.stringify(entry) + "\n", "utf8");
+}
+function readLog(projectRoot, limit = 100) {
+  const path = logPath(projectRoot);
+  if (!existsSync7(path)) return [];
+  const lines = readFileSync7(path, "utf8").split("\n").filter((l) => l.trim() !== "");
+  const entries = lines.map((l) => JSON.parse(l));
+  return entries.reverse().slice(0, limit);
+}
+var init_log = __esm({
+  "src/server/log.ts"() {
+    "use strict";
+    init_glotfile_dir();
+  }
+});
+
 // src/server/ai/batch-run.ts
 function buildBatchJobs(reqs, batchSize) {
   const byLocale = /* @__PURE__ */ new Map();
@@ -3180,7 +3291,9 @@ async function submitBatchTranslation(state, provider, reqs, batchSize, model, p
   return pending;
 }
 async function applyBatchResults(load, persist, provider, pending, projectRoot, ai) {
+  provider.takeUsage?.();
   const outcomes = await provider.translationBatchResults(pending.batchId);
+  const batchUsage = provider.takeUsage?.();
   const fresh = load();
   const isStale = (r) => {
     const entry = fresh.keys[r.key];
@@ -3189,13 +3302,19 @@ async function applyBatchResults(load, persist, provider, pending, projectRoot,
   const applied = [];
   const results = [];
   const retryReqs = [];
-  let staleSkipped = 0;
+  const stale = [];
+  const jobFailures = [];
   for (const job of pending.jobs) {
     const outcome = outcomes.get(job.customId);
     const itemsById = outcome?.type === "items" ? new Map(outcome.items.map((i) => [i.id, i])) : null;
+    if (!itemsById) {
+      if (!outcome) jobFailures.push({ customId: job.customId, locale: job.locale, type: "missing" });
+      else if (outcome.type === "malformed") jobFailures.push({ customId: job.customId, locale: job.locale, type: "malformed", raw: outcome.raw });
+      else if (outcome.type === "failed") jobFailures.push({ customId: job.customId, locale: job.locale, type: "failed", error: outcome.error });
+    }
     for (const stored of job.requests) {
       if (isStale(stored)) {
-        staleSkipped++;
+        stale.push({ key: stored.key, locale: stored.targetLocale });
         continue;
       }
       const { sourceHash: _hash, ...req } = stored;
@@ -3214,7 +3333,20 @@ async function applyBatchResults(load, persist, provider, pending, projectRoot,
     const retryResults = await runLocaleParallel(
       retryReqs,
       provider,
-      {},
+      {
+        // Record the raw reply so an unparseable retry response is diagnosable
+        // from the activity log instead of vanishing into per-item errors.
+        onMalformedReply: (raw, batchSize, locale) => {
+          appendLog(projectRoot, {
+            at: (/* @__PURE__ */ new Date()).toISOString(),
+            kind: "translate",
+            summary: `Malformed model reply (${locale}, batch of ${batchSize})`,
+            model: pending.model,
+            locale,
+            raw
+          });
+        }
+      },
       ai.concurrency,
       void 0,
       ai.batchSize
@@ -3222,10 +3354,34 @@ async function applyBatchResults(load, persist, provider, pending, projectRoot,
     applied.push(...retryReqs);
     results.push(...retryResults);
   }
+  const retryUsage = provider.takeUsage?.();
+  const pricing = resolvePricing({ ...ai, model: pending.model });
+  let estimatedCostUsd;
+  if (pricing && (batchUsage || retryUsage)) {
+    estimatedCostUsd = (batchUsage ? estimateUsageCostUsd(batchUsage, pricing, BATCH_PRICE_MULTIPLIER) : 0) + (retryUsage ? estimateUsageCostUsd(retryUsage, pricing) : 0);
+  }
+  let usage;
+  if (batchUsage || retryUsage) {
+    usage = batchUsage ?? { inputTokens: 0, outputTokens: 0, cacheCreationInputTokens: 0, cacheReadInputTokens: 0 };
+    if (retryUsage) addUsage(usage, retryUsage);
+  }
   const { written, errors } = applyResults(fresh, applied, results);
   persist(fresh);
   clearPendingBatch(projectRoot);
-  return { written, errors, staleSkipped, retried: retryReqs.length, screenshotsSkipped };
+  const costSuffix = estimatedCostUsd !== void 0 ? ` (~$${estimatedCostUsd.toFixed(2)})` : "";
+  appendLog(projectRoot, {
+    at: (/* @__PURE__ */ new Date()).toISOString(),
+    kind: "translate",
+    summary: `Applied batch ${pending.batchId}: wrote ${written}, ${errors.length} error(s), ${retryReqs.length} retried, ${stale.length} stale${costSuffix}`,
+    model: pending.model,
+    items: applied.map((r) => ({ id: r.id, key: r.key, source: r.source, targetLocale: r.targetLocale })),
+    results,
+    jobFailures: jobFailures.length ? jobFailures : void 0,
+    stale: stale.length ? stale : void 0,
+    usage,
+    estimatedCostUsd
+  });
+  return { written, errors, staleSkipped: stale.length, retried: retryReqs.length, screenshotsSkipped };
 }
 var init_batch_run = __esm({
   "src/server/ai/batch-run.ts"() {
@@ -3234,55 +3390,8 @@ var init_batch_run = __esm({
     init_batch();
     init_run();
     init_pending_batch();
-  }
-});
-
-// src/server/ai/pricing.ts
-function bareModelId(model) {
-  let id = model.trim().toLowerCase();
-  const slash = id.lastIndexOf("/");
-  if (slash !== -1) id = id.slice(slash + 1);
-  const anth = id.lastIndexOf("anthropic.");
-  if (anth !== -1) id = id.slice(anth + "anthropic.".length);
-  return id;
-}
-function resolvePricing(ai) {
-  if (ai.inputPricePerMTok !== void 0 && ai.outputPricePerMTok !== void 0) {
-    return { source: "profile", inputPerMTok: ai.inputPricePerMTok, outputPerMTok: ai.outputPricePerMTok };
-  }
-  if (FREE_PROVIDERS.has(ai.provider)) return { source: "builtin", inputPerMTok: 0, outputPerMTok: 0 };
-  const id = bareModelId(ai.model);
-  let best;
-  for (const row of PRICE_TABLE) {
-    if (id.startsWith(row[0]) && (!best || row[0].length > best[0].length)) best = row;
-  }
-  return best ? { source: "builtin", inputPerMTok: best[1], outputPerMTok: best[2] } : null;
-}
-var PRICE_TABLE, FREE_PROVIDERS;
-var init_pricing = __esm({
-  "src/server/ai/pricing.ts"() {
-    "use strict";
-    PRICE_TABLE = [
-      ["claude-fable-5", 10, 50],
-      ["claude-mythos-5", 10, 50],
-      // Deprecated Opus 4.1 / 4.0 cost 3x the 4.5+ generation — they must outrank
-      // the shorter "claude-opus-4" prefix (4-2025 covers the dated Opus 4 full IDs).
-      ["claude-opus-4-1", 15, 75],
-      ["claude-opus-4-0", 15, 75],
-      ["claude-opus-4-2025", 15, 75],
-      ["claude-opus-4", 5, 25],
-      ["claude-sonnet-4", 3, 15],
-      ["claude-haiku-4", 1, 5],
-      ["claude-3-5-haiku", 0.8, 4],
-      ["gpt-5.5-pro", 30, 180],
-      ["gpt-5.5", 5, 30],
-      ["gpt-5.4-pro", 30, 180],
-      ["gpt-5.4-mini", 0.75, 4.5],
-      ["gpt-5.4-nano", 0.2, 1.25],
-      ["gpt-5.4", 2.5, 15],
-      ["gpt-5.3-codex", 1.75, 14]
-    ];
-    FREE_PROVIDERS = /* @__PURE__ */ new Set(["ollama", "claude-code"]);
+    init_log();
+    init_pricing();
   }
 });
 
@@ -3349,30 +3458,6 @@ var init_estimate = __esm({
   }
 });
 
-// src/server/log.ts
-import { appendFileSync, readFileSync as readFileSync7, existsSync as existsSync7 } from "fs";
-import { resolve as resolve5 } from "path";
-function logPath(projectRoot) {
-  return resolve5(projectRoot, ".glotfile", "log.jsonl");
-}
-function appendLog(projectRoot, entry) {
-  ensureGlotfileDir(projectRoot);
-  appendFileSync(logPath(projectRoot), JSON.stringify(entry) + "\n", "utf8");
-}
-function readLog(projectRoot, limit = 100) {
-  const path = logPath(projectRoot);
-  if (!existsSync7(path)) return [];
-  const lines = readFileSync7(path, "utf8").split("\n").filter((l) => l.trim() !== "");
-  const entries = lines.map((l) => JSON.parse(l));
-  return entries.reverse().slice(0, limit);
-}
-var init_log = __esm({
-  "src/server/log.ts"() {
-    "use strict";
-    init_glotfile_dir();
-  }
-});
-
 // src/server/scan.ts
 import { existsSync as existsSync8, readFileSync as readFileSync8 } from "fs";
 import { resolve as resolve6 } from "path";
@@ -6836,6 +6921,7 @@ function createApi(deps) {
           persist(fresh);
           totalWritten += written;
           allErrors.push(...errors);
+          const usage = provider.takeUsage?.();
           appendLog(projectRoot, {
             at: (/* @__PURE__ */ new Date()).toISOString(),
             kind: "translate",
@@ -6846,7 +6932,9 @@ function createApi(deps) {
               const req = reqById.get(r.id);
               return { id: r.id, key: req?.key ?? "", source: req?.source ?? "", targetLocale: req?.targetLocale, context: req?.context, glossary: req?.glossary, screenshot: req ? fresh.keys[req.key]?.screenshot : void 0 };
             }),
-            results: batchResults
+            results: batchResults,
+            usage,
+            estimatedCostUsd: usageCostUsd(usage, aiCfg)
           });
           const ld = (localeDone.get(locale) ?? 0) + batchResults.length;
           localeDone.set(locale, ld);
@@ -6918,11 +7006,14 @@ function createApi(deps) {
       }, aiCfg.concurrency, void 0, aiCfg.batchSize);
       const latest = load();
       ({ written, errors } = applyResults(latest, toTranslate, results, void 0, force));
+      const usage = provider.takeUsage?.();
       const entry = {
         at: (/* @__PURE__ */ new Date()).toISOString(),
         kind: "translate",
         summary: `Translated ${toTranslate.length} item(s)`,
         model: aiCfg.model,
+        usage,
+        estimatedCostUsd: usageCostUsd(usage, aiCfg),
         system: buildSystemPrompt(toTranslate.some((r) => r.plural !== void 0)),
         // Log the screenshot PATH only — never the image bytes.
         items: toTranslate.map((r) => ({
@@ -7020,17 +7111,7 @@ function createApi(deps) {
     if (!supportsBatchTranslate(provider)) {
       return c.json({ error: `Provider "${aiCfg.provider}" does not support batch mode.` }, 400);
     }
-    const outcome = await applyBatchResults(load, persist, provider, pending, projectRoot, {
-      batchSize: aiCfg.batchSize,
-      concurrency: aiCfg.concurrency
-    });
-    appendLog(projectRoot, {
-      at: (/* @__PURE__ */ new Date()).toISOString(),
-      kind: "translate",
-      summary: `Applied batch ${pending.batchId}: wrote ${outcome.written}, ${outcome.retried} retried, ${outcome.staleSkipped} stale`,
-      model: aiCfg.model,
-      results: []
-    });
+    const outcome = await applyBatchResults(load, persist, provider, pending, projectRoot, aiCfg);
     console.log(`[batch] applied ${pending.batchId} \u2014 wrote ${outcome.written}, ${outcome.errors.length} error(s)`);
     return c.json(outcome);
   }));
@@ -7181,6 +7262,7 @@ function createApi(deps) {
           const batch = raw;
           const fresh = load();
           const { written, errors } = applyContext(fresh, chunk2, batch.items ?? [], void 0, body.force === true);
+          const usage = provider.takeUsage?.();
           appendLog(projectRoot, {
             at: (/* @__PURE__ */ new Date()).toISOString(),
             kind: "context",
@@ -7188,7 +7270,9 @@ function createApi(deps) {
             model: aiCfg.model,
             system,
             items: chunk2.map((t) => ({ id: t.id, key: t.key, source: t.source })),
-            results: (batch.items ?? []).map((r) => ({ id: r.id, value: r.context, error: r.error }))
+            results: (batch.items ?? []).map((r) => ({ id: r.id, value: r.context, error: r.error })),
+            usage,
+            estimatedCostUsd: usageCostUsd(usage, aiCfg)
           });
           persist(fresh);
           totalWritten += written;
@@ -7229,6 +7313,7 @@ var init_api = __esm({
     init_batch_run();
     init_pending_batch();
     init_estimate();
+    init_pricing();
     init_log();
     init_schema();
     init_run3();
@@ -7391,6 +7476,7 @@ init_provider();
 init_batch_run();
 init_pending_batch();
 init_estimate();
+init_pricing();
 init_log();
 init_scan();
 init_scanner();
@@ -7713,11 +7799,14 @@ async function runTranslate(args) {
     if (!batchCallbackFired) {
       ({ written, errors } = applyResults(state, toTranslate, results));
     }
+    const usage = provider.takeUsage?.();
     appendLog(projectRoot, {
       at: (/* @__PURE__ */ new Date()).toISOString(),
       kind: "translate",
       summary: `Translated ${toTranslate.length} item(s)`,
       model: ai.model,
+      usage,
+      estimatedCostUsd: usageCostUsd(usage, ai),
       system: buildSystemPrompt(toTranslate.some((r) => r.plural !== void 0)),
       items: toTranslate.map((r) => ({
         id: r.id,
@@ -7752,7 +7841,7 @@ async function applyPending(args, provider, pending, ai) {
     provider,
     pending,
     projectRoot,
-    { batchSize: ai.batchSize, concurrency: ai.concurrency }
+    ai
   );
   reportApply(outcome);
 }