ai-speedometer 2.1.6 → 2.2.0

package/dist/ai-speedometer

@@ -1712,7 +1712,7 @@ var exports_benchmark = {};
 __export(exports_benchmark, {
   benchmarkSingleModelRest: () => benchmarkSingleModelRest
 });
-async function benchmarkSingleModelRest(model) {
+async function benchmarkSingleModelRest(model, logger) {
   try {
     if (!model.providerConfig || !model.providerConfig.apiKey) {
       throw new Error(`Missing API key for provider ${model.providerName}`);
@@ -1729,6 +1729,7 @@ async function benchmarkSingleModelRest(model) {
       actualModelId = model.name;
     }
     actualModelId = actualModelId.trim();
+    await logger?.logHeader(model.name, model.providerName, model.providerConfig.apiKey);
     const startTime = Date.now();
     let firstTokenTime = null;
     let streamedText = "";
@@ -1801,33 +1802,36 @@ async function benchmarkSingleModelRest(model) {
     const reader = response.body.getReader();
     const decoder = new TextDecoder;
     let buffer = "";
-    let isFirstChunk = true;
+    let firstParsedTokenTime = null;
     while (true) {
       const { done, value } = await reader.read();
       if (done)
         break;
-      if (isFirstChunk && !firstTokenTime) {
+      if (!firstTokenTime)
         firstTokenTime = Date.now();
-        isFirstChunk = false;
-      }
       buffer += decoder.decode(value, { stream: true });
       const lines = buffer.split(`
 `);
       buffer = lines.pop() || "";
       for (const line of lines) {
         const trimmedLine = line.trim();
+        if (trimmedLine)
+          await logger?.logRaw(trimmedLine);
         if (!trimmedLine)
           continue;
         try {
           if (model.providerType === "anthropic") {
-            if (trimmedLine.startsWith("data: ")) {
-              const jsonStr = trimmedLine.slice(6);
+            const anthropicDataPrefix = trimmedLine.startsWith("data: ") ? 6 : trimmedLine.startsWith("data:") ? 5 : -1;
+            if (anthropicDataPrefix !== -1) {
+              const jsonStr = trimmedLine.slice(anthropicDataPrefix);
               if (jsonStr === "[DONE]")
                 continue;
               const chunk = JSON.parse(jsonStr);
               const chunkTyped = chunk;
-              if (chunkTyped.type === "content_block_delta" && chunkTyped.delta?.text) {
-                streamedText += chunkTyped.delta.text;
+              if (chunkTyped.type === "content_block_delta" && (chunkTyped.delta?.text || chunkTyped.delta?.thinking)) {
+                if (!firstParsedTokenTime)
+                  firstParsedTokenTime = Date.now();
+                streamedText += chunkTyped.delta?.text || chunkTyped.delta?.thinking || "";
               } else if (chunkTyped.type === "message_start" && chunkTyped.message?.usage) {
                 inputTokens = chunkTyped.message.usage.input_tokens || 0;
               } else if (chunkTyped.type === "message_delta") {
@@ -1840,8 +1844,10 @@ async function benchmarkSingleModelRest(model) {
               continue;
             } else {
               const chunk = JSON.parse(trimmedLine);
-              if (chunk.type === "content_block_delta" && chunk.delta?.text) {
-                streamedText += chunk.delta.text;
+              if (chunk.type === "content_block_delta" && (chunk.delta?.text || chunk.delta?.thinking)) {
+                if (!firstParsedTokenTime)
+                  firstParsedTokenTime = Date.now();
+                streamedText += chunk.delta?.text || chunk.delta?.thinking || "";
               } else if (chunk.type === "message_start" && chunk.message?.usage) {
                 inputTokens = chunk.message.usage.input_tokens || 0;
               } else if (chunk.type === "message_delta") {
@@ -1854,6 +1860,8 @@ async function benchmarkSingleModelRest(model) {
           } else if (model.providerType === "google") {
             const chunk = JSON.parse(trimmedLine);
             if (chunk.candidates?.[0]?.content?.parts?.[0]?.text) {
+              if (!firstParsedTokenTime)
+                firstParsedTokenTime = Date.now();
               streamedText += chunk.candidates[0].content.parts[0].text;
             }
             if (chunk.usageMetadata?.promptTokenCount)
@@ -1861,19 +1869,42 @@ async function benchmarkSingleModelRest(model) {
             if (chunk.usageMetadata?.candidatesTokenCount)
               outputTokens = chunk.usageMetadata.candidatesTokenCount;
           } else {
-            if (trimmedLine.startsWith("data: ")) {
-              const jsonStr = trimmedLine.slice(6);
-              if (jsonStr === "[DONE]")
-                continue;
-              const chunk = JSON.parse(jsonStr);
-              if (chunk.choices?.[0]?.delta?.content)
-                streamedText += chunk.choices[0].delta.content;
-              else if (chunk.choices?.[0]?.delta?.reasoning)
-                streamedText += chunk.choices[0].delta.reasoning;
-              if (chunk.usage?.prompt_tokens)
-                inputTokens = chunk.usage.prompt_tokens;
-              if (chunk.usage?.completion_tokens)
-                outputTokens = chunk.usage.completion_tokens;
+            const dataPrefix = trimmedLine.startsWith("data: ") ? 6 : trimmedLine.startsWith("data:") ? 5 : -1;
+            if (dataPrefix === -1)
+              continue;
+            const jsonStr = trimmedLine.slice(dataPrefix);
+            if (jsonStr === "[DONE]")
+              continue;
+            const chunk = JSON.parse(jsonStr);
+            if (chunk.choices?.[0]?.delta?.content) {
+              if (!firstParsedTokenTime)
+                firstParsedTokenTime = Date.now();
+              streamedText += chunk.choices[0].delta.content;
+            } else if (chunk.choices?.[0]?.delta?.reasoning) {
+              if (!firstParsedTokenTime)
+                firstParsedTokenTime = Date.now();
+              streamedText += chunk.choices[0].delta.reasoning;
+            } else if (chunk.choices?.[0]?.delta?.reasoning_content) {
+              if (!firstParsedTokenTime)
+                firstParsedTokenTime = Date.now();
+              streamedText += chunk.choices[0].delta.reasoning_content;
+            } else if (chunk.type === "content_block_delta" && chunk.delta?.text) {
+              if (!firstParsedTokenTime)
+                firstParsedTokenTime = Date.now();
+              streamedText += chunk.delta.text;
+            } else if (chunk.type === "content_block_delta" && chunk.delta?.thinking) {
+              if (!firstParsedTokenTime)
+                firstParsedTokenTime = Date.now();
+              streamedText += chunk.delta.thinking;
+            }
+            if (chunk.usage?.prompt_tokens)
+              inputTokens = chunk.usage.prompt_tokens;
+            if (chunk.usage?.completion_tokens)
+              outputTokens = chunk.usage.completion_tokens;
+            if (chunk.type === "message_start" && chunk.message?.usage?.input_tokens)
+              inputTokens = chunk.message.usage.input_tokens;
+            if (chunk.type === "message_delta" && chunk.usage?.output_tokens) {
+              outputTokens = chunk.usage.output_tokens;
             }
           }
         } catch {
@@ -1881,15 +1912,18 @@ async function benchmarkSingleModelRest(model) {
         }
       }
     }
+    await logger?.flush();
     const endTime = Date.now();
     const totalTime = endTime - startTime;
-    const timeToFirstToken = firstTokenTime ? firstTokenTime - startTime : totalTime;
+    const effectiveFirstToken = firstParsedTokenTime ?? firstTokenTime;
+    const timeToFirstToken = effectiveFirstToken ? effectiveFirstToken - startTime : totalTime;
+    const generationTime = totalTime - timeToFirstToken;
     const usedEstimateForOutput = !outputTokens;
     const usedEstimateForInput = !inputTokens;
     const finalOutputTokens = outputTokens || Math.round(streamedText.length / 4);
     const finalInputTokens = inputTokens || Math.round(TEST_PROMPT.length / 4);
     const totalTokens = finalInputTokens + finalOutputTokens;
-    const tokensPerSecond = totalTime > 0 ? finalOutputTokens / totalTime * 1000 : 0;
+    const tokensPerSecond = generationTime > 0 ? finalOutputTokens / generationTime * 1000 : 0;
     return {
       model: model.name,
       provider: model.providerName,
@@ -1904,6 +1938,7 @@ async function benchmarkSingleModelRest(model) {
       success: true
     };
   } catch (error) {
+    await logger?.flush();
     return {
       model: model.name,
       provider: model.providerName,
@@ -2085,6 +2120,66 @@ var init_headless = __esm(() => {
   init_benchmark();
 });
 
+// ../core/src/logger.ts
+var exports_logger = {};
+__export(exports_logger, {
+  getLogPath: () => getLogPath,
+  createRunId: () => createRunId,
+  createBenchLogger: () => createBenchLogger
+});
+import { mkdir, appendFile } from "fs/promises";
+import { homedir as homedir4 } from "os";
+import { join } from "path";
+function generateRunId() {
+  const now = new Date;
+  const date = now.toISOString().slice(0, 10);
+  const time = now.toTimeString().slice(0, 8).replace(/:/g, "");
+  const rand = Math.random().toString(16).slice(2, 6);
+  return `${date}_${time}_${rand}`;
+}
+function redactSecrets(line, apiKey) {
+  if (!apiKey)
+    return line;
+  return line.split(apiKey).join("[REDACTED]");
+}
+function createRunId() {
+  return generateRunId();
+}
+function getLogPath(runId) {
+  return join(homedir4(), ".local", "share", "ai-speedometer", "logs", `${runId}.log`);
+}
+async function createBenchLogger(runId) {
+  const logPath = getLogPath(runId);
+  const logDir = join(homedir4(), ".local", "share", "ai-speedometer", "logs");
+  await mkdir(logDir, { recursive: true });
+  let currentApiKey = "";
+  let buffer = "";
+  return {
+    logPath,
+    runId,
+    logHeader: async (modelName, providerName, apiKey = "") => {
+      currentApiKey = apiKey;
+      const ts = new Date().toISOString();
+      buffer = `
+=== ${modelName} | ${providerName} | ${ts} ===
+`;
+    },
+    logRaw: async (line) => {
+      const safe = redactSecrets(line, currentApiKey);
+      buffer += safe + `
+`;
+    },
+    flush: async () => {
+      buffer += `
+` + "=".repeat(60) + `
+`;
+      await appendFile(logPath, buffer, "utf8");
+      buffer = "";
+    }
+  };
+}
+var init_logger = () => {};
+
 // src/tui/context/AppContext.tsx
 import { createContext, useContext, useReducer, useEffect } from "react";
 import { jsxDEV } from "@opentui/react/jsx-dev-runtime";
@@ -2121,12 +2216,23 @@ function appReducer(state, action) {
       };
     case "BENCH_RESET":
       return { ...state, benchResults: [], selectedModels: [] };
+    case "SET_LOG_INFO":
+      return { ...state, logMode: action.logMode, logPath: action.logPath, runId: action.runId };
     default:
       return state;
   }
 }
-function AppProvider({ children }) {
+function AppProvider({ children, logMode = false }) {
   const [state, dispatch] = useReducer(appReducer, initialState);
+  useEffect(() => {
+    if (logMode) {
+      Promise.resolve().then(() => (init_logger(), exports_logger)).then(({ createRunId: createRunId2, getLogPath: getLogPath2 }) => {
+        const runId = createRunId2();
+        const logPath = getLogPath2(runId);
+        dispatch({ type: "SET_LOG_INFO", logMode: true, logPath, runId });
+      });
+    }
+  }, [logMode]);
   useEffect(() => {
     let cancelled = false;
     async function loadConfig2() {
@@ -2169,7 +2275,10 @@ var init_AppContext = __esm(() => {
     config: null,
     selectedModels: [],
     benchResults: [],
-    isLoadingConfig: true
+    isLoadingConfig: true,
+    logMode: false,
+    logPath: null,
+    runId: null
   };
   AppContext = createContext(null);
 });
@@ -2179,7 +2288,7 @@ var package_default;
 var init_package = __esm(() => {
   package_default = {
     name: "ai-speedometer",
-    version: "2.1.6",
+    version: "2.2.0",
     description: "A comprehensive CLI tool for benchmarking AI models across multiple providers with parallel execution and professional metrics",
     bin: {
       "ai-speedometer": "dist/ai-speedometer",
@@ -2969,6 +3078,7 @@ var init_ResultsTable = () => {};
 // src/tui/screens/BenchmarkScreen.tsx
 import { useState as useState4, useEffect as useEffect4, useRef as useRef3, useMemo } from "react";
 import { useKeyboard as useKeyboard5 } from "@opentui/react";
+import { engine, Timeline } from "@opentui/core";
 import { jsxDEV as jsxDEV10 } from "@opentui/react/jsx-dev-runtime";
 function rankBadge(rank) {
   if (rank === 1)
@@ -2990,9 +3100,23 @@ function BenchmarkScreen() {
   const navigate = useNavigate();
   const [modelStates, setModelStates] = useState4([]);
   const [spinnerFrame, setSpinnerFrame] = useState4(0);
+  const [shimmerPos, setShimmerPos] = useState4(0);
   const [allDone, setAllDone] = useState4(false);
   const spinnerRef = useRef3(null);
   const startedRef = useRef3(false);
+  useEffect4(() => {
+    const tl = new Timeline({ loop: true, duration: 1400 });
+    const target = { pos: 0 };
+    tl.add(target, { pos: 1, duration: 1400, ease: "inOutSine", onUpdate: (anim) => {
+      setShimmerPos(anim.targets[0].pos);
+    } }, 0);
+    tl.play();
+    engine.register(tl);
+    return () => {
+      tl.pause();
+      engine.unregister(tl);
+    };
+  }, []);
   useEffect4(() => {
     if (startedRef.current)
       return;
@@ -3005,9 +3129,12 @@ function BenchmarkScreen() {
     setModelStates((prev) => prev.map((s) => ({ ...s, status: "running", startedAt: Date.now() })));
     async function runAll() {
       const { benchmarkSingleModelRest: benchmarkSingleModelRest2 } = await Promise.resolve().then(() => (init_benchmark(), exports_benchmark));
+      const logEnabled = state.logMode && !!state.runId;
+      const { createBenchLogger: createBenchLogger2 } = logEnabled ? await Promise.resolve().then(() => (init_logger(), exports_logger)) : { createBenchLogger: null };
       const promises = models.map(async (model) => {
+        const logger = logEnabled && createBenchLogger2 ? await createBenchLogger2(state.runId) : undefined;
         try {
-          const result = await benchmarkSingleModelRest2(model);
+          const result = await benchmarkSingleModelRest2(model, logger);
           if (!result.success) {
             const errMsg = result.error ?? "Request failed";
             setModelStates((prev) => prev.map((s) => s.model.id === model.id && s.model.providerId === model.providerId ? { ...s, status: "error", error: errMsg } : s));
@@ -3058,6 +3185,27 @@ function BenchmarkScreen() {
       const total = modelStates.length || 1;
       const filled = Math.round((done.length + errors.length) / total * BAR_W);
       const empty = BAR_W - filled;
+      const shimmerCenter = shimmerPos * (BAR_W - 1);
+      const shimmerHalf = 1.5;
+      const barChars = Array.from({ length: BAR_W }, (_, i) => {
+        const isFilled = i < filled;
+        const dist = Math.abs(i - shimmerCenter);
+        const inWindow = dist <= shimmerHalf;
+        const intensity = inWindow ? 1 - dist / shimmerHalf : 0;
+        if (isFilled) {
+          if (intensity > 0.6)
+            return { ch: "\u2588", fg: "#c8eeff" };
+          if (intensity > 0)
+            return { ch: "\u2593", fg: "#a0d8f0" };
+          return { ch: "\u2588", fg: "#4a9abb" };
+        } else {
+          if (intensity > 0.6)
+            return { ch: "\u2591", fg: "#2a3a52" };
+          if (intensity > 0)
+            return { ch: "\u2591", fg: "#222d40" };
+          return { ch: "\u2591", fg: "#1a2030" };
+        }
+      });
       rows.push(/* @__PURE__ */ jsxDEV10("box", {
         height: 1,
         flexDirection: "row",
@@ -3076,14 +3224,10 @@ function BenchmarkScreen() {
               "  "
             ]
           }, undefined, true, undefined, this),
-          /* @__PURE__ */ jsxDEV10("text", {
-            fg: "#7dcfff",
-            children: "\u2588".repeat(filled)
-          }, undefined, false, undefined, this),
-          /* @__PURE__ */ jsxDEV10("text", {
-            fg: "#292e42",
-            children: "\u2591".repeat(empty)
-          }, undefined, false, undefined, this),
+          barChars.map((b, i) => /* @__PURE__ */ jsxDEV10("text", {
+            fg: b.fg,
+            children: b.ch
+          }, i, false, undefined, this)),
           /* @__PURE__ */ jsxDEV10("text", {
             fg: "#ff9e64",
             children: [
@@ -3419,7 +3563,7 @@ function BenchmarkScreen() {
       }, "results-empty", false, undefined, this));
     }
     return rows;
-  }, [modelStates, allDone, tpsRanked, ttftRanked, doneResults, pendingCount, maxTps, maxTtftForBar]);
+  }, [modelStates, allDone, shimmerPos, tpsRanked, ttftRanked, doneResults, pendingCount, maxTps, maxTtftForBar]);
   useKeyboard5((key) => {
     if (!allDone)
       return;
@@ -3428,10 +3572,22 @@ function BenchmarkScreen() {
       navigate("main-menu");
     }
   });
-  const statusLine = allDone ? /* @__PURE__ */ jsxDEV10("text", {
-    fg: "#9ece6a",
-    children: "All done!  [Enter]/[q] return  [\u2191\u2193/PgUp/PgDn/wheel] scroll"
-  }, undefined, false, undefined, this) : /* @__PURE__ */ jsxDEV10("box", {
+  const statusLine = allDone ? /* @__PURE__ */ jsxDEV10("box", {
+    flexDirection: "row",
+    children: [
+      /* @__PURE__ */ jsxDEV10("text", {
+        fg: "#9ece6a",
+        children: "All done!  [Enter]/[Q] return  [\u2191\u2193/PgUp/PgDn/wheel] scroll"
+      }, undefined, false, undefined, this),
+      state.logMode && state.logPath && /* @__PURE__ */ jsxDEV10("text", {
+        fg: "#565f89",
+        children: [
+          "  log: ",
+          state.logPath
+        ]
+      }, undefined, true, undefined, this)
+    ]
+  }, undefined, true, undefined, this) : /* @__PURE__ */ jsxDEV10("box", {
     flexDirection: "row",
     children: [
       running.length > 0 && /* @__PURE__ */ jsxDEV10("text", {
@@ -5100,8 +5256,9 @@ function Shell() {
     ]
   }, undefined, true, undefined, this);
 }
-function App() {
+function App({ logMode = false }) {
   return /* @__PURE__ */ jsxDEV15(AppProvider, {
+    logMode,
     children: /* @__PURE__ */ jsxDEV15(Shell, {}, undefined, false, undefined, this)
   }, undefined, false, undefined, this);
 }
@@ -5127,7 +5284,7 @@ __export(exports_tui, {
 import { createCliRenderer } from "@opentui/core";
 import { createRoot } from "@opentui/react";
 import { jsxDEV as jsxDEV16 } from "@opentui/react/jsx-dev-runtime";
-async function startTui() {
+async function startTui(logMode = false) {
   const renderer = await createCliRenderer({
     exitOnCtrlC: false
   });
@@ -5141,7 +5298,9 @@ async function startTui() {
     renderer.destroy();
     process.exit(0);
   });
-  createRoot(renderer).render(/* @__PURE__ */ jsxDEV16(App, {}, undefined, false, undefined, this));
+  createRoot(renderer).render(/* @__PURE__ */ jsxDEV16(App, {
+    logMode
+  }, undefined, false, undefined, this));
 }
 var ENABLE_BRACKETED_PASTE = "\x1B[?2004h", DISABLE_BRACKETED_PASTE = "\x1B[?2004l";
 var init_tui = __esm(() => {
@@ -5153,6 +5312,7 @@ function parseCliArgs() {
   const args = process.argv.slice(2);
   const parsed = {
     debug: false,
+    log: false,
     bench: null,
     benchCustom: null,
     apiKey: null,
@@ -5165,6 +5325,8 @@ function parseCliArgs() {
     const arg = args[i];
     if (arg === "--debug")
       parsed.debug = true;
+    else if (arg === "--log")
+      parsed.log = true;
     else if (arg === "--bench")
       parsed.bench = args[++i] ?? null;
     else if (arg === "--bench-custom")
@@ -5197,6 +5359,7 @@ function showHelp() {
   console.log("  --api-key <key>                 API key for custom provider");
   console.log("  --endpoint-format <format>      Endpoint format (default: chat/completions)");
   console.log("  --formatted                     Format JSON output for human readability");
+  console.log("  --log                           Log raw SSE streams to ~/.local/share/ai-speedometer/logs/");
   console.log("  --debug                         Enable debug logging");
   console.log("  --help, -h                      Show this help message");
   console.log("");
@@ -5214,5 +5377,5 @@ if (cliArgs.help) {
   await runHeadlessBenchmark2(cliArgs);
 } else {
   const { startTui: startTui2 } = await Promise.resolve().then(() => (init_tui(), exports_tui));
-  await startTui2();
+  await startTui2(cliArgs.log);
 }

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "ai-speedometer",
-  "version": "2.1.6",
+  "version": "2.2.0",
   "description": "A comprehensive CLI tool for benchmarking AI models across multiple providers with parallel execution and professional metrics",
   "bin": {
     "ai-speedometer": "dist/ai-speedometer",