@flotorch/loadtest 0.2.1 → 0.2.3

package/README.md

@@ -41,7 +41,7 @@ This launches an interactive wizard that asks for:
 
 - **Provider adapter** — `openai` or `sagemaker` (default: `openai`)
 - **Model name** — the model identifier your endpoint expects
-- **Base URL** — API endpoint (default: `https://api.openai.com/v1/chat/completions`)
+- **Base URL** — API endpoint (default: `https://api.openai.com/v1`)
 - **Concurrency** — number of parallel requests (default: `10`)
 - **Input tokens mean** — average input token count per request (default: `512`)
 - **Output tokens mean** — average output token count per request (default: `256`)
@@ -129,7 +129,7 @@ The config file is JSON with four sections:
   "provider": {
     "adapter": "openai", // "openai" | "sagemaker"
     "model": "gpt-4o", // model identifier (required)
-    "baseURL": "https://api.openai.com/v1/chat/completions", // API endpoint
+    "baseURL": "https://api.openai.com/v1", // API endpoint
     "systemPrompt": "You are a helpful assistant.", // optional system message
     "config": {}, // backend-specific options
   },
@@ -199,7 +199,7 @@ The config file is JSON with four sections:
   "provider": {
     "adapter": "openai",
     "model": "gpt-4o",
-    "baseURL": "https://api.openai.com/v1/chat/completions"
+    "baseURL": "https://api.openai.com/v1"
   },
   "benchmark": {
     "concurrency": 20,
@@ -239,7 +239,7 @@ The config file is JSON with four sections:
   "provider": {
     "adapter": "openai",
     "model": "gpt-4o-mini",
-    "baseURL": "https://api.openai.com/v1/chat/completions"
+    "baseURL": "https://api.openai.com/v1"
   },
   "benchmark": {
     "concurrency": 10,

package/dist/index.js

@@ -455,7 +455,7 @@ var cyan = wrap("36", "39");
 var magenta = wrap("35", "39");
 
 // src/cli/args.ts
-var VERSION = true ? "0.2.1" : "dev";
+var VERSION = true ? "0.2.3" : "dev";
 var VALID_COMMANDS = /* @__PURE__ */ new Set(["run", "generate", "bench", "report", "init"]);
 var HELP_TEXT = `
 ${bold("FLOTorch Load Tester")} ${dim(`v${VERSION}`)}
@@ -628,8 +628,8 @@ async function runInit(outputPath) {
   }
   let baseURL;
   if (adapter === "openai") {
-    const url = await prompt(rl, "Base URL", "https://api.openai.com/v1/chat/completions");
-    if (url !== "https://api.openai.com/v1/chat/completions") {
+    const url = await prompt(rl, "Base URL", "https://api.openai.com/v1");
+    if (url !== "https://api.openai.com/v1") {
       baseURL = url;
     }
   }
@@ -868,11 +868,11 @@ var OpenAIBackend = class _OpenAIBackend {
   apiKey;
   static create(baseURL) {
     const env = validateEnv(EnvSchema, "openai");
-    const url = baseURL ?? "https://api.openai.com/v1/chat/completions";
+    const url = baseURL ?? "https://api.openai.com/v1";
     return new _OpenAIBackend(url, env.OPENAI_API_KEY);
   }
   constructor(baseURL, apiKey) {
-    this.url = baseURL;
+    this.url = baseURL.endsWith("/") ? baseURL : `${baseURL.replace(/\/+$/, "")}/chat/completions`;
     this.apiKey = apiKey;
   }
   async request(prompt2, model, maxTokens, systemPrompt, params, streaming, signal) {
@@ -882,12 +882,13 @@ var OpenAIBackend = class _OpenAIBackend {
     }
     messages.push({ role: "user", content: prompt2 });
     const body = {
-      model,
       messages,
       stream: streaming,
-      ...params
+      ...params,
+      model,
+      max_tokens: maxTokens
     };
-    if (body.max_tokens && this.isOpenAIHost()) {
+    if (this.isOpenAIHost()) {
       body.max_completion_tokens = body.max_tokens;
       delete body.max_tokens;
     }
@@ -1017,7 +1018,7 @@ var SageMakerBackend = class _SageMakerBackend {
   }
   constructor(config) {
     this.baseURL = config.baseURL ?? `https://runtime.sagemaker.${config.region}.amazonaws.com`;
-    this.requestFormat = config.requestFormat ?? "sagemaker" /* Sagemaker */;
+    this.requestFormat = config.requestFormat ?? "openai" /* OpenAI */;
     this.signer = new SignatureV4({
       service: "sagemaker",
       region: config.region,
@@ -1068,22 +1069,25 @@ var SageMakerBackend = class _SageMakerBackend {
     return this.parseResponse(response);
   }
   buildRequestBody(prompt2, maxTokens, systemPrompt, params, streaming) {
-    const messages = [];
-    if (systemPrompt) messages.push({ role: "system", content: systemPrompt });
-    messages.push({ role: "user", content: prompt2 });
     if (this.requestFormat === "openai" /* OpenAI */) {
+      const messages = [];
+      if (systemPrompt) messages.push({ role: "system", content: systemPrompt });
+      messages.push({ role: "user", content: prompt2 });
       return {
         messages,
-        max_tokens: maxTokens,
         stream: streaming,
-        ...params
+        ...params,
+        max_tokens: maxTokens
       };
     }
+    const rawPrompt = systemPrompt ? `${systemPrompt}
+
+${prompt2}` : prompt2;
     return {
-      inputs: [messages],
+      inputs: rawPrompt,
       parameters: {
-        max_new_tokens: maxTokens,
-        ...params
+        ...params,
+        max_new_tokens: maxTokens
       }
     };
   }
@@ -1262,7 +1266,7 @@ function createBackend(config) {
     case "openai":
       return OpenAIBackend.create(baseURL);
     case "sagemaker": {
-      const requestFormat = config.provider.config?.["requestFormat"] ?? "sagemaker" /* Sagemaker */;
+      const requestFormat = config.provider.config?.["requestFormat"] ?? "openai" /* OpenAI */;
       return SageMakerBackend.create(baseURL, requestFormat);
     }
     default:

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@flotorch/loadtest",
-  "version": "0.2.1",
+  "version": "0.2.3",
   "description": "LLM inference load testing and benchmarking tool",
   "license": "MIT",
   "repository": {