open-agents-ai 0.16.2 → 0.16.4

package/README.md

@@ -276,6 +276,48 @@ oa --backend vllm --backend-url http://localhost:8000/v1 "add tests"
 oa --backend-url http://10.0.0.5:11434 "refactor auth"
 ```
 
+## Supported Inference Providers
+
+Open Agents auto-detects your provider from the endpoint URL and configures auth + health checks accordingly. All providers use standard `Authorization: Bearer <key>` authentication.
+
+| Provider | Endpoint URL | API Key | Notes |
+|----------|-------------|---------|-------|
+| **Ollama** (local) | `http://localhost:11434` | None | Default. Auto-detects, auto-expands context window |
+| **vLLM** (local) | `http://localhost:8000` | Optional | Self-hosted OpenAI-compatible server |
+| **LM Studio** (local) | `http://localhost:1234` | None | Local model server with GUI |
+| **Chutes AI** | `https://llm.chutes.ai` | `cpk_...` | Bearer auth. Fast cloud inference |
+| **Together AI** | `https://api.together.xyz` | Required | Large model catalog |
+| **Groq** | `https://api.groq.com/openai` | `gsk_...` | Ultra-fast LPU inference |
+| **OpenRouter** | `https://openrouter.ai/api` | `sk-or-...` | Multi-provider routing |
+| **Fireworks AI** | `https://api.fireworks.ai/inference` | `fw_...` | Fast serverless inference |
+| **DeepInfra** | `https://api.deepinfra.com` | Required | Cost-effective inference |
+| **Mistral AI** | `https://api.mistral.ai` | Required | Mistral models |
+| **Cerebras** | `https://api.cerebras.ai` | `csk-...` | Wafer-scale inference |
+| **SambaNova** | `https://api.sambanova.ai` | Required | RDU-accelerated inference |
+| **NVIDIA NIM** | `https://integrate.api.nvidia.com` | `nvapi-...` | NVIDIA cloud inference |
+| **Hyperbolic** | `https://api.hyperbolic.xyz` | Required | GPU cloud inference |
+| **OpenAI** | `https://api.openai.com` | `sk-...` | GPT models (tool calling) |
+
+### Connecting to a Provider
+
+Use `/endpoint` in the TUI or pass via CLI:
+
+```bash
+# Chutes AI
+/endpoint https://llm.chutes.ai --auth cpk_your_key_here
+
+# Groq
+/endpoint https://api.groq.com/openai --auth gsk_your_key_here
+
+# Together AI
+/endpoint https://api.together.xyz --auth your_key_here
+
+# Self-hosted vLLM on LAN
+/endpoint http://10.0.0.5:8000
+```
+
+The agent auto-detects the provider, normalizes the URL (strips `/v1/chat/completions` if pasted), tests connectivity, and saves the configuration. You can paste full endpoint URLs — they'll be cleaned up automatically.
+
 ## Evaluation Suite
 
 23 evaluation tasks test the agent's autonomous capabilities across coding, web research, SDLC analysis, and tool creation:

package/dist/index.js

@@ -497,13 +497,100 @@ var init_sleep = __esm({
 function normalizeBaseUrl(url) {
   let u = url.trim();
   u = u.replace(/\/+$/, "");
-  u = u.replace(/\/v1(?:\/.*)?$/, "");
+  u = u.replace(/\/chat\/completions$/, "");
+  u = u.replace(/\/completions$/, "");
+  u = u.replace(/\/embeddings$/, "");
+  u = u.replace(/\/models(?:\/.*)?$/, "");
+  u = u.replace(/\/+$/, "");
+  u = u.replace(/\/v1\/openai$/, "");
+  u = u.replace(/\/+$/, "");
+  u = u.replace(/\/v1$/, "");
   u = u.replace(/\/+$/, "");
   return u;
 }
+function detectProvider(url) {
+  const normalized = url.trim().toLowerCase();
+  for (const { match, info } of PROVIDERS) {
+    if (match(normalized))
+      return info;
+  }
+  const isLocal = /localhost|127\.0\.0\.1|0\.0\.0\.0/i.test(normalized);
+  return {
+    id: "unknown",
+    label: isLocal ? "Local OpenAI-compatible" : "OpenAI-compatible",
+    local: isLocal,
+    authRequired: !isLocal,
+    modelsPath: isLocal ? "/v1/models" : "/v1/models"
+  };
+}
+var PROVIDERS;
 var init_normalizeUrl = __esm({
   "packages/backend-vllm/dist/normalizeUrl.js"() {
     "use strict";
+    PROVIDERS = [
+      // --- Cloud providers (specific domains) ---
+      {
+        match: (u) => /api\.openai\.com/i.test(u),
+        info: { id: "openai", label: "OpenAI", local: false, authRequired: true, keyPrefix: "sk-", modelsPath: "/v1/models" }
+      },
+      {
+        match: (u) => /api\.together\.xyz/i.test(u),
+        info: { id: "together", label: "Together AI", local: false, authRequired: true, modelsPath: "/v1/models" }
+      },
+      {
+        match: (u) => /api\.groq\.com/i.test(u),
+        info: { id: "groq", label: "Groq", local: false, authRequired: true, keyPrefix: "gsk_", modelsPath: "/v1/models" }
+      },
+      {
+        match: (u) => /openrouter\.ai/i.test(u),
+        info: { id: "openrouter", label: "OpenRouter", local: false, authRequired: true, keyPrefix: "sk-or-", modelsPath: "/v1/models" }
+      },
+      {
+        match: (u) => /api\.fireworks\.ai/i.test(u),
+        info: { id: "fireworks", label: "Fireworks AI", local: false, authRequired: true, keyPrefix: "fw_", modelsPath: "/v1/models" }
+      },
+      {
+        match: (u) => /api\.deepinfra\.com/i.test(u),
+        info: { id: "deepinfra", label: "DeepInfra", local: false, authRequired: true, modelsPath: "/v1/models" }
+      },
+      {
+        match: (u) => /api\.mistral\.ai/i.test(u),
+        info: { id: "mistral", label: "Mistral AI", local: false, authRequired: true, modelsPath: "/v1/models" }
+      },
+      {
+        match: (u) => /llm\.chutes\.ai|chutes\.ai/i.test(u),
+        info: { id: "chutes", label: "Chutes AI", local: false, authRequired: true, keyPrefix: "cpk_", modelsPath: "/v1/models" }
+      },
+      {
+        match: (u) => /api\.cerebras\.ai/i.test(u),
+        info: { id: "cerebras", label: "Cerebras", local: false, authRequired: true, keyPrefix: "csk-", modelsPath: "/v1/models" }
+      },
+      {
+        match: (u) => /api\.sambanova\.ai/i.test(u),
+        info: { id: "sambanova", label: "SambaNova", local: false, authRequired: true, modelsPath: "/v1/models" }
+      },
+      {
+        match: (u) => /integrate\.api\.nvidia\.com/i.test(u),
+        info: { id: "nvidia", label: "NVIDIA NIM", local: false, authRequired: true, keyPrefix: "nvapi-", modelsPath: "/v1/models" }
+      },
+      {
+        match: (u) => /api\.hyperbolic\.xyz/i.test(u),
+        info: { id: "hyperbolic", label: "Hyperbolic", local: false, authRequired: true, modelsPath: "/v1/models" }
+      },
+      // --- Local providers (port-based detection) ---
+      {
+        match: (u) => /(?:localhost|127\.0\.0\.1|0\.0\.0\.0):11434/i.test(u),
+        info: { id: "ollama", label: "Ollama (local)", local: true, authRequired: false, modelsPath: "/api/tags" }
+      },
+      {
+        match: (u) => /(?:localhost|127\.0\.0\.1|0\.0\.0\.0):1234/i.test(u),
+        info: { id: "lmstudio", label: "LM Studio (local)", local: true, authRequired: false, modelsPath: "/v1/models" }
+      },
+      {
+        match: (u) => /(?:localhost|127\.0\.0\.1|0\.0\.0\.0):8000/i.test(u),
+        info: { id: "vllm", label: "vLLM (local)", local: true, authRequired: false, modelsPath: "/v1/models" }
+      }
+    ];
   }
 });
 
@@ -9094,11 +9181,11 @@ ${newerSummary}` : newerSummary;
         this.model = model;
         this.apiKey = apiKey ?? "";
       }
-      /** Build auth headers — handles Chutes (cpk_ raw), Bearer, or none. */
+      /** Build auth headers — all providers use standard Bearer token auth. */
       authHeaders() {
         const headers = { "Content-Type": "application/json" };
         if (this.apiKey) {
-          headers["Authorization"] = this.apiKey.startsWith("cpk_") ? this.apiKey : `Bearer ${this.apiKey}`;
+          headers["Authorization"] = `Bearer ${this.apiKey}`;
         }
         return headers;
       }
@@ -9820,6 +9907,84 @@ function getColorsEnabled() {
 function getTermWidth() {
   return process.stdout.columns ?? 80;
 }
+function formatMarkdownLine(line) {
+  const headingMatch = line.match(/^(#{1,6})\s+(.*)/);
+  if (headingMatch) {
+    const level = headingMatch[1].length;
+    const text = headingMatch[2];
+    const colors = [MD.heading1, MD.heading2, MD.heading3, MD.heading3, 183, 183];
+    return c2.bold(fg256(colors[level - 1] ?? 147, formatInlineMarkdown(text)));
+  }
+  if (/^[-*_]{3,}\s*$/.test(line)) {
+    const w = getTermWidth() - 10;
+    return fg256(MD.hr, "\u2500".repeat(Math.min(w, 60)));
+  }
+  if (/^>\s?/.test(line)) {
+    const content = line.replace(/^>\s?/, "");
+    return fg256(MD.blockquote, "\u2502 ") + c2.italic(fg256(MD.blockquote, formatInlineMarkdown(content)));
+  }
+  if (/^\|(.+)\|/.test(line)) {
+    if (/^\|[\s:_-]+\|/.test(line)) {
+      return fg256(MD.tableBar, line);
+    }
+    return line.replace(/([^|]+)/g, (cell) => {
+      const trimmed = cell.trim();
+      if (!trimmed)
+        return cell;
+      const leading = cell.match(/^(\s*)/)?.[1] ?? "";
+      const trailing = cell.match(/(\s*)$/)?.[1] ?? "";
+      return leading + formatInlineMarkdown(trimmed) + trailing;
+    });
+  }
+  const ulMatch = line.match(/^(\s*)([-*+])\s+(.*)/);
+  if (ulMatch) {
+    return ulMatch[1] + fg256(MD.listBullet, "\u2022") + " " + formatInlineMarkdown(ulMatch[3]);
+  }
+  const olMatch = line.match(/^(\s*)(\d+[.)])\s+(.*)/);
+  if (olMatch) {
+    return olMatch[1] + fg256(MD.listBullet, olMatch[2]) + " " + formatInlineMarkdown(olMatch[3]);
+  }
+  return formatInlineMarkdown(line);
+}
+function formatInlineMarkdown(text) {
+  let result = text;
+  result = result.replace(/`([^`]+)`/g, (_m, code) => fg256(MD.inlineCode, code));
+  result = result.replace(/\*{3}([^*]+)\*{3}/g, (_m, t) => c2.bold(c2.italic(t)));
+  result = result.replace(/\*{2}([^*]+)\*{2}/g, (_m, t) => c2.bold(t));
+  result = result.replace(/(?<!\*)\*([^*]+)\*(?!\*)/g, (_m, t) => c2.italic(t));
+  result = result.replace(/\[([^\]]+)\]\(([^)]+)\)/g, (_m, label, url) => c2.bold(fg256(MD.link, label)) + " " + c2.dim(fg256(MD.link, `(${url})`)));
+  result = result.replace(/__([^_]+)__/g, (_m, t) => c2.bold(t));
+  result = result.replace(/(?<!_)_([^_]+)_(?!_)/g, (_m, t) => c2.italic(t));
+  result = result.replace(/~~([^~]+)~~/g, (_m, t) => c2.dim(t));
+  return result;
+}
+function formatMarkdownBlock(text) {
+  const lines = text.split("\n");
+  const result = [];
+  let inCodeBlock = false;
+  let codeLang = "";
+  for (const line of lines) {
+    const trimmedLine = line.trimStart();
+    if (trimmedLine.startsWith("```")) {
+      if (inCodeBlock) {
+        result.push(c2.dim("  ```"));
+        inCodeBlock = false;
+        codeLang = "";
+      } else {
+        codeLang = trimmedLine.slice(3).trim();
+        result.push(c2.dim("  ```" + codeLang));
+        inCodeBlock = true;
+      }
+      continue;
+    }
+    if (inCodeBlock) {
+      result.push("  " + c2.dim(line));
+    } else {
+      result.push(formatMarkdownLine(line));
+    }
+  }
+  return result.join("\n");
+}
 function renderUserMessage(text) {
   process.stdout.write(`
 ${c2.bold(c2.blue("> "))}${c2.bold(text)}
@@ -9828,7 +9993,8 @@ ${c2.bold(c2.blue("> "))}${c2.bold(text)}
 function renderAssistantText(text) {
   if (!text.trim())
     return;
-  const lines = text.split("\n");
+  const formatted = formatMarkdownBlock(text);
+  const lines = formatted.split("\n");
   for (const line of lines) {
     process.stdout.write(`  ${line}
 `);
@@ -9925,8 +10091,9 @@ function renderToolResult(toolName, success, output) {
 `);
       return;
     }
-    const trimmed = line.length > maxW ? line.slice(0, maxW - 3) + "..." : line;
-    process.stdout.write(`${prefix}${highlightToolOutput(trimmed)}
+    const cropped = line.length > maxW ? line.slice(0, maxW - 3) + "..." : line;
+    const formatted = formatMarkdownLine(cropped);
+    process.stdout.write(`${prefix}${formatted === cropped ? highlightToolOutput(cropped) : formatted}
 `);
   }
   if (lines.length > maxLines) {
@@ -10010,6 +10177,9 @@ function highlightToolOutput(line) {
     return c2.green(line);
   if (/\bfail(ed|ing|ure)?\b/i.test(line))
     return c2.red(line);
+  const formatted = formatInlineMarkdown(line);
+  if (formatted !== line)
+    return formatted;
   return c2.dim(line);
 }
 function renderTaskComplete(summary, turns, toolCalls, durationMs, tokens) {
@@ -10023,7 +10193,8 @@ ${c2.green("\u2714")} ${c2.bold("Task completed")} ${c2.dim(`(${turns} turns, ${
 `);
   }
   if (summary) {
-    const lines = summary.split("\n");
+    const formatted = formatMarkdownBlock(summary);
+    const lines = formatted.split("\n");
     for (const line of lines) {
       process.stdout.write(`  ${line}
 `);
@@ -10234,13 +10405,14 @@ function renderConfig(config) {
   process.stdout.write("\n");
 }
 function formatToolArgs(toolName, args) {
+  const maxArg = Math.max(40, getTermWidth() - 20);
   switch (toolName) {
     case "file_read":
     case "file_write":
     case "file_edit":
       return String(args["path"] ?? "");
     case "shell": {
-      const cmd = truncStr(String(args["command"] ?? ""), 60);
+      const cmd = truncStr(String(args["command"] ?? ""), maxArg);
       return args["stdin"] ? `${cmd} ${c2.dim("(with stdin)")}` : cmd;
     }
     case "grep_search":
@@ -10250,21 +10422,21 @@ function formatToolArgs(toolName, args) {
     case "list_directory":
       return String(args["path"] ?? ".");
     case "web_search":
-      return `"${truncStr(String(args["query"] ?? ""), 50)}"`;
+      return `"${truncStr(String(args["query"] ?? ""), maxArg - 2)}"`;
     case "web_fetch":
-      return truncStr(String(args["url"] ?? ""), 60);
+      return truncStr(String(args["url"] ?? ""), maxArg);
     case "memory_read":
       return `${args["topic"]}${args["key"] ? "." + args["key"] : ""}`;
     case "memory_write":
       return `${args["topic"]}.${args["key"]}`;
     case "task_complete":
-      return truncStr(String(args["summary"] ?? ""), 60);
+      return truncStr(String(args["summary"] ?? ""), maxArg);
     case "aiwg_setup":
       return String(args["framework"] ?? "sdlc");
     case "aiwg_health":
       return args["detailed"] ? "detailed" : "summary";
     case "aiwg_workflow":
-      return truncStr(String(args["command"] ?? ""), 60);
+      return truncStr(String(args["command"] ?? ""), maxArg);
     case "batch_edit": {
       const edits = args["edits"];
       return edits ? `${edits.length} edit(s)` : "";
@@ -10278,14 +10450,14 @@ function formatToolArgs(toolName, args) {
     case "git_info":
       return args["show_diff"] ? "with diff" : "summary";
     case "background_run":
-      return truncStr(String(args["command"] ?? ""), 60);
+      return truncStr(String(args["command"] ?? ""), maxArg);
     case "task_status":
     case "task_output":
     case "task_stop":
       return String(args["task_id"] ?? "all");
     case "sub_agent": {
       const bg = args["background"] ? " (background)" : "";
-      return truncStr(String(args["task"] ?? ""), 50) + bg;
+      return truncStr(String(args["task"] ?? ""), maxArg - 15) + bg;
     }
     case "image_read":
       return String(args["path"] ?? "");
@@ -10296,9 +10468,9 @@ function formatToolArgs(toolName, args) {
     case "transcribe_file":
       return `${args["path"] ?? ""}${args["model"] ? ` (${args["model"]})` : ""}`;
     case "transcribe_url":
-      return truncStr(String(args["url"] ?? ""), 60);
+      return truncStr(String(args["url"] ?? ""), maxArg);
     default:
-      return Object.entries(args).map(([k, v]) => `${k}=${truncStr(String(v), 30)}`).join(", ");
+      return Object.entries(args).map(([k, v]) => `${k}=${truncStr(String(v), Math.max(30, maxArg / 3))}`).join(", ");
   }
 }
 function truncStr(s, max) {
@@ -10314,7 +10486,7 @@ function formatDuration2(ms) {
   const secs = Math.floor(totalSecs % 60);
   return `${mins}m ${secs}s`;
 }
-var isTTY2, c2, pastel, _emojisEnabled, _colorsEnabled, TOOL_ICONS, TOOL_LABELS, TOOL_COLORS, _contentWriteHook, HINTS, TOOL_NAMES, COMMAND_NAMES;
+var isTTY2, c2, pastel, _emojisEnabled, _colorsEnabled, MD, TOOL_ICONS, TOOL_LABELS, TOOL_COLORS, _contentWriteHook, HINTS, TOOL_NAMES, COMMAND_NAMES;
 var init_render = __esm({
   "packages/cli/dist/tui/render.js"() {
     "use strict";
@@ -10347,6 +10519,26 @@ var init_render = __esm({
     };
     _emojisEnabled = true;
     _colorsEnabled = true;
+    MD = {
+      heading1: 75,
+      // blue
+      heading2: 117,
+      // sky blue
+      heading3: 147,
+      // light blue
+      inlineCode: 223,
+      // light peach
+      link: 111,
+      // periwinkle
+      blockquote: 245,
+      // grey
+      hr: 240,
+      // dark grey
+      listBullet: 245,
+      // grey
+      tableBar: 245
+      // grey
+    };
     TOOL_ICONS = {
       file_read: "\u{1F4C4}",
       file_write: "\u{1F4DD}",
@@ -11054,15 +11246,18 @@ async function promptForCustomEndpoint(config, rl) {
 `);
   const modelName = await ask(rl, `  ${c2.bold("Model name")} (Enter for ${c2.dim(config.model)}): `);
   const chosenModel = modelName || config.model;
+  const provider = detectProvider(endpoint);
   process.stdout.write(`
-  ${c2.cyan("\u25CF")} Testing endpoint ${c2.bold(cleanUrl)}...
+  ${c2.cyan("\u25CF")} Detected provider: ${c2.bold(provider.label)}
+`);
+  process.stdout.write(`  ${c2.cyan("\u25CF")} Testing endpoint ${c2.bold(cleanUrl)}...
 `);
   let testOk = false;
   try {
-    const testUrl = `${cleanUrl}/v1/models`;
+    const testUrl = `${cleanUrl}${provider.modelsPath}`;
     const headers = { "Content-Type": "application/json" };
     if (apiKey) {
-      headers["Authorization"] = apiKey.startsWith("cpk_") ? apiKey : `Bearer ${apiKey}`;
+      headers["Authorization"] = `Bearer ${apiKey}`;
     }
     const resp = await fetch(testUrl, { headers, signal: AbortSignal.timeout(1e4) });
     if (resp.ok) {
@@ -11070,14 +11265,16 @@ async function promptForCustomEndpoint(config, rl) {
 `);
       testOk = true;
     } else {
-      try {
-        const ollamaResp = await fetch(`${cleanUrl}/api/tags`, { signal: AbortSignal.timeout(1e4) });
-        if (ollamaResp.ok) {
-          process.stdout.write(`  ${c2.green("\u2714")} Ollama endpoint detected.
+      if (provider.id !== "ollama") {
+        try {
+          const ollamaResp = await fetch(`${cleanUrl}/api/tags`, { signal: AbortSignal.timeout(1e4) });
+          if (ollamaResp.ok) {
+            process.stdout.write(`  ${c2.green("\u2714")} Ollama endpoint detected.
 `);
-          testOk = true;
+            testOk = true;
+          }
+        } catch {
         }
-      } catch {
       }
       if (!testOk) {
         process.stdout.write(`  ${c2.yellow("\u26A0")} Endpoint returned HTTP ${resp.status}
@@ -11089,6 +11286,10 @@ async function promptForCustomEndpoint(config, rl) {
 `);
   }
   if (!testOk) {
+    if (provider.authRequired && !apiKey) {
+      process.stdout.write(`  ${c2.dim(`${provider.label} typically requires an API key.`)}
+`);
+    }
     const startAnyway = await ask(rl, `
   ${c2.bold("Endpoint unreachable. Start anyway?")} (y/n) `);
     if (startAnyway.toLowerCase() !== "y" && startAnyway.toLowerCase() !== "yes") {
@@ -11103,11 +11304,12 @@ async function promptForCustomEndpoint(config, rl) {
   if (apiKey) {
     setConfigValue("apiKey", apiKey);
   }
-  const isLocalOllama = /localhost|127\.0\.0\.1/.test(cleanUrl) && cleanUrl.includes("11434");
-  const backendType = isLocalOllama ? "ollama" : "vllm";
+  const backendType = provider.id === "ollama" ? "ollama" : "vllm";
   setConfigValue("backendType", backendType);
   process.stdout.write(`
   ${c2.green("\u2714")} Configured: ${c2.bold(chosenModel)} at ${c2.bold(cleanUrl)}
+`);
+  process.stdout.write(`  ${c2.green("\u2714")} Provider: ${c2.bold(provider.label)}
 `);
   if (apiKey)
     process.stdout.write(`  ${c2.green("\u2714")} API key saved.
@@ -11728,9 +11930,12 @@ async function showModelPicker(ctx) {
 }
 async function handleEndpoint(arg, ctx, local = false) {
   if (!arg) {
+    const currentProvider = detectProvider(ctx.config.backendUrl);
     process.stdout.write(`
   ${c2.bold("Current endpoint:")}
 
+`);
+    process.stdout.write(`  ${c2.cyan("Provider".padEnd(12))} ${currentProvider.label}
 `);
     process.stdout.write(`  ${c2.cyan("URL".padEnd(12))} ${ctx.config.backendUrl}
 `);
@@ -11741,11 +11946,15 @@ async function handleEndpoint(arg, ctx, local = false) {
     process.stdout.write(`
   ${c2.dim("Usage: /endpoint <url> [--auth <token>]")}
 `);
-    process.stdout.write(`  ${c2.dim("  /endpoint http://localhost:11434         (Ollama, no auth)")}
+    process.stdout.write(`  ${c2.dim("  /endpoint http://localhost:11434                          Ollama")}
+`);
+    process.stdout.write(`  ${c2.dim("  /endpoint https://llm.chutes.ai --auth cpk_...           Chutes AI")}
 `);
-    process.stdout.write(`  ${c2.dim("  /endpoint http://remote:8000/v1 --auth sk-...  (OpenAI-compatible)")}
+    process.stdout.write(`  ${c2.dim("  /endpoint https://api.groq.com/openai --auth gsk_...     Groq")}
 `);
-    process.stdout.write(`  ${c2.dim("  /endpoint http://remote:8000/v1          (OpenAI-compatible, no auth)")}
+    process.stdout.write(`  ${c2.dim("  /endpoint https://api.together.xyz --auth ...             Together AI")}
+`);
+    process.stdout.write(`  ${c2.dim("  /endpoint http://localhost:8000                           vLLM")}
 
 `);
     return;
@@ -11764,15 +11973,17 @@ async function handleEndpoint(arg, ctx, local = false) {
     return;
   }
   const normalizedUrl = normalizeBaseUrl(url);
-  const isOllama = /localhost|127\.0\.0\.1/.test(url) && !url.includes("/v1") && !apiKey;
-  let backendType = isOllama ? "ollama" : "vllm";
+  const provider = detectProvider(url);
+  const backendType = provider.id === "ollama" ? "ollama" : "vllm";
   process.stdout.write(`
-  ${c2.dim("Testing connection...")} `);
+  ${c2.dim("Detected:")} ${c2.bold(provider.label)}
+`);
+  process.stdout.write(`  ${c2.dim("Testing connection...")} `);
   try {
-    const healthUrl = backendType === "ollama" ? `${normalizedUrl}/api/tags` : `${normalizedUrl}/v1/models`;
+    const healthUrl = `${normalizedUrl}${provider.modelsPath}`;
     const headers = {};
     if (apiKey) {
-      headers["Authorization"] = apiKey.startsWith("cpk_") ? apiKey : `Bearer ${apiKey}`;
+      headers["Authorization"] = `Bearer ${apiKey}`;
     }
     const resp = await fetch(healthUrl, {
       headers,
@@ -11786,6 +11997,9 @@ async function handleEndpoint(arg, ctx, local = false) {
     process.stdout.write(`${c2.yellow("\u26A0")} Could not verify
 `);
     renderWarning(`Endpoint may not be reachable: ${err instanceof Error ? err.message : String(err)}`);
+    if (provider.authRequired && !apiKey) {
+      renderInfo(`${provider.label} typically requires an API key. Use: /endpoint ${url} --auth <key>`);
+    }
     renderInfo("Setting endpoint anyway \u2014 it may come online later.");
   }
   ctx.setEndpoint(normalizedUrl, backendType, apiKey);
@@ -11803,15 +12017,17 @@ async function handleEndpoint(arg, ctx, local = false) {
   process.stdout.write(`
   ${c2.green("\u2714")} Endpoint updated and saved${local ? " (project-local)" : ""}:
 `);
-  process.stdout.write(`    ${c2.cyan("URL".padEnd(8))} ${url}
+  process.stdout.write(`    ${c2.cyan("Provider".padEnd(12))} ${provider.label}
+`);
+  process.stdout.write(`    ${c2.cyan("URL".padEnd(12))} ${normalizedUrl}
 `);
-  process.stdout.write(`    ${c2.cyan("Type".padEnd(8))} ${backendType}
+  process.stdout.write(`    ${c2.cyan("Type".padEnd(12))} ${backendType}
 `);
   if (apiKey) {
-    process.stdout.write(`    ${c2.cyan("Auth".padEnd(8))} Bearer ${apiKey.slice(0, 8)}...
+    process.stdout.write(`    ${c2.cyan("Auth".padEnd(12))} Bearer ${apiKey.slice(0, 8)}...
 `);
   } else {
-    process.stdout.write(`    ${c2.cyan("Auth".padEnd(8))} none
+    process.stdout.write(`    ${c2.cyan("Auth".padEnd(12))} ${provider.authRequired ? c2.yellow("none (may be required)") : "none"}
 `);
   }
   process.stdout.write("\n");
@@ -15587,17 +15803,19 @@ async function startInteractive(config, repoPath) {
   if (!isResumed) {
     try {
       const baseUrl = normalizeBaseUrl(config.backendUrl);
-      const healthUrl = config.backendType === "ollama" ? `${baseUrl}/api/tags` : `${baseUrl}/v1/models`;
+      const provider = detectProvider(config.backendUrl);
+      const healthUrl = `${baseUrl}${provider.modelsPath}`;
       const headers = {};
       if (config.apiKey) {
-        headers["Authorization"] = config.apiKey.startsWith("cpk_") ? config.apiKey : `Bearer ${config.apiKey}`;
+        headers["Authorization"] = `Bearer ${config.apiKey}`;
       }
       const resp = await fetch(healthUrl, { headers, signal: AbortSignal.timeout(1e4) });
       if (!resp.ok)
         throw new Error(`HTTP ${resp.status}`);
     } catch {
-      renderWarning(`Cannot reach ${config.backendType} at ${config.backendUrl}`);
-      if (config.backendType === "ollama") {
+      const provider = detectProvider(config.backendUrl);
+      renderWarning(`Cannot reach ${provider.label} at ${config.backendUrl}`);
+      if (provider.id === "ollama") {
         renderInfo("Start Ollama with: ollama serve");
       }
       renderInfo("Use /endpoint to configure a different backend. Starting anyway...");
@@ -16205,17 +16423,19 @@ async function runWithTUI(task, config, repoPath) {
   }
   try {
     const baseUrl2 = normalizeBaseUrl(config.backendUrl);
-    const healthUrl = config.backendType === "ollama" ? `${baseUrl2}/api/tags` : `${baseUrl2}/v1/models`;
+    const provider2 = detectProvider(config.backendUrl);
+    const healthUrl = `${baseUrl2}${provider2.modelsPath}`;
     const headers = {};
     if (config.apiKey) {
-      headers["Authorization"] = config.apiKey.startsWith("cpk_") ? config.apiKey : `Bearer ${config.apiKey}`;
+      headers["Authorization"] = `Bearer ${config.apiKey}`;
     }
     const resp = await fetch(healthUrl, { headers, signal: AbortSignal.timeout(1e4) });
     if (!resp.ok)
       throw new Error(`HTTP ${resp.status}`);
   } catch {
-    renderWarning(`Cannot reach ${config.backendType} at ${config.backendUrl}`);
-    if (config.backendType === "ollama") {
+    const provider2 = detectProvider(config.backendUrl);
+    renderWarning(`Cannot reach ${provider2.label} at ${config.backendUrl}`);
+    if (provider2.id === "ollama") {
       renderInfo("Start Ollama with: ollama serve");
     }
     renderInfo("The agent will retry when you submit a task. Use /endpoint to reconfigure.");

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "open-agents-ai",
-  "version": "0.16.2",
+  "version": "0.16.4",
   "description": "AI coding agent powered by open-source models (Ollama/vLLM) — interactive TUI with agentic tool-calling loop",
   "type": "module",
   "main": "./dist/index.js",