openclaw-freerouter 2.1.0 → 2.1.1

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "openclaw-freerouter",
-  "version": "2.1.0",
+  "version": "2.1.1",
   "description": "Smart LLM router plugin for OpenClaw — classify requests and route to the best model using your own API keys. 14-dimension scoring, <1ms classification, per-prompt/session model switching.",
   "type": "module",
   "openclaw": {

package/src/provider.ts

@@ -12,6 +12,8 @@ import { join } from "node:path";
 import { homedir } from "node:os";
 import type { ServerResponse } from "node:http";
 
+const AUTH_PROFILES_PATH = join(homedir(), ".openclaw", "agents", "main", "agent", "auth-profiles.json");
+
 // ─── Timeout Configuration ───
 const TIER_TIMEOUTS: Record<string, number> = {
   SIMPLE: 30_000,
@@ -29,6 +31,15 @@ export class TimeoutError extends Error {
   }
 }
 
+export type PreflightResult = {
+  ok: boolean;
+  issues: Array<{
+    provider: string;
+    reason: string;
+    model?: string;
+  }>;
+};
+
 // ─── Provider Config ───
 
 type ProviderDef = {
@@ -58,9 +69,8 @@ type AuthProfiles = {
 };
 
 function loadAuthProfiles(): Map<string, { token?: string; apiKey?: string }> {
-  const filePath = join(homedir(), ".openclaw", "agents", "main", "agent", "auth-profiles.json");
   try {
-    const data: AuthProfiles = JSON.parse(readFileSync(filePath, "utf-8"));
+    const data: AuthProfiles = JSON.parse(readFileSync(AUTH_PROFILES_PATH, "utf-8"));
     const map = new Map<string, { token?: string; apiKey?: string }>();
     const lastGood = data.lastGood ?? {};
 
@@ -100,8 +110,9 @@ export function createForwarder(openclawConfig: any, logger: any) {
   function getProviderDef(provider: string): ProviderDef | undefined {
     // Check OpenClaw config for custom provider definitions
     const ocProviders = (openclawConfig as any)?.providers;
-    if (ocProviders?.[provider]) {
-      const p = ocProviders[provider];
+    const modelsProviders = (openclawConfig as any)?.models?.providers;
+    const p = ocProviders?.[provider] ?? modelsProviders?.[provider];
+    if (p) {
       return {
         baseUrl: p.baseUrl ?? DEFAULT_PROVIDERS[provider]?.baseUrl,
         api: p.api === "anthropic-messages" ? "anthropic" : (p.api ?? DEFAULT_PROVIDERS[provider]?.api ?? "openai"),
@@ -111,6 +122,38 @@ export function createForwarder(openclawConfig: any, logger: any) {
     return DEFAULT_PROVIDERS[provider];
   }
 
+  function preflightModels(modelIds: string[]): PreflightResult {
+    const seen = new Set<string>();
+    const issues: PreflightResult["issues"] = [];
+
+    for (const modelId of modelIds) {
+      const parsed = parseModelId(modelId);
+      if (seen.has(parsed.provider)) continue;
+      seen.add(parsed.provider);
+
+      const providerDef = getProviderDef(parsed.provider);
+      if (!providerDef) {
+        issues.push({
+          provider: parsed.provider,
+          reason: `Unsupported provider: ${parsed.provider}`,
+          model: modelId,
+        });
+        continue;
+      }
+
+      const auth = getAuth(parsed.provider);
+      if (!auth?.token && !auth?.apiKey) {
+        issues.push({
+          provider: parsed.provider,
+          model: modelId,
+          reason: `No API key found for provider \"${parsed.provider}\". Auth store: ${AUTH_PROFILES_PATH}`,
+        });
+      }
+    }
+
+    return { ok: issues.length === 0, issues };
+  }
+
   function getThinkingConfig(thinkingMode: string, _modelId: string): { type: string; budget_tokens?: number } | undefined {
     if (thinkingMode === "adaptive") return { type: "adaptive" };
     const budgetMatch = thinkingMode.match(/^enabled\((\d+)\)$/);
@@ -452,7 +495,15 @@ export function createForwarder(openclawConfig: any, logger: any) {
       messages: req.messages,
       stream,
     };
-    if (req.max_tokens) body.max_tokens = req.max_tokens;
+    // GPT-5.x and newer OpenAI models require max_completion_tokens instead of max_tokens
+    const isNewOpenAI = provider === "openai" && /^gpt-[5-9]/.test(modelName);
+    if (req.max_tokens || req.max_completion_tokens) {
+      if (isNewOpenAI) {
+        body.max_completion_tokens = req.max_completion_tokens ?? req.max_tokens;
+      } else {
+        body.max_tokens = req.max_tokens ?? req.max_completion_tokens;
+      }
+    }
     if (req.temperature !== undefined) body.temperature = req.temperature;
     if (req.top_p !== undefined) body.top_p = req.top_p;
 
@@ -546,6 +597,7 @@ export function createForwarder(openclawConfig: any, logger: any) {
   // ─── Main forward function ───
 
   return {
+    preflight: preflightModels,
     forward: async (chatReq: any, routedModel: string, tier: string, thinkingMode: string, res: ServerResponse, stream: boolean) => {
       const { provider, model } = parseModelId(routedModel);
       const providerDef = getProviderDef(provider);

package/src/service.ts

@@ -403,6 +403,17 @@ export function createProxyServer(options: ProxyOptions): { server: Server; stat
     byModel: {},
   };
 
+  function preflightHealth() {
+    const modelSet = new Set<string>(["freerouter/auto"]);
+    for (const tier of Object.values(routingConfig.tiers)) {
+      if (tier?.primary) modelSet.add(tier.primary);
+      for (const fb of tier?.fallback ?? []) {
+        modelSet.add(fb);
+      }
+    }
+    return forwarder.preflight(Array.from(modelSet));
+  }
+
   async function handleChatCompletions(req: IncomingMessage, res: ServerResponse) {
     const bodyStr = await readBody(req);
     let chatReq: any;
@@ -660,6 +671,22 @@ export function createProxyServer(options: ProxyOptions): { server: Server; stat
       }
     }
 
+    // Preflight check: ensure auth/config for every provider we might hit.
+    const preflightModels = [requestedModel, ...modelsToTry];
+    const preflight = forwarder.preflight(preflightModels);
+    if (!preflight.ok) {
+      const issue = preflight.issues[0];
+      if (!issue) {
+        sendError(res, 503, "Freerouter preflight failed", "auth_error");
+      } else {
+        const details = preflight.issues
+          .map((i) => `${i.provider}: ${i.reason}`)
+          .join("; ");
+        sendError(res, 503, `Freerouter preflight failed: ${details}`, "auth_error");
+      }
+      return;
+    }
+
     let lastError = "";
     for (const modelToTry of modelsToTry) {
       try {
@@ -669,7 +696,15 @@ export function createProxyServer(options: ProxyOptions): { server: Server; stat
         }
 
         // KEY FIX: Forward with the REAL model name — not "freerouter/X"
-        await forwarder.forward(chatReq, modelToTry, tier, thinkingMode, res, stream);
+        // Inject model identity into messages so the agent knows which model it's running on
+        const injectedMessages = [...chatReq.messages];
+        const modelHint = `[FreeRouter] You are running on model: ${modelToTry} | Tier: ${tier} | Thinking: ${thinkingMode}`;
+        // Insert as a developer message right after system prompts (before user messages)
+        const lastSystemIdx = injectedMessages.reduce((acc: number, m: any, i: number) =>
+          (m.role === "system" || m.role === "developer") ? i : acc, -1);
+        injectedMessages.splice(lastSystemIdx + 1, 0, { role: "developer", content: modelHint });
+        const injectedReq = { ...chatReq, messages: injectedMessages };
+        await forwarder.forward(injectedReq, modelToTry, tier, thinkingMode, res, stream);
         return;
       } catch (err: any) {
         lastError = err.message ?? String(err);
@@ -712,8 +747,21 @@ export function createProxyServer(options: ProxyOptions): { server: Server; stat
   }
 
   function handleHealth(_req: IncomingMessage, res: ServerResponse) {
-    res.writeHead(200, { "Content-Type": "application/json" });
-    res.end(JSON.stringify({ status: "ok", version: "2.0.0", uptime: process.uptime(), stats }));
+    const pf = preflightHealth();
+    const status = pf.ok ? "ok" : "degraded";
+    const payload: Record<string, unknown> = {
+      status,
+      version: "2.0.0",
+      uptime: process.uptime(),
+      stats,
+    };
+
+    if (!pf.ok) {
+      payload.issues = pf.issues;
+    }
+
+    res.writeHead(pf.ok ? 200 : 503, { "Content-Type": "application/json" });
+    res.end(JSON.stringify(payload));
   }
 
   function handleStats(_req: IncomingMessage, res: ServerResponse) {