clauderipple 0.8.9 → 0.9.0

package/dist/router/src/proxy.js

@@ -18,7 +18,7 @@ import https from "node:https";
 import net from "node:net";
 import tls from "node:tls";
 import zlib from "node:zlib";
-import { UpstreamHealth } from "./health.js";
+import { ProviderHealth, providerExhaustion, UpstreamHealth } from "./health.js";
 import { BOOTSTRAP_PATH, injectBootstrap } from "./bootstrap.js";
 import { THREAD_UNSUPPORTED, effortOf, resolve, rewriteBody, stripThreadFields, threadDecision, unroutableReason } from "./routing.js";
 import { defaultAgentDir, withAgentAliases } from "./agents.js";
@@ -32,7 +32,7 @@ import { OpenAiCompatibleAdapter } from "./providers/openai/index.js";
 import { GoogleAdapter } from "./providers/google/index.js";
 import { conversationKey } from "./providers/chatgpt/translate.js";
 import { providerFor, searchProviderFor, terminateHosts } from "./config.js";
-import { injectPickerModels, isBootstrapPath } from "./picker.js";
+import { chatPickerModels, injectPickerModels, isBootstrapPath } from "./picker.js";
 import { effortLevelsForModel } from "./admin.js";
 import { ResponseUsageTap } from "./requestlog.js";
 import { credentialHeaderValues, redactErrorText, redactHeaders } from "./redact.js";
@@ -198,6 +198,7 @@ export class Proxy {
      * its quarantine again on the first request.
      */
     pool = new CredentialPool();
+    providerHealth = new ProviderHealth();
     /** Latest rate-limit snapshot reported by any chatgpt provider (for the admin GUI). */
     get chatgptRateLimits() {
         const out = {};
@@ -784,6 +785,8 @@ export class Proxy {
                     // The adapter reports each account's outcome to the shared pool itself, so a slot pointing
                     // here fails over once every account is out (chooseTarget asks the adapter).
                     const o = await this.chatgpt(route.provider, provider).handle(req, res, path, json, route.model, effortOf(json));
+                    if (!path.startsWith("/v1/messages/count_tokens"))
+                        this.providerHealth.observe(route.provider, o.status, o.exhaustion);
                     finish(String(o.status), o.bytes, o.note, o.status >= 400, { ...(o.usage ? { usage: o.usage } : {}), ...(o.stopReason ? { stopReason: o.stopReason } : {}) });
                 }
                 catch (e) {
@@ -813,6 +816,8 @@ export class Proxy {
                 try {
                     const o = await this.openai(route.provider, provider).handle(req, res, path, json, route.model, routeEffort);
                     this.recordOutcome(route.provider, o.status);
+                    if (!path.startsWith("/v1/messages/count_tokens"))
+                        this.providerHealth.observe(route.provider, o.status, o.exhaustion);
                     finish(String(o.status), o.bytes, o.note, o.status >= 400, { ...(o.usage ? { usage: o.usage } : {}), ...(o.stopReason ? { stopReason: o.stopReason } : {}) });
                 }
                 catch (e) {
@@ -845,6 +850,8 @@ export class Proxy {
                 try {
                     const o = await this.google(route.provider, provider).handle(req, res, path, json, route.model, routeEffort);
                     this.recordOutcome(route.provider, o.status);
+                    if (!path.startsWith("/v1/messages/count_tokens"))
+                        this.providerHealth.observe(route.provider, o.status, o.exhaustion);
                     finish(String(o.status), o.bytes, o.note, o.status >= 400, { ...(o.usage ? { usage: o.usage } : {}), ...(o.stopReason ? { stopReason: o.stopReason } : {}) });
                 }
                 catch (e) {
@@ -1274,7 +1281,7 @@ export class Proxy {
                     else if (status === 200 && isPickerBootstrap) {
                         try {
                             const j = JSON.parse(out.toString("utf8"));
-                            const r = injectPickerModels(j, cfg.cli.extraModels, cfg.cli.autoCompactWindow, (model) => effortLevelsForModel(cfg, model));
+                            const r = injectPickerModels(j, chatPickerModels(cfg), cfg.cli.autoCompactWindow, (model) => effortLevelsForModel(cfg, model));
                             if (r.surfaces.length > 0) {
                                 this.lastPickerInjection = { at: new Date().toISOString(), ...r };
                                 out = Buffer.from(JSON.stringify(j));
@@ -1337,6 +1344,10 @@ export class Proxy {
                 observedUsage = observed.usage;
                 observedStopReason = observed.stopReason;
                 res.end();
+                const errorBody = status >= 400 ? decodeBodyToText(Buffer.concat(errorHead), upRes.headers["content-encoding"]) : "";
+                if (route && path.startsWith("/v1/messages") && !path.startsWith("/v1/messages/count_tokens")) {
+                    this.providerHealth.observe(route.provider, status, providerExhaustion(status, errorBody, retryAfterMs(upRes.headers)));
+                }
                 const note = status >= 400 ? `upstream ${status}: ${errorSnippet(Buffer.concat(errorHead), upRes.headers["content-encoding"], [...errorSecrets], upRes.headers["content-type"])}` : undefined;
                 // Anthropic forgets a conversation's server-side thread after a few idle minutes, and the CLI
                 // answers the 404 by replaying the whole conversation, which hits the prompt cache. From CLI

package/dist/router/src/version.js

@@ -5,4 +5,4 @@
 // from the manifests. Until 0.1.1 the router and the CLI each carried their own literal, and both
 // still said "0.1.0" in the 0.1.1 release: after an update nobody could tell which router was
 // running, and the tray app had nothing to compare (2026-09-15, reported from a Windows install).
-export const VERSION = "0.8.9";
+export const VERSION = "0.9.0";