refactor(providers): move defaults and error policy into plugins

src/agents/pi-embedded-helpers/provider-error-patterns.test.ts

@@ -59,6 +59,12 @@ describe("classifyProviderSpecificError", () => {
     expect(classifyProviderSpecificError("concurrency limit reached")).toBe("rate_limit");
   });
 
+  it("classifies Cloudflare Workers AI quota errors as rate_limit", () => {
+    expect(classifyProviderSpecificError("workers_ai gateway error: quota limit exceeded")).toBe(
+      "rate_limit",
+    );
+  });
+
   it("does not match generic 'model is not ready' without Bedrock prefix", () => {
     expect(classifyProviderSpecificError("model is not ready")).toBeNull();
   });

src/agents/pi-embedded-helpers/provider-error-patterns.ts

@@ -1,11 +1,15 @@
 /**
- * Provider-specific error patterns that improve failover classification accuracy.
+ * Provider-owned error-pattern dispatch plus legacy fallback patterns.
  *
- * Many providers return errors in non-standard formats. Without these patterns,
- * errors get misclassified (e.g., a context overflow classified as "format"),
- * causing the failover engine to choose wrong recovery strategies.
+ * Most provider-specific failover classification now lives on provider-plugin
+ * hooks. This module keeps only fallback patterns for providers that do not
+ * yet ship a dedicated provider plugin hook surface.
  */
 
+import {
+  classifyProviderFailoverReasonWithPlugin,
+  matchesProviderContextOverflowWithPlugin,
+} from "../../plugins/provider-runtime.js";
 import type { FailoverReason } from "./types.js";
 
 type ProviderErrorPattern = {
@@ -21,30 +25,9 @@ type ProviderErrorPattern = {
  * to catch provider-specific wording that the generic regex misses.
  */
 export const PROVIDER_CONTEXT_OVERFLOW_PATTERNS: readonly RegExp[] = [
-  // AWS Bedrock
-  /ValidationException.*(?:input is too long|max input token|input token.*exceed)/i,
-  /ValidationException.*(?:exceeds? the (?:maximum|max) (?:number of )?(?:input )?tokens)/i,
-  /ModelStreamErrorException.*(?:Input is too long|too many input tokens)/i,
-
-  // Azure OpenAI (sometimes wraps OpenAI errors differently)
-  /content_filter.*(?:prompt|input).*(?:too long|exceed)/i,
-
-  // Ollama / local models
-  /\bollama\b.*(?:context length|too many tokens|context window)/i,
-  /\btruncating input\b.*\btoo long\b/i,
-
-  // Mistral
-  /\bmistral\b.*(?:input.*too long|token limit.*exceeded)/i,
-
-  // Cohere
+  // Cohere does not currently ship a bundled provider hook.
   /\btotal tokens?.*exceeds? (?:the )?(?:model(?:'s)? )?(?:max|maximum|limit)/i,
 
-  // DeepSeek
-  /\bdeepseek\b.*(?:input.*too long|context.*exceed)/i,
-
-  // Google Vertex / Gemini: INVALID_ARGUMENT with token-related messages is context overflow.
-  /INVALID_ARGUMENT.*(?:exceeds? the (?:maximum|max)|input.*too (?:long|large))/i,
-
   // Generic "input too long" pattern that isn't covered by existing checks
   /\binput (?:is )?too long for (?:the )?model\b/i,
 ];
@@ -55,38 +38,11 @@ export const PROVIDER_CONTEXT_OVERFLOW_PATTERNS: readonly RegExp[] = [
  * produce wrong results for specific providers.
  */
 export const PROVIDER_SPECIFIC_PATTERNS: readonly ProviderErrorPattern[] = [
-  // AWS Bedrock: ThrottlingException is rate limit
-  {
-    test: /ThrottlingException|Too many concurrent requests/i,
-    reason: "rate_limit",
-  },
-
-  // AWS Bedrock: ModelNotReadyException (require class prefix to avoid false positives)
-  {
-    test: /ModelNotReadyException/i,
-    reason: "overloaded",
-  },
-
-  // Azure: content_policy_violation should not trigger failover
-  // (it's a content moderation rejection, not a transient error)
-
-  // Groq: model_deactivated is permanent
+  // Groq does not currently ship a bundled provider hook.
   {
     test: /model(?:_is)?_deactivated|model has been deactivated/i,
     reason: "model_not_found",
   },
-
-  // Together AI / Fireworks: specific rate limit messages
-  {
-    test: /\bconcurrency limit\b.*\breached\b/i,
-    reason: "rate_limit",
-  },
-
-  // Cloudflare Workers AI
-  {
-    test: /\bworkers?_ai\b.*\b(?:rate|limit|quota)\b/i,
-    reason: "rate_limit",
-  },
 ];
 
 /**
@@ -94,7 +50,11 @@ export const PROVIDER_SPECIFIC_PATTERNS: readonly ProviderErrorPattern[] = [
  * Called from `isContextOverflowError()` to catch provider-specific wording.
  */
 export function matchesProviderContextOverflow(errorMessage: string): boolean {
-  return PROVIDER_CONTEXT_OVERFLOW_PATTERNS.some((pattern) => pattern.test(errorMessage));
+  return (
+    matchesProviderContextOverflowWithPlugin({
+      context: { errorMessage },
+    }) || PROVIDER_CONTEXT_OVERFLOW_PATTERNS.some((pattern) => pattern.test(errorMessage))
+  );
 }
 
 /**
@@ -102,6 +62,12 @@ export function matchesProviderContextOverflow(errorMessage: string): boolean {
  * Returns null if no provider-specific pattern matches (fall through to generic classification).
  */
 export function classifyProviderSpecificError(errorMessage: string): FailoverReason | null {
+  const pluginReason = classifyProviderFailoverReasonWithPlugin({
+    context: { errorMessage },
+  });
+  if (pluginReason) {
+    return pluginReason;
+  }
   for (const pattern of PROVIDER_SPECIFIC_PATTERNS) {
     if (pattern.test.test(errorMessage)) {
       return pattern.reason;