@code-yeongyu/senpi-ai 2026.9.29 → 2026.9.30

package/dist/utils/prompt-cache-ttl.js

@@ -323,6 +323,38 @@ function hasOpenAIExtendedPromptCache(model) {
     }
     return OPENAI_EXTENDED_CACHE_MODEL_ID.test(model.id);
 }
+/** Env that puts Claude Code on API-key, gateway or cloud billing, where it caches the main conversation for 5 minutes. */
+const CLAUDE_CODE_API_BILLING_ENV = [
+    "ANTHROPIC_API_KEY",
+    "ANTHROPIC_AUTH_TOKEN",
+    "ANTHROPIC_BASE_URL",
+    "CLAUDE_CODE_USE_BEDROCK",
+    "CLAUDE_CODE_USE_VERTEX",
+    "CLAUDE_CODE_USE_FOUNDRY",
+];
+function isEnabledFlag(value) {
+    return value !== undefined && !/^(?:|0|false|no|off)$/i.test(value.trim());
+}
+/**
+ * The Claude SDK lane's cache TTL is chosen by Claude Code, not senpi: `CLAUDE_CODE_PROMPT_CACHE_TTL`
+ * (`5m` | `1h`) wins, then `FORCE_PROMPT_CACHING_5M` and `ENABLE_PROMPT_CACHING_1H`; otherwise a Claude
+ * subscription gets 1 hour and API-key, gateway or cloud billing gets 5 minutes. A subscription past its
+ * usage limits also drops to 5 minutes, which nothing here can observe.
+ */
+function claudeCodePromptCacheTtlSeconds(env) {
+    const explicit = getProviderEnvValue("CLAUDE_CODE_PROMPT_CACHE_TTL", env)?.trim().toLowerCase();
+    if (explicit === "5m")
+        return PROMPT_CACHE_TTL_SHORT_SECONDS;
+    if (explicit === "1h")
+        return PROMPT_CACHE_TTL_LONG_SECONDS;
+    if (isEnabledFlag(getProviderEnvValue("FORCE_PROMPT_CACHING_5M", env)))
+        return PROMPT_CACHE_TTL_SHORT_SECONDS;
+    if (isEnabledFlag(getProviderEnvValue("ENABLE_PROMPT_CACHING_1H", env)))
+        return PROMPT_CACHE_TTL_LONG_SECONDS;
+    return CLAUDE_CODE_API_BILLING_ENV.some((name) => getProviderEnvValue(name, env) !== undefined)
+        ? PROMPT_CACHE_TTL_SHORT_SECONDS
+        : PROMPT_CACHE_TTL_LONG_SECONDS;
+}
 /**
  * Classify the active model's prompt-cache lifetime from the provider's documented cache contract.
  * `cacheRetention: "none"` (or `PI_CACHE_RETENTION` resolving to it) always wins.
@@ -330,8 +362,7 @@ function hasOpenAIExtendedPromptCache(model) {
 export function resolvePromptCacheLifetime(model, env) {
     switch (model.api) {
         case "claude-sdk-oauth":
-            // The Claude SDK owns prompt caching for this lane and uses Anthropic's default 5m TTL.
-            return ttl(PROMPT_CACHE_TTL_SHORT_SECONDS);
+            return ttl(claudeCodePromptCacheTtlSeconds(env));
         case "anthropic-messages": {
             const anthropicModel = model;
             const retention = resolveAnthropicCacheRetention(anthropicModel.cacheRetention, env, "short");

package/dist/utils/provider-failure-description.d.ts

@@ -16,4 +16,11 @@ export declare function stripTurnRetrySuppressionPrefix(message: string): string
  * the caller can fall back to the (marker-stripped) raw text.
  */
 export declare function describeProviderFailureForUser(errorMessage: string | undefined, options?: ProviderStallDescriptionOptions): string | undefined;
+/**
+ * The turn-ending message for a provider module that can no longer be imported because the
+ * installed package was replaced under the running process (#2358), or undefined for any other
+ * failure. It carries the no-turn-retry marker: every model in a fallback chain needs a module from
+ * the same install, so a retry or fallback would only repeat the failure.
+ */
+export declare function describeReplacedInstall(error: unknown): string | undefined;
 //# sourceMappingURL=provider-failure-description.d.ts.map

package/dist/utils/provider-failure-description.js

@@ -60,4 +60,25 @@ export function describeProviderFailureForUser(errorMessage, options = {}) {
         sentences.push(RECOVERY_SENTENCE[options.recovery]);
     return sentences.join(" ");
 }
+// Node: "Cannot find module '/…/chunks/x-HASH.js' imported from …"; Bun: the same with a relative
+// specifier, or "ENOENT reading \"/…/chunks/x-HASH.js\"" when the file vanished after resolution.
+const MISSING_MODULE_PATTERNS = [/^Cannot find module ['"]([^'"]+)['"]/, /^ENOENT reading ['"]([^'"]+)['"]/];
+const SHIPPED_MODULE = /^(?:\.{1,2}[\\/]|[\\/]|[A-Za-z]:[\\/]|file:).*\.m?js$/;
+/**
+ * The turn-ending message for a provider module that can no longer be imported because the
+ * installed package was replaced under the running process (#2358), or undefined for any other
+ * failure. It carries the no-turn-retry marker: every model in a fallback chain needs a module from
+ * the same install, so a retry or fallback would only repeat the failure.
+ */
+export function describeReplacedInstall(error) {
+    if (!(error instanceof Error))
+        return undefined;
+    for (const pattern of MISSING_MODULE_PATTERNS) {
+        const missing = pattern.exec(error.message)?.[1];
+        if (missing !== undefined && SHIPPED_MODULE.test(missing)) {
+            return `${TURN_RETRY_SUPPRESSION_PREFIX}The installed package changed while this session was running, so ${missing.split(/[\\/]/).at(-1)} can no longer be loaded. Restart and resume this session to continue.`;
+        }
+    }
+    return undefined;
+}
 //# sourceMappingURL=provider-failure-description.js.map

package/dist/utils/retry.js

@@ -179,6 +179,15 @@ const RETRYABLE_PROVIDER_ERROR_PATTERN = buildProviderErrorPattern([
     escapeRegExp(FORWARDED_EMPTY_TOOL_USE_ERROR),
     // gRPC based providers (e.g. NVIDIA NIM)
     "ResourceExhausted",
+    // Claude subscription per-request rejection that names no policy reason:
+    // `{"type":"error","error":{"type":"forbidden","message":"Request not allowed"}}`
+    // (senpi#2376). The same credential answered neighbouring requests with 200 and
+    // the burst ended by itself, so a bounded same-model retry recovers where an
+    // immediate fallback hop stranded the session. Anchored on the exact `forbidden`
+    // type plus the reason-less message, in either field order, so permission_error
+    // and forbidden rejections that carry a reason stay terminal.
+    '"type"\\s*:\\s*"forbidden"\\s*,\\s*"message"\\s*:\\s*"Request not allowed\\.?"',
+    '"message"\\s*:\\s*"Request not allowed\\.?"\\s*,\\s*"type"\\s*:\\s*"forbidden"',
     // Claude Agent SDK session.json lock contention. A second stream/resume
     // hits proper-lockfile while the previous subprocess still holds the file.
     // Same-process retry recovers; hopping providers cannot release that lock.

package/dist/utils/tool-schema-compat.d.ts

@@ -24,4 +24,6 @@ export declare function normalizeToolParametersForMoonshot(schema: Record<string
  * Schemas that are already plain objects are returned untouched.
  */
 export declare function resolveRootObjectSchema(schema: Record<string, unknown>): Record<string, unknown>;
+/** Bedrock Converse rejects root combiners, including allOf, even with an object type (#1947). */
+export declare function normalizeToolParametersForBedrock(schema: Record<string, unknown>): Record<string, unknown>;
 //# sourceMappingURL=tool-schema-compat.d.ts.map

package/dist/utils/tool-schema-compat.js

@@ -92,9 +92,10 @@ function collapseConstUnion(node) {
  * so the root's own `properties` and `required` are merged with the branches'
  * rather than replaced by them.
  */
-function mergeRootObjectUnion(schema) {
-    const branches = Array.isArray(schema.anyOf) ? schema.anyOf : Array.isArray(schema.oneOf) ? schema.oneOf : undefined;
-    if (branches === undefined || branches.length === 0)
+function mergeRootObjectUnion(schema, combiner) {
+    const selected = combiner ?? (Array.isArray(schema.anyOf) ? "anyOf" : "oneOf");
+    const branches = schema[selected];
+    if (!Array.isArray(branches) || branches.length === 0)
         return undefined;
     if (schema.properties !== undefined && !isJsonObject(schema.properties))
         return undefined;
@@ -121,12 +122,11 @@ function mergeRootObjectUnion(schema) {
             properties[name] =
                 existing === undefined || JSON.stringify(existing) === JSON.stringify(propertySchema)
                     ? propertySchema
-                    : { anyOf: [existing, propertySchema] };
+                    : { [selected === "allOf" ? "allOf" : "anyOf"]: [existing, propertySchema] };
         }
     }
-    // Only names required by EVERY branch stay required; a name required by one
-    // branch alone would reject payloads the union accepts. Root-level `required`
-    // applies to all branches, so it is unioned back in.
+    // Unions require only branch-common names; intersections require every branch's
+    // names. Root-level required always applies.
     const rootRequired = Array.isArray(schema.required)
         ? schema.required.filter((name) => typeof name === "string")
         : [];
@@ -137,8 +137,15 @@ function mergeRootObjectUnion(schema) {
     const commonBranchRequired = firstBranchRequired
         ? [...firstBranchRequired].filter((name) => branchRequiredSets.every((names) => names.has(name)))
         : [];
-    const required = [...new Set([...rootRequired, ...commonBranchRequired])];
-    const { anyOf: _anyOf, oneOf: _oneOf, ...rest } = schema;
+    const required = [
+        ...new Set([
+            ...rootRequired,
+            ...(selected === "allOf" ? branchRequiredSets.flatMap((names) => [...names]) : commonBranchRequired),
+        ]),
+    ];
+    const rest = { ...schema };
+    for (const key of combiner ? [combiner] : ["anyOf", "oneOf"])
+        delete rest[key];
     return {
         ...rest,
         type: "object",
@@ -230,6 +237,18 @@ export function normalizeToolParametersForMoonshot(schema) {
 export function resolveRootObjectSchema(schema) {
     return mergeRootObjectUnion(structuredClone(schema)) ?? schema;
 }
+/** Bedrock Converse rejects root combiners, including allOf, even with an object type (#1947). */
+export function normalizeToolParametersForBedrock(schema) {
+    let normalized = structuredClone(schema);
+    for (const combiner of COMBINER_KEYS) {
+        const branches = normalized[combiner];
+        if (Array.isArray(branches)) {
+            normalized[combiner] = branches.map((branch) => isJsonObject(branch) ? normalizeToolParametersForBedrock(branch) : branch);
+        }
+        normalized = mergeRootObjectUnion(normalized, combiner) ?? normalized;
+    }
+    return { type: "object", ...normalized };
+}
 function stripMoonshotAnnotations(node) {
     if (Array.isArray(node)) {
         return node.map((child) => stripMoonshotAnnotations(child));

package/package.json

@@ -1,6 +1,6 @@
 {
 	"name": "@code-yeongyu/senpi-ai",
-	"version": "2026.9.29",
+	"version": "2026.9.30",
 	"description": "Unified LLM API with automatic model discovery and provider configuration",
 	"type": "module",
 	"main": "./dist/index.js",
@@ -65,6 +65,7 @@
 	},
 	"files": [
 		"dist",
+		"!dist/**/*.map",
 		"README.md"
 	],
 	"scripts": {
@@ -84,7 +85,7 @@
 	"dependencies": {
 		"@anthropic-ai/sdk": "0.127.0",
 		"@aws-sdk/client-bedrock-runtime": "3.1136.0",
-		"@earendil-works/pi-telemetry": "npm:@code-yeongyu/senpi-telemetry@2026.9.29",
+		"@earendil-works/pi-telemetry": "npm:@code-yeongyu/senpi-telemetry@2026.9.30",
 		"@google/genai": "2.23.0",
 		"@smithy/node-http-handler": "4.12.1",
 		"http-proxy-agent": "9.1.0",

package/dist/api/anthropic-messages.d.ts.map

@@ -1 +0,0 @@
-{"version":3,"file":"anthropic-messages.d.ts","sourceRoot":"","sources":["../../src/api/anthropic-messages.ts"],"names":[],"mappings":"AAAA,OAAO,SAAS,MAAM,mBAAmB,CAAC;AAC1C,OAAO,KAAK,EAMX,+BAA+B,EAK/B,MAAM,uDAAuD,CAAC;AAI/D,OAAO,KAAK,EACX,wBAAwB,EAKxB,OAAO,EAGP,KAAK,EAGL,mBAAmB,EAEnB,cAAc,EACd,aAAa,EAGb,IAAI,EAIJ,MAAM,aAAa,CAAC;AAwDrB,OAAO,EAAE,kBAAkB,EAAE,MAAM,8BAA8B,CAAC;AA6ElE,eAAO,MAAM,gBAAgB,SAAU,MAAM,WAC2B,CAAC;AACzE,eAAO,MAAM,kBAAkB,SAAU,MAAM,UAAU,IAAI,EAAE,WAS9D,CAAC;AA4EF,MAAM,MAAM,eAAe,GAAG,KAAK,GAAG,QAAQ,GAAG,MAAM,GAAG,OAAO,GAAG,KAAK,CAAC;AAE1E,MAAM,MAAM,wBAAwB,GAAG,YAAY,GAAG,SAAS,CAAC;AAsGhE,MAAM,WAAW,gBAAiB,SAAQ,aAAa;IACtD;;;;;;OAMG;IACH,eAAe,CAAC,EAAE,OAAO,CAAC;IAC1B;;;;OAIG;IACH,oBAAoB,CAAC,EAAE,MAAM,CAAC;IAC9B;;;;;;;;;;;OAWG;IACH,MAAM,CAAC,EAAE,eAAe,CAAC;IACzB;;;;;;;;;;;OAWG;IACH,eAAe,CAAC,EAAE,wBAAwB,CAAC;IAC3C;;;;;OAKG;IACH,mBAAmB,CAAC,EAAE,OAAO,CAAC;IAC9B,wEAAwE;IACxE,gBAAgB,CAAC,EAAE,wBAAwB,CAAC;IAC5C;;;;OAIG;IACH,UAAU,CAAC,EAAE,MAAM,GAAG,KAAK,GAAG,MAAM,GAAG;QAAE,IAAI,EAAE,MAAM,CAAC;QAAC,IAAI,EAAE,MAAM,CAAA;KAAE,CAAC;IACtE;;;;OAIG;IACH,MAAM,CAAC,EAAE,SAAS,CAAC;CACnB;AAmvBD,eAAO,MAAM,MAAM,EAAE,cAAc,CAAC,oBAAoB,EAAE,gBAAgB,CA0gBzE,CAAC;AAyFF,eAAO,MAAM,YAAY,EAAE,cAAc,CAAC,oBAAoB,EAAE,mBAAmB,CA+ClF,CAAC;AAoJF,wBAAgB,mCAAmC,CAClD,KAAK,EAAE,KAAK,CAAC,oBAAoB,CAAC,EAClC,OAAO,EAAE,OAAO,EAChB,OAAO,CAAC,EAAE,gBAAgB,GACxB,+BAA+B,CAkBjC"}