@poolzin/pool-bot 2026.2.23 → 2026.2.25

package/dist/agents/model-fallback.js

@@ -1,64 +1,75 @@
+import { ensureAuthProfileStore, getSoonestCooldownExpiry, isProfileInCooldown, resolveAuthProfileOrder, } from "./auth-profiles.js";
 import { DEFAULT_MODEL, DEFAULT_PROVIDER } from "./defaults.js";
 import { coerceToFailoverError, describeFailoverError, isFailoverError, isTimeoutError, } from "./failover-error.js";
-import { buildModelAliasIndex, modelKey, parseModelRef, resolveConfiguredModelRef, resolveModelRefFromString, } from "./model-selection.js";
-import { ensureAuthProfileStore, getSoonestCooldownExpiry, isProfileInCooldown, resolveAuthProfileOrder, } from "./auth-profiles.js";
-import { isProviderRateLimited, recordFallbackError } from "./provider/integration.js";
-function isAbortError(err) {
-    if (!err || typeof err !== "object")
+import { buildConfiguredAllowlistKeys, buildModelAliasIndex, modelKey, normalizeModelRef, resolveConfiguredModelRef, resolveModelRefFromString, } from "./model-selection.js";
+import { isLikelyContextOverflowError } from "./pi-embedded-helpers.js";
+/**
+ * Fallback abort check. Only treats explicit AbortError names as user aborts.
+ * Message-based checks (e.g., "aborted") can mask timeouts and skip fallback.
+ */
+function isFallbackAbortError(err) {
+    if (!err || typeof err !== "object") {
         return false;
-    if (isFailoverError(err))
+    }
+    if (isFailoverError(err)) {
         return false;
+    }
     const name = "name" in err ? String(err.name) : "";
-    // Only treat explicit AbortError names as user aborts.
-    // Message-based checks (e.g., "aborted") can mask timeouts and skip fallback.
     return name === "AbortError";
 }
 function shouldRethrowAbort(err) {
-    return isAbortError(err) && !isTimeoutError(err);
+    return isFallbackAbortError(err) && !isTimeoutError(err);
 }
-function buildAllowedModelKeys(cfg, defaultProvider) {
-    const rawAllowlist = (() => {
-        const modelMap = cfg?.agents?.defaults?.models ?? {};
-        return Object.keys(modelMap);
-    })();
-    if (rawAllowlist.length === 0)
-        return null;
-    const keys = new Set();
-    for (const raw of rawAllowlist) {
-        const parsed = parseModelRef(String(raw ?? ""), defaultProvider);
-        if (!parsed)
-            continue;
-        keys.add(modelKey(parsed.provider, parsed.model));
-    }
-    return keys.size > 0 ? keys : null;
-}
-function resolveImageFallbackCandidates(params) {
-    const aliasIndex = buildModelAliasIndex({
-        cfg: params.cfg ?? {},
-        defaultProvider: params.defaultProvider,
-    });
-    const allowlist = buildAllowedModelKeys(params.cfg, params.defaultProvider);
+function createModelCandidateCollector(allowlist) {
     const seen = new Set();
     const candidates = [];
     const addCandidate = (candidate, enforceAllowlist) => {
-        if (!candidate.provider || !candidate.model)
+        if (!candidate.provider || !candidate.model) {
             return;
+        }
         const key = modelKey(candidate.provider, candidate.model);
-        if (seen.has(key))
+        if (seen.has(key)) {
             return;
-        if (enforceAllowlist && allowlist && !allowlist.has(key))
+        }
+        if (enforceAllowlist && allowlist && !allowlist.has(key)) {
             return;
+        }
         seen.add(key);
         candidates.push(candidate);
     };
+    return { candidates, addCandidate };
+}
+function sameModelCandidate(a, b) {
+    return a.provider === b.provider && a.model === b.model;
+}
+function throwFallbackFailureSummary(params) {
+    if (params.attempts.length <= 1 && params.lastError) {
+        throw params.lastError;
+    }
+    const summary = params.attempts.length > 0 ? params.attempts.map(params.formatAttempt).join(" | ") : "unknown";
+    throw new Error(`All ${params.label} failed (${params.attempts.length || params.candidates.length}): ${summary}`, {
+        cause: params.lastError instanceof Error ? params.lastError : undefined,
+    });
+}
+function resolveImageFallbackCandidates(params) {
+    const aliasIndex = buildModelAliasIndex({
+        cfg: params.cfg ?? {},
+        defaultProvider: params.defaultProvider,
+    });
+    const allowlist = buildConfiguredAllowlistKeys({
+        cfg: params.cfg,
+        defaultProvider: params.defaultProvider,
+    });
+    const { candidates, addCandidate } = createModelCandidateCollector(allowlist);
     const addRaw = (raw, enforceAllowlist) => {
         const resolved = resolveModelRefFromString({
             raw: String(raw ?? ""),
             defaultProvider: params.defaultProvider,
             aliasIndex,
         });
-        if (!resolved)
+        if (!resolved) {
             return;
+        }
         addCandidate(resolved.ref, enforceAllowlist);
     };
     if (params.modelOverride?.trim()) {
@@ -67,8 +78,9 @@ function resolveImageFallbackCandidates(params) {
     else {
         const imageModel = params.cfg?.agents?.defaults?.imageModel;
         const primary = typeof imageModel === "string" ? imageModel.trim() : imageModel?.primary;
-        if (primary?.trim())
+        if (primary?.trim()) {
             addRaw(primary, false);
+        }
     }
     const imageFallbacks = (() => {
         const imageModel = params.cfg?.agents?.defaults?.imageModel;
@@ -92,33 +104,33 @@ function resolveFallbackCandidates(params) {
         : null;
     const defaultProvider = primary?.provider ?? DEFAULT_PROVIDER;
     const defaultModel = primary?.model ?? DEFAULT_MODEL;
-    const provider = String(params.provider ?? "").trim() || defaultProvider;
-    const model = String(params.model ?? "").trim() || defaultModel;
+    const providerRaw = String(params.provider ?? "").trim() || defaultProvider;
+    const modelRaw = String(params.model ?? "").trim() || defaultModel;
+    const normalizedPrimary = normalizeModelRef(providerRaw, modelRaw);
+    const configuredPrimary = normalizeModelRef(defaultProvider, defaultModel);
     const aliasIndex = buildModelAliasIndex({
         cfg: params.cfg ?? {},
         defaultProvider,
     });
-    const allowlist = buildAllowedModelKeys(params.cfg, defaultProvider);
-    const seen = new Set();
-    const candidates = [];
-    const addCandidate = (candidate, enforceAllowlist) => {
-        if (!candidate.provider || !candidate.model)
-            return;
-        const key = modelKey(candidate.provider, candidate.model);
-        if (seen.has(key))
-            return;
-        if (enforceAllowlist && allowlist && !allowlist.has(key))
-            return;
-        seen.add(key);
-        candidates.push(candidate);
-    };
-    addCandidate({ provider, model }, false);
+    const allowlist = buildConfiguredAllowlistKeys({
+        cfg: params.cfg,
+        defaultProvider,
+    });
+    const { candidates, addCandidate } = createModelCandidateCollector(allowlist);
+    addCandidate(normalizedPrimary, false);
     const modelFallbacks = (() => {
-        if (params.fallbacksOverride !== undefined)
+        if (params.fallbacksOverride !== undefined) {
             return params.fallbacksOverride;
+        }
+        // Skip configured fallback chain when the user runs a non-default override.
+        // In that case, retry should return directly to configured primary.
+        if (!sameModelCandidate(normalizedPrimary, configuredPrimary)) {
+            return []; // Override model failed → go straight to configured default
+        }
         const model = params.cfg?.agents?.defaults?.model;
-        if (model && typeof model === "object")
+        if (model && typeof model === "object") {
             return model.fallbacks ?? [];
+        }
         return [];
     })();
     for (const raw of modelFallbacks) {
@@ -127,8 +139,9 @@ function resolveFallbackCandidates(params) {
             defaultProvider,
             aliasIndex,
         });
-        if (!resolved)
+        if (!resolved) {
             continue;
+        }
         addCandidate(resolved.ref, true);
     }
     if (params.fallbacksOverride === undefined && primary?.provider && primary.model) {
@@ -177,8 +190,8 @@ export async function runWithModelFallback(params) {
         ? ensureAuthProfileStore(params.agentDir, { allowKeychainPrompt: false })
         : null;
     const attempts = [];
-    const hasFallbackCandidates = candidates.length > 1;
     let lastError;
+    const hasFallbackCandidates = candidates.length > 1;
     for (let i = 0; i < candidates.length; i += 1) {
         const candidate = candidates[i];
         if (authStore) {
@@ -219,21 +232,6 @@ export async function runWithModelFallback(params) {
                 lastProbeAttempt.set(probeThrottleKey, now);
             }
         }
-        // Provider-level rate limit check (from token pool / rate limit tracking).
-        // Complements the per-profile cooldown above — skips candidates when all
-        // API keys in the pool are known to be rate-limited.
-        const poolLimit = isProviderRateLimited(candidate.provider);
-        if (poolLimit.isLimited && hasFallbackCandidates && i > 0) {
-            // Skip non-primary candidates that are pool-rate-limited.
-            // Primary (i === 0) is never skipped here — it gets a chance to probe.
-            attempts.push({
-                provider: candidate.provider,
-                model: candidate.model,
-                error: `Provider ${candidate.provider} pool is rate-limited (wait ${poolLimit.waitTimeMs}ms)`,
-                reason: "rate_limit",
-            });
-            continue;
-        }
         try {
             const result = await params.run(candidate.provider, candidate.model);
             return {
@@ -244,14 +242,24 @@ export async function runWithModelFallback(params) {
             };
         }
         catch (err) {
-            if (shouldRethrowAbort(err))
+            if (shouldRethrowAbort(err)) {
+                throw err;
+            }
+            // Context overflow errors should be handled by the inner runner's
+            // compaction/retry logic, not by model fallback.  If one escapes as a
+            // throw, rethrow it immediately rather than trying a different model
+            // that may have a smaller context window and fail worse.
+            const errMessage = err instanceof Error ? err.message : String(err);
+            if (isLikelyContextOverflowError(errMessage)) {
                 throw err;
+            }
             const normalized = coerceToFailoverError(err, {
                 provider: candidate.provider,
                 model: candidate.model,
             }) ?? err;
-            if (!isFailoverError(normalized))
+            if (!isFailoverError(normalized)) {
                 throw err;
+            }
             lastError = normalized;
             const described = describeFailoverError(normalized);
             attempts.push({
@@ -262,14 +270,6 @@ export async function runWithModelFallback(params) {
                 status: described.status,
                 code: described.code,
             });
-            // Feed error into provider infrastructure (rate limits + monitoring)
-            recordFallbackError({
-                provider: candidate.provider,
-                model: candidate.model,
-                status: described.status,
-                reason: described.reason,
-                error: described.message,
-            });
             await params.onError?.({
                 provider: candidate.provider,
                 model: candidate.model,
@@ -279,15 +279,12 @@ export async function runWithModelFallback(params) {
             });
         }
     }
-    if (attempts.length <= 1 && lastError)
-        throw lastError;
-    const summary = attempts.length > 0
-        ? attempts
-            .map((attempt) => `${attempt.provider}/${attempt.model}: ${attempt.error}${attempt.reason ? ` (${attempt.reason})` : ""}`)
-            .join(" | ")
-        : "unknown";
-    throw new Error(`All models failed (${attempts.length || candidates.length}): ${summary}`, {
-        cause: lastError instanceof Error ? lastError : undefined,
+    throwFallbackFailureSummary({
+        attempts,
+        candidates,
+        lastError,
+        label: "models",
+        formatAttempt: (attempt) => `${attempt.provider}/${attempt.model}: ${attempt.error}${attempt.reason ? ` (${attempt.reason})` : ""}`,
     });
 }
 export async function runWithImageModelFallback(params) {
@@ -313,8 +310,9 @@ export async function runWithImageModelFallback(params) {
             };
         }
         catch (err) {
-            if (shouldRethrowAbort(err))
+            if (shouldRethrowAbort(err)) {
                 throw err;
+            }
             lastError = err;
             attempts.push({
                 provider: candidate.provider,
@@ -330,14 +328,11 @@ export async function runWithImageModelFallback(params) {
             });
         }
     }
-    if (attempts.length <= 1 && lastError)
-        throw lastError;
-    const summary = attempts.length > 0
-        ? attempts
-            .map((attempt) => `${attempt.provider}/${attempt.model}: ${attempt.error}`)
-            .join(" | ")
-        : "unknown";
-    throw new Error(`All image models failed (${attempts.length || candidates.length}): ${summary}`, {
-        cause: lastError instanceof Error ? lastError : undefined,
+    throwFallbackFailureSummary({
+        attempts,
+        candidates,
+        lastError,
+        label: "image models",
+        formatAttempt: (attempt) => `${attempt.provider}/${attempt.model}: ${attempt.error}`,
     });
 }