fix(auto-capture): provider order, phase errors, AI session cleanup

README.md

@@ -456,7 +456,7 @@ Slow reasoning models (or very large profile prompts) can exceed the default 90
 
 Supported providers: any provider listed by `opencode providers list` (e.g. `anthropic`, `openai`, `github-copilot`, ...).
 
-If `opencodeProvider` and `opencodeModel` are set, they take precedence over the manual `memoryProvider` settings below.
+If `opencodeProvider` and `opencodeModel` are set, they take precedence over the manual `memoryProvider` settings below by default (`"autoCaptureProviderOrder": "opencode-first"`). Set `"autoCaptureProviderOrder": "external-first"` to prefer the manual API when both are configured (useful when the host session API is busy/slow); the other channel is used as fallback. Lower `"opencodeTimeoutMs"` to fail over from the opencode channel sooner.
 
 **Follow the session model:** set `"opencodeModel": "inherit"` to use a concrete OpenCode model at call time instead of a pinned id. For **auto-capture**, each prompt is recorded via the `chat.params` hook and the capture request reuses that prompt's provider/model. For **profile learning** and other structured-output paths (which are not tied to a single user message), `inherit` falls back to the most recent model in OpenCode's `model.json` recent list (preferring the configured `opencodeProvider`). Sending the literal model id `inherit` is never valid and previously caused `ProviderModelNotFoundError: Model not found: <provider>/inherit` on those paths. `opencodeProvider` is still required as the normal config gate.
 

opencode-mem.example.jsonc

@@ -182,6 +182,13 @@
 
   "autoCaptureEnabled": true,
 
+  // When both opencodeProvider and manual memoryModel/memoryApiUrl are set, which
+  // channel to try first for auto-capture summaries. Default "opencode-first".
+  // Prefer "external-first" if the host session API is slow/busy and you want the
+  // faster external API path. Failed primary falls back to the other channel.
+  // Faster opencode failover: also lower "opencodeTimeoutMs" (default 90000).
+  // "autoCaptureProviderOrder": "external-first",
+
   // Provider type: "atlas-cloud" | "cheaper-inference" | "openai-chat" | "openai-responses" | "anthropic" | "minimax" | "orcarouter"
   // Note: "openai-chat" is a generic OpenAI API-compatible mode.
   // Any service that follows the OpenAI Chat Completions API can use it via custom "memoryApiUrl".

src/config.ts

@@ -65,6 +65,13 @@ interface OpenCodeMemConfig {
   autoCaptureMaxRetries?: number;
   autoCaptureMaxContextBytes?: number;
   autoCaptureLanguage?: string;
+  /**
+   * Which auto-capture summary channel to try first when both opencode and
+   * manual (memoryModel/memoryApiUrl) providers are configured.
+   * Default "opencode-first". Use "external-first" for lower latency when the
+   * host session API is busy; failed primary channel falls back to the other.
+   */
+  autoCaptureProviderOrder?: "opencode-first" | "external-first";
   memoryProvider?:
     | "atlas-cloud"
     | "cheaper-inference"
@@ -256,6 +263,7 @@ const DEFAULTS: Required<
   autoCaptureIterationTimeout: 30000,
   autoCaptureMaxRetries: 3,
   autoCaptureMaxContextBytes: 131072,
+  autoCaptureProviderOrder: "opencode-first",
   aiSessionRetentionDays: 7,
   webServerEnabled: true,
   webServerPort: 4747,
@@ -600,6 +608,13 @@ const CONFIG_TEMPLATE = `{
   // Prevents HTTP 400 context overflows on models with ~131K token windows (e.g. Groq Llama).
   // Rough guide: tokens ≈ bytes / 4 for mixed code/prose.
   "autoCaptureMaxContextBytes": 131072,
+
+  // When both opencodeProvider and manual memoryModel/memoryApiUrl are set, which
+  // channel to try first for auto-capture summaries. Default "opencode-first".
+  // Prefer "external-first" if the host session API is slow/busy (~15–50s) and
+  // you want the faster external API path (~3s). Failed primary falls back.
+  // Faster opencode failover also: lower "opencodeTimeoutMs" (default 90000).
+  // "autoCaptureProviderOrder": "external-first",
 
   // Days to keep AI session history before cleanup
   "aiSessionRetentionDays": 7,
@@ -899,6 +914,13 @@ export function normalizeAutoCaptureMaxContextBytes(value: number): number {
   return value;
 }
 
+export type AutoCaptureProviderOrder = "opencode-first" | "external-first";
+
+/** Invalid / unset values fall back to the historical default (opencode-first). */
+export function normalizeAutoCaptureProviderOrder(value: unknown): AutoCaptureProviderOrder {
+  return value === "external-first" ? "external-first" : "opencode-first";
+}
+
 export function normalizeAutoCleanupRetentionDays(value: number): number {
   if (!Number.isSafeInteger(value) || value <= 0) {
     throw new Error(`Invalid autoCleanupRetentionDays config: ${value}`);
@@ -1059,6 +1081,9 @@ function buildConfig(fileConfig: OpenCodeMemConfig) {
     autoCaptureMaxRetries: fileConfig.autoCaptureMaxRetries ?? DEFAULTS.autoCaptureMaxRetries,
     autoCaptureMaxContextBytes,
     autoCaptureLanguage: fileConfig.autoCaptureLanguage,
+    autoCaptureProviderOrder: normalizeAutoCaptureProviderOrder(
+      fileConfig.autoCaptureProviderOrder
+    ),
     memoryProvider,
     memoryModel: fileConfig.memoryModel,
     memoryApiUrl: fileConfig.memoryApiUrl,

src/services/api-handlers.ts

@@ -896,7 +896,12 @@ export async function handleUnpinMemory(id: string): Promise<ApiResponse<void>>
 }
 
 export async function handleRunCleanup(): Promise<
-  ApiResponse<{ deletedCount: number; userCount: number; projectCount: number }>
+  ApiResponse<{
+    deletedCount: number;
+    userCount: number;
+    projectCount: number;
+    aiSessionsDeleted?: number;
+  }>
 > {
   try {
     const { cleanupService } = await import("./cleanup-service.js");