@hienlh/ppm 0.8.85 → 0.8.86

package/docs/system-architecture.md

@@ -140,24 +140,10 @@ POST   /api/db/connections/:id/query              → Execute query (readonly ch
 PATCH  /api/db/connections/:id/cell               → Update cell value (single)
 GET    /api/upgrade/status                        → Get current + available versions, install method
 POST   /api/upgrade/apply                         → Install new version, trigger supervisor self-replace
-GET    /api/project/:name/workspace               → Get saved workspace layout + metadata
-PUT    /api/project/:name/workspace               → Save workspace layout (layout JSON)
 WS     /ws/project/:name/chat/:sessionId          → Chat streaming
 WS     /ws/project/:name/terminal/:id             → Terminal I/O
 ```
 
-**URL Format (Deterministic Tabs, v0.8.77+):**
-```
-/project/{name}                          → Project root (project switcher)
-/project/{name}/editor/{filePath}        → Open editor tab (e.g., src/index.ts)
-/project/{name}/chat/{provider}/{sessionId} → Open chat tab
-/project/{name}/terminal/{index}         → Open terminal tab
-/project/{name}/database/{connId}/{table} → Open database browser
-/project/{name}/git-graph                → Git history graph (singleton)
-/project/{name}/settings                 → Settings panel (singleton)
-```
-Tab IDs are deterministic: `{type}:{identifier}` (e.g., `editor:src/index.ts`, `chat:claude/abc123`). Deep links auto-create missing tabs.
-
 ---
 
 ### Service Layer (Business Logic)
@@ -175,7 +161,7 @@ Tab IDs are deterministic: `{type}:{identifier}` (e.g., `editor:src/index.ts`, `
 |---------|---------|-------------|
 | **ChatService** | Session management, message streaming | createSession, streamMessage, getHistory |
 | **ConfigService** | Config loading (YAML→SQLite migration) | load, save, getToken |
-| **DbService** | SQLite persistence (10 tables, WAL, connections/accounts/workspace CRUD) | getDb, openTestDb, getWorkspace, setWorkspace, getConnections, insertConnection, deleteConnection, getTableCache |
+| **DbService** | SQLite persistence (9 tables, WAL, connections/accounts CRUD) | getDb, openTestDb, getConnections, insertConnection, deleteConnection, getTableCache |
 | **TableCacheService** | Cache table metadata, search tables | syncTables, searchTables, invalidateCache |
 | **GitService** | Git command execution | status, diff, commit, stage, branch |
 | **FileService** | File operations with validation | read, write, tree, delete, mkdir |
@@ -219,47 +205,7 @@ interface AIProvider {
 **Implementations:**
 - **claude-agent-sdk** (Primary) — @anthropic-ai/claude-agent-sdk, streaming, tool use. Reads model/effort/maxTurns/budget/thinking from config. Settings refreshed per query. Windows CLI fallback for Bun subprocess pipe issues. .env poisoning mitigation. **Multi-account support:** Injects account API token from AccountService instead of relying on ANTHROPIC_API_KEY env var when accounts configured.
 - **mock-provider** (Testing) — Returns canned responses
-- **cursor-cli** (CLI-based) — Spawns `cursor-agent` CLI binary with NDJSON streaming. Extends `CliProvider` base class.
-- **codex/gemini** (Planned) — Pluggable via `CliProvider` extension (~100-150 lines each)
-
-#### Multi-Provider Architecture (v0.8.61+)
-
-PPM supports multiple AI providers through a generic `AIProvider` interface and extensible base classes:
-
-**Provider Types:**
-1. **SDK-based** (claude-agent-sdk) — Uses Anthropic SDK for rich features (approvals, thinking blocks)
-2. **CLI-based** (cursor-cli, codex, gemini) — Spawns external binary with NDJSON streaming
-
-**Base Classes:**
-- `AIProvider` interface — Defines required methods (createSession, sendMessage) + optional capabilities (abortQuery, getMessages, listSessionsByDir, ensureProjectPath)
-- `CliProvider` abstract class — Shared spawn/parse/abort logic for all CLI-spawning providers
-- Provider-specific subclasses implement: `buildArgs()`, `mapEvent()`, `extractSessionId()`, `isAvailable()`
-
-**Streaming Infrastructure:**
-- `parseNdjsonLines()` utility — Async generator that buffers partial TCP packets, yields complete JSON lines
-- `ChatEvent` union type — Normalized event format across all providers (text, tool_use, thinking, approval_request, system, done, error)
-- Event mappers translate provider-specific JSON → ChatEvent (e.g., Cursor's `reasoning` type → `thinking` event)
-
-**Provider Registration & Bootstrap:**
-- `ProviderRegistry` maintains active provider instances
-- `bootstrapProviders()` async function checks `isAvailable()` on CLI providers before registering
-- Graceful fallback: if Cursor binary not found, provider skips registration (no crash, logged as info)
-- Config type `AIProviderConfig.type` union: `"agent-sdk" | "cli" | "mock"`
-
-**CLI-Provider Features:**
-- **Session capture** — Extract session ID from provider's init event, re-key process tracking
-- **Workspace trust auto-retry** — Detect trust prompts in stderr, retry once with `--trust` flag
-- **Process lifecycle** — Track active processes per session, escalate SIGTERM → SIGKILL on abort
-- **History loading** — Override `listSessions()` to read native provider history (e.g., Cursor SQLite DAG)
-- **Graceful degradation** — Missing binary → provider skipped, not fatal
-
-**New Files (v0.8.61):**
-- `src/utils/ndjson-line-parser.ts` — NDJSON streaming parser
-- `src/providers/cli-provider-base.ts` — Abstract base class for CLI providers
-- `src/providers/cursor-cli/cursor-provider.ts` — CursorCliProvider implementation
-- `src/providers/cursor-cli/cursor-event-mapper.ts` — NDJSON → ChatEvent mapping
-- `src/providers/cursor-cli/cursor-history.ts` — SQLite DAG reader for Cursor history
-- `src/web/components/chat/provider-selector.tsx` — UI component for provider selection
+- **Note:** CLI provider removed (v2); agent SDK is sole AI provider with Windows CLI fallback
 
 ---
 
@@ -274,11 +220,11 @@ PPM supports multiple AI providers through a generic `AIProvider` interface and
 - Enforce security (no parent directory access)
 
 **Key Patterns:**
-- SQLite: WAL mode, foreign keys, lazy init, schema v10 (10 tables: config, connections, accounts, usage_history, session_logs, push_subscriptions, session_map, table_metadata, session_logs, workspace_state)
+- SQLite: WAL mode, foreign keys, lazy init, schema v1 with 6 tables
 - Path validation: `projectPath/relativePath` only, reject `..`
 - Caching: Directory trees cached with TTL
 - Error handling: Descriptive messages (file not found, permission denied)
-- Migration: Automatic YAML→SQLite migration on first run with new db.service; schema auto-upgrade on version bump
+- Migration: Automatic YAML→SQLite migration on first run with new db.service
 
 ---
 
@@ -298,24 +244,6 @@ PPM supports multiple AI providers through a generic `AIProvider` interface and
 const messages = chatStore((s) => s.messages); // Subscribe to messages only
 ```
 
-#### Workspace Sync (v0.8.77+)
-
-**Deterministic Tab IDs & URL Routing:**
-- Tab IDs derived from type + metadata: `deriveTabId(type, metadata) → {type}:{identifier}`
-- Examples: `editor:src/index.ts`, `chat:claude/abc123`, `terminal:1`, `git-graph`
-- URLs rebuilt from active tab: `/project/{name}/{type}/{identifier}`
-- Deep linking: URL → `parseUrlState()` → auto-create tabs if missing
-
-**Workspace Persistence:**
-1. **Client**: PanelStore layout (grid, panels, tabs) cached in localStorage per project
-2. **Server**: Workspace JSON persisted in `workspace_state` SQLite table
-3. **Sync Flow:**
-   - User loads project → fetch workspace from server (GET `/api/project/:name/workspace`)
-   - Latest-wins: server `updated_at` vs client localStorage timestamp
-   - Panel layout changes debounced (1.5s) → POST to server
-   - On reconnect: server layout restored, client edits queued
-4. **Cross-Device:** Any device can load workspace, browser restores exact grid + active tabs
-
 ---
 
 ## Communication Protocols
@@ -525,27 +453,7 @@ Returns full updated config. Validates ranges/enums before writing.
 
 ---
 
-## Chat Streaming Flow (Persistent AsyncGenerator Sessions)
-
-### Architecture Overview (v0.8.55+)
-
-PPM uses a **persistent streaming session** model instead of per-message query execution:
-
-**Key Changes:**
-- Provider maintains **long-lived AsyncGenerator streaming input** per chat session (not per message)
-- Follow-up messages **push into the existing generator** instead of abort-and-replace
-- **Single streaming loop** per session decoupled from WebSocket message handler
-- Message priority support: `now` (interrupt current), `next` (queue first), `later` (queue at end)
-- Supports image attachments in messages
-
-**Design Benefits:**
-- Continuous context preservation — multi-turn conversations flow naturally
-- No SDK subprocess restarts between messages (faster)
-- Clean separation: BE owns Claude connection, FE disconnect doesn't abort
-- Message buffering on reconnect — clients that lose WS connection sync turn events
-- Tool approvals don't restart the query — integrated into streaming loop
-
-### Message Flow
+## Chat Streaming Flow
 
 ```
 User types: "Debug this function"
@@ -554,26 +462,18 @@ MessageInput.tsx calls useChat.sendMessage()
     ↓
 useChat opens WebSocket: WS /ws/project/:name/chat/:sessionId
     ↓
-Sends: { type: "message", content: "Debug...", priority?: "now"|"next"|"later" }
-    ↓
-WS handler in chat.ts receives message
+Sends: { type: "message", content: "Debug..." }
     ↓
-If already streaming with different content → abort previous + wait cleanup
-If streaming, new message priority determines queue behavior:
-    • priority: "now" → abort current, restart with new content
-    • priority: "next" → push into pending queue (higher priority)
-    • priority: "later" → push to end of queue (FIFO)
-    ↓
-runStreamLoop() executes in detached async context
+Server routes to ChatService.streamMessage()
     ↓
 ChatService calls provider.sendMessage() (async generator)
     ↓
-Provider (Claude SDK) yields events:
-    1. { type: "text", content: "Here's what..." }
-    2. { type: "text", content: " happens..." }
-    3. { type: "tool_use", tool: "read_file", input: {...} }
+Provider (Claude SDK) streams response:
+    1. Yields: { type: "text", content: "Here's what..." }
+    2. Yields: { type: "text", content: " happens..." }
+    3. Yields: { type: "tool_use", tool: "read_file", input: {...} }
     ↓
-Stream loop buffers + broadcasts to all connected clients:
+ChatService wraps as WebSocket messages:
     { type: "text", content: "Here's what..." }
     { type: "text", content: " happens..." }
     { type: "tool_use", tool: "read_file", input: {...} }
@@ -585,112 +485,15 @@ User sees tool approval prompt, clicks "Approve"
     ↓
 Client sends: { type: "approval_response", requestId, approved: true }
     ↓
-Provider continues streaming with tool result (no restart)
+ChatService.onToolApproval() executes tool (file_read, git commands, etc.)
     ↓
-If multiple messages queued, next message processes after done event
+Provider continues streaming with tool result
     ↓
 Final response streamed, then: { type: "done", sessionId }
     ↓
-Phase transitions to idle, clients can send new message
-    ↓
-useChat saves message to store, displays in chat history
-```
-
-### Session State Management
-
-**Session Entry** (BE-owned, persists across FE disconnections):
-```typescript
-interface SessionEntry {
-  providerId: string;              // Which AI provider (e.g., "claude")
-  clients: Set<ChatWsSocket>;      // Connected FE clients (may be empty)
-  abort?: AbortController;         // Current stream abort handle
-  projectPath?: string;            // Project context
-  projectName?: string;
-  pingIntervals: Map<...>;         // Per-client keepalive
-  phase: SessionPhase;             // "initializing" | "connecting" | "thinking" | "streaming" | "idle"
-  cleanupTimer?: ReturnType<...>;  // Auto-cleanup if no FE reconnects (5min)
-  pendingApprovalEvent?: {...};    // Current tool approval waiting
-  turnEvents: unknown[];           // Buffered events (for reconnect sync)
-  streamPromise?: Promise<void>;   // Track ongoing runStreamLoop
-  permissionMode?: string;         // Sticky permission mode for session
-}
+useChat closes WebSocket, saves message to store
 ```
 
-**Client Connection States:**
-- **Active streaming + FE connected** → Events broadcast to all clients in real-time
-- **Active streaming + FE disconnected** → Events buffered in turnEvents array, BE stream continues
-- **FE reconnects** → Receive session_state + buffered turnEvents, resync with stream
-- **Idle (no query running)** → Phase is "idle", ready for next message
-- **Idle + no FE for 5min** → Cleanup timer removes session from memory
-
-### Follow-up Messages
-
-**Abort-and-Replace Pattern:**
-```typescript
-if (entry.phase !== "idle" && entry.abort) {
-  console.log(`[chat] aborting current query for new message`);
-  entry.abort.abort();
-  await entry.streamPromise;  // Wait for cleanup
-  // Re-fetch entry — may have been mutated during cleanup
-  entry = activeSessions.get(sessionId)!;
-}
-```
-
-**Multiple Message Queueing:**
-- First message: immediately starts runStreamLoop
-- Second message (while streaming): abort current, wait, start new runStreamLoop
-- Priority modes (future): could queue messages for intelligent interleaving
-
-### WebSocket Reconnection Sync
-
-```
-FE WebSocket closes (network issue, tab closes)
-    ↓
-BE keeps session alive, streaming continues
-    ↓
-FE reconnects: WS /ws/project/:name/chat/:sessionId
-    ↓
-open() handler checks activeSessions.get(sessionId)
-    ↓
-If exists (entry found):
-    1. Clear cleanup timer (FE is back)
-    2. Send session_state with current phase + pendingApproval
-    3. If phase !== "idle", send buffered turnEvents
-    4. Add WS to clients Set
-    ↓
-FE processes session_state, renders current phase
-    ↓
-FE applies buffered events to rebuild turn state
-    ↓
-FE displays: "reconnected, current phase: streaming" etc.
-```
-
-### Phase Transitions
-
-```
-idle → initializing → connecting → thinking/streaming ↔ thinking/streaming → idle
-  ^                                      ↑                                    ↓
-  └──────────────────────────────────────────────────────────────────────────┘
-```
-
-**Phase Descriptions:**
-- **idle** — No query running, ready to accept new message
-- **initializing** — Preparing (permission checks, session resume)
-- **connecting** — Waiting for first SDK event (heartbeat: "connecting" with elapsed time every 5s)
-- **thinking** — Receiving thinking content (extended thinking)
-- **streaming** — Receiving text/tool_use content (dynamic switch between thinking/streaming)
-
-### Image Attachment Support
-
-Messages can now include images:
-```typescript
-type ChatWsClientMessage =
-  | { type: "message"; content: string; images?: { id: string; data: string }[]; priority?: string }
-  | ...
-```
-
-Images are passed to provider's message context and included in tool input/output.
-
 ---
 
 ## Terminal Flow

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@hienlh/ppm",
-  "version": "0.8.85",
+  "version": "0.8.86",
   "description": "Personal Project Manager — mobile-first web IDE with AI assistance",
   "author": "hienlh",
   "license": "MIT",
@@ -11,7 +11,7 @@
   },
   "scripts": {
     "dev": "concurrently \"bun run dev:server\" \"bun run dev:web\"",
-    "dev:server": "bun run --hot src/server/index.ts __serve__ 8081 0.0.0.0 '' dev",
+    "dev:server": "bun run --hot src/index.ts start --profile dev -f",
     "dev:web": "bun run vite --config vite.config.ts",
     "build:web": "bun run vite build --config vite.config.ts",
     "build": "bun run build:web && bun build src/index.ts --compile --outfile dist/ppm",
@@ -45,7 +45,6 @@
     "@radix-ui/react-switch": "^1.2.6",
     "@skitee3000/bun-pty": "^0.3.3",
     "@tanstack/react-table": "^8.21.3",
-    "@tanstack/react-virtual": "^3.13.23",
     "@uiw/react-codemirror": "^4.25.8",
     "@xterm/addon-fit": "^0.11.0",
     "@xterm/addon-web-links": "^0.12.0",

package/src/cli/commands/autostart.ts

@@ -10,7 +10,7 @@ export function registerAutoStartCommands(program: Command): void {
     .command("enable")
     .description("Register PPM to start automatically on boot")
     .option("-p, --port <port>", "Override port")
-    .option("-s, --share", "(deprecated) Tunnel is now always enabled")
+    .option("-s, --share", "Enable Cloudflare tunnel on boot")
     .option("-c, --config <path>", "Config file path")
     .option("--profile <name>", "DB profile name")
     .action(async (options) => {

package/src/cli/commands/restart.ts

@@ -9,7 +9,7 @@ const RESTARTING_FLAG = resolve(PPM_DIR, ".restarting");
 const RESTART_RESULT = resolve(PPM_DIR, ".restart-result");
 
 /** Restart only the server process, keeping the tunnel alive */
-export async function restartServer(options: { config?: string; force?: boolean }) {
+export async function restartServer(options: { config?: string }) {
   // Ignore SIGHUP so this process survives when PPM terminal dies
   process.on("SIGHUP", () => {});
 
@@ -34,14 +34,6 @@ export async function restartServer(options: { config?: string; force?: boolean
       process.exit(1);
     }
 
-    // Check if supervisor is paused — require --force to resume
-    const state = status.state as string | undefined;
-    if (state === "paused" && !options.force) {
-      console.log("\n  Server is paused (crashed too many times).");
-      console.log("  Use 'ppm restart --force' to resume.\n");
-      process.exit(1);
-    }
-
     const oldServerPid = status.pid as number | undefined;
     console.log("\n  Restarting PPM server via supervisor...");
     console.log("  If you're using PPM terminal, wait a few seconds for auto-reconnect.\n");

package/src/cli/commands/status.ts

@@ -15,10 +15,6 @@ interface DaemonStatus {
   tunnelAlive: boolean;
   supervisorPid: number | null;
   supervisorAlive: boolean;
-  state: string | null;
-  pausedAt: string | null;
-  pauseReason: string | null;
-  lastCrashError: string | null;
 }
 
 function isAlive(pid: number): boolean {
@@ -30,7 +26,6 @@ function getDaemonStatus(): DaemonStatus {
     running: false, pid: null, port: null, host: null,
     shareUrl: null, tunnelPid: null, tunnelAlive: false,
     supervisorPid: null, supervisorAlive: false,
-    state: null, pausedAt: null, pauseReason: null, lastCrashError: null,
   };
 
   if (existsSync(STATUS_FILE)) {
@@ -51,10 +46,6 @@ function getDaemonStatus(): DaemonStatus {
         tunnelAlive,
         supervisorPid,
         supervisorAlive,
-        state: (data.state as string) ?? null,
-        pausedAt: (data.pausedAt as string) ?? null,
-        pauseReason: (data.pauseReason as string) ?? null,
-        lastCrashError: (data.lastCrashError as string) ?? null,
       };
     } catch { return dead; }
   }
@@ -170,16 +161,6 @@ export async function showStatus(options: { json?: boolean; all?: boolean }) {
   if (status.supervisorPid) {
     console.log(`  Supervisor: ${status.supervisorAlive ? "running" : "stopped"} (PID: ${status.supervisorPid})`);
   }
-  // Show state info
-  const state = status.state ?? (status.running ? "running" : "stopped");
-  if (state === "paused") {
-    console.log(`  State:   PAUSED — ${status.pauseReason ?? "unknown reason"}`);
-    if (status.pausedAt) console.log(`  Paused:  ${status.pausedAt}`);
-    if (status.lastCrashError) console.log(`  Error:   ${status.lastCrashError}`);
-    console.log(`\n  Resume:  ppm restart --force`);
-  } else if (state === "upgrading") {
-    console.log(`  State:   UPGRADING`);
-  }
   console.log(`  Server:  ${status.running ? "running" : "stopped"} (PID: ${status.pid})`);
   if (status.port) console.log(`  Local:   http://localhost:${status.port}/`);
   if (status.tunnelPid) {

package/src/index.ts

@@ -16,7 +16,9 @@ program
   .command("start")
   .description("Start the PPM server (background by default)")
   .option("-p, --port <port>", "Port to listen on")
-  .option("-s, --share", "(deprecated) Tunnel is now always enabled")
+  .option("-f, --foreground", "Run in foreground (default: background daemon)")
+  .option("-d, --daemon", "Run as background daemon (default, kept for compat)")
+  .option("-s, --share", "Share via public URL (Cloudflare tunnel)")
   .option("-c, --config <path>", "Path to config file (YAML import into DB)")
   .option("--profile <name>", "DB profile name (e.g. 'dev' → ppm.dev.db)")
   .action(async (options) => {
@@ -49,7 +51,6 @@ program
   .command("restart")
   .description("Restart the server (keeps tunnel alive)")
   .option("-c, --config <path>", "Path to config file")
-  .option("--force", "Force resume from paused state")
   .action(async (options) => {
     const { restartServer } = await import("./cli/commands/restart.ts");
     await restartServer(options);

package/src/providers/claude-agent-sdk.ts

@@ -13,7 +13,7 @@ import type {
 } from "./provider.interface.ts";
 import { configService } from "../services/config.service.ts";
 import { updateFromSdkEvent } from "../services/claude-usage.service.ts";
-import { getSessionMapping, setSessionMapping, getSessionTitles, getSessionTitle } from "../services/db.service.ts";
+import { getSessionMapping, setSessionMapping } from "../services/db.service.ts";
 import { accountSelector } from "../services/account-selector.service.ts";
 import { accountService } from "../services/account.service.ts";
 import { resolve } from "node:path";
@@ -131,16 +131,6 @@ export class ClaudeAgentSdkProvider implements AIProvider {
     return null;
   }
 
-  /** Extract text content from an SDK assistant message */
-  private extractAssistantText(msg: unknown): string {
-    const content = (msg as any)?.message?.content;
-    if (!Array.isArray(content)) return "";
-    return content
-      .filter((b: any) => b.type === "text" && typeof b.text === "string")
-      .map((b: any) => b.text)
-      .join("");
-  }
-
   /** Read current provider config from yaml (fresh each call) */
   private getProviderConfig(): Partial<import("../types/config.ts").AIProviderConfig> {
     const ai = configService.get("ai");
@@ -176,11 +166,11 @@ export class ClaudeAgentSdkProvider implements AIProvider {
         (s) => s.sessionId === sessionId || s.sessionId === mappedSdkId,
       );
       if (found) {
-        const dbTitle = getSessionTitle(found.sessionId);
         const meta: Session = {
           id: sessionId,
           providerId: this.id,
-          title: dbTitle ?? found.customTitle ?? found.summary ?? "Resumed Chat",
+          title: found.customTitle ?? found.summary ?? "Resumed Chat",
+          projectPath: (found as any).cwd || undefined,
           createdAt: new Date(found.lastModified).toISOString(),
         };
         this.activeSessions.set(sessionId, meta);
@@ -211,13 +201,10 @@ export class ClaudeAgentSdkProvider implements AIProvider {
   async listSessionsByDir(dir?: string): Promise<SessionInfo[]> {
     try {
       const sdkSessions = await sdkListSessions({ dir, limit: 50 });
-      // Overlay DB titles (user-set) over SDK titles
-      const ids = sdkSessions.map((s) => s.sessionId);
-      const dbTitles = getSessionTitles(ids);
       return sdkSessions.map((s) => ({
         id: s.sessionId,
         providerId: this.id,
-        title: dbTitles[s.sessionId] ?? s.customTitle ?? s.summary ?? s.firstPrompt ?? "Chat",
+        title: s.customTitle ?? s.summary ?? s.firstPrompt ?? "Chat",
         createdAt: new Date(s.lastModified).toISOString(),
         updatedAt: new Date(s.lastModified).toISOString(),
       }));
@@ -495,8 +482,9 @@ export class ClaudeAgentSdkProvider implements AIProvider {
           if ((msg as any).type === "result" && (msg as any).subtype === "error_during_execution" && ((msg as any).num_turns ?? 0) === 0 && retryCount < MAX_RETRIES) {
             retryCount++;
             console.warn(`[sdk] transient error on first event — retrying (attempt ${retryCount}/${MAX_RETRIES})`);
-            // Re-create query for retry — don't reuse sessionId in case SDK partially created it
-            const retryOpts = { ...queryOptions, sessionId: undefined, resume: undefined };
+            // Resume existing SDK session on retry to preserve conversation history.
+            // If first message, SDK may have partially created the session — resume with sdkId.
+            const retryOpts = { ...queryOptions, sessionId: undefined, resume: sdkId };
             const rq = query({
               prompt: message,
               options: { ...retryOpts, ...(permissionHooks && { hooks: permissionHooks }), canUseTool } as any,
@@ -516,24 +504,22 @@ export class ClaudeAgentSdkProvider implements AIProvider {
 
         // Log all system events for debugging SDK lifecycle
         if (msg.type === "system") {
-          const subtype = (msg as any).subtype ?? "none";
-          console.log(`[sdk] session=${sessionId} system: subtype=${subtype} ${JSON.stringify(msg).slice(0, 500)}`);
-
-          // Capture SDK session metadata from init message
-          if (subtype === "init") {
-            const initMsg = msg as any;
-            if (initMsg.session_id && initMsg.session_id !== sessionId) {
-              setSessionMapping(sessionId, initMsg.session_id);
-              const oldMeta = this.activeSessions.get(sessionId);
-              if (oldMeta) {
-                this.activeSessions.set(initMsg.session_id, { ...oldMeta, id: initMsg.session_id });
-              }
+          console.log(`[sdk] session=${sessionId} system: subtype=${(msg as any).subtype ?? "none"} ${JSON.stringify(msg).slice(0, 500)}`);
+        }
+
+        // Capture SDK session metadata from init message
+        if (msg.type === "system" && (msg as any).subtype === "init") {
+          const initMsg = msg as any;
+          // SDK may assign a different session_id than our UUID
+          if (initMsg.session_id && initMsg.session_id !== sessionId) {
+            // Persist mapping so resume works after server restart
+            setSessionMapping(sessionId, initMsg.session_id);
+            // Update our in-memory mapping
+            const oldMeta = this.activeSessions.get(sessionId);
+            if (oldMeta) {
+              this.activeSessions.set(initMsg.session_id, { ...oldMeta, id: initMsg.session_id });
             }
           }
-
-          // Yield system events so streaming loop can transition phases
-          // (e.g. connecting → thinking when hooks/init arrive)
-          yield { type: "system" as any, subtype } as any;
           continue;
         }
 
@@ -630,26 +616,7 @@ export class ClaudeAgentSdkProvider implements AIProvider {
         // Full assistant message
         if (msg.type === "assistant") {
           // SDK assistant messages can carry an error field for auth/billing/rate-limit failures
-          let assistantError = (msg as any).error as string | undefined;
-
-          // SDK sometimes returns errors as text content without setting a specific error field.
-          // Detect known HTTP error patterns in text and reclassify accordingly.
-          if (!assistantError || assistantError === "unknown") {
-            const textContent = this.extractAssistantText(msg);
-            if (textContent) {
-              if (/API Error:\s*401\b.*authentication_error/i.test(textContent)) {
-                assistantError = "authentication_failed";
-                console.warn(`[sdk] session=${sessionId} detected 401 in assistant text content — treating as auth error`);
-              } else if (/API Error:\s*5\d{2}\b/i.test(textContent) || /internal server error/i.test(textContent)) {
-                assistantError = "server_error";
-                console.warn(`[sdk] session=${sessionId} detected 5xx in assistant text content — treating as server error`);
-              } else if (/API Error:\s*429\b/i.test(textContent) || /rate.?limit/i.test(textContent)) {
-                assistantError = "rate_limit";
-                console.warn(`[sdk] session=${sessionId} detected 429/rate-limit in assistant text content — treating as rate limit`);
-              }
-            }
-          }
-
+          const assistantError = (msg as any).error as string | undefined;
           if (assistantError) {
             // Dump full SDK message for debugging
             console.error(`[sdk] session=${sessionId} cwd=${effectiveCwd} assistant error: ${assistantError} (isFirst=${isFirstMessage} retry=${retryCount})`);
@@ -660,13 +627,13 @@ export class ClaudeAgentSdkProvider implements AIProvider {
               authRetried = true;
               try {
                 await accountService.refreshAccessToken(account.id, false);
+                console.log(`[sdk] session=${sessionId} OAuth token refreshed for ${account.id} — retrying`);
+                // Re-build env with refreshed token
                 const refreshedAccount = accountService.getWithTokens(account.id);
                 if (refreshedAccount) {
-                  const label = refreshedAccount.label ?? refreshedAccount.email ?? "Unknown";
-                  console.log(`[sdk] session=${sessionId} OAuth token refreshed for ${account.id} (${label}) — retrying`);
-                  yield { type: "account_retry" as const, reason: "Token refreshed", accountId: refreshedAccount.id, accountLabel: label };
                   const retryEnv = this.buildQueryEnv(meta.projectPath, refreshedAccount);
-                  const retryOpts = { ...queryOptions, sessionId: undefined, resume: undefined, env: retryEnv };
+                  // Resume existing SDK session to preserve conversation history
+                  const retryOpts = { ...queryOptions, sessionId: undefined, resume: sdkId, env: retryEnv };
                   const rq = query({
                     prompt: message,
                     options: { ...retryOpts, ...(permissionHooks && { hooks: permissionHooks }), canUseTool } as any,
@@ -691,8 +658,6 @@ export class ClaudeAgentSdkProvider implements AIProvider {
             };
             const hint = errorHints[assistantError] ?? `API error: ${assistantError}`;
             yield { type: "error", message: hint };
-            // Skip emitting the raw 401 error as text content — already shown as error event
-            continue;
           }
           const content = (msg as any).message?.content;
           if (Array.isArray(content)) {
@@ -748,30 +713,11 @@ export class ClaudeAgentSdkProvider implements AIProvider {
               yield { type: "error", message: "Rate limited. This account is now on cooldown. Please retry." };
               break;
             } else if (errCode === 401) {
-              // Refresh token and retry with fresh session (same logic as assistant-level auth retry)
-              if (!authRetried) {
-                authRetried = true;
-                try {
-                  await accountService.refreshAccessToken(account.id, false);
-                  const refreshedAccount = accountService.getWithTokens(account.id);
-                  if (refreshedAccount) {
-                    const label = refreshedAccount.label ?? refreshedAccount.email ?? "Unknown";
-                    console.log(`[sdk] 401 in result on account ${account.id} (${label}) — token refreshed, retrying`);
-                    yield { type: "account_retry" as const, reason: "Token refreshed", accountId: refreshedAccount.id, accountLabel: label };
-                    const retryEnv = this.buildQueryEnv(meta.projectPath, refreshedAccount);
-                    const retryOpts = { ...queryOptions, sessionId: undefined, resume: undefined, env: retryEnv };
-                    const rq = query({
-                      prompt: message,
-                      options: { ...retryOpts, ...(permissionHooks && { hooks: permissionHooks }), canUseTool } as any,
-                    });
-                    this.activeQueries.set(sessionId, rq);
-                    eventSource = rq;
-                    continue retryLoop;
-                  }
-                } catch {
-                  accountSelector.onAuthError(account.id);
-                }
-              } else {
+              // Try refresh once
+              try {
+                await accountService.refreshAccessToken(account.id, false);
+                console.log(`[sdk] 401 on account ${account.id} — token refreshed`);
+              } catch {
                 accountSelector.onAuthError(account.id);
               }
             } else {
@@ -813,16 +759,7 @@ export class ClaudeAgentSdkProvider implements AIProvider {
           }
 
           // Surface non-success subtypes as errors so FE can display them
-          // But suppress abort errors — user-initiated cancel is not a real error
           if (subtype && subtype !== "success") {
-            const errorsArr0 = Array.isArray(result.errors) ? result.errors : [];
-            const abortDetail = errorsArr0.join(" ") + " " + (typeof result.error === "string" ? result.error : "");
-            if (subtype === "error_during_execution" && /abort|request was aborted/i.test(abortDetail)) {
-              console.log(`[sdk] session=${sessionId} suppressing abort error (user-initiated cancel)`);
-              resultSubtype = subtype;
-              resultNumTurns = result.num_turns as number | undefined;
-              break;
-            }
             // SDK error results use `errors: string[]` array (not singular `error`)
             const errorsArr = Array.isArray(result.errors) ? result.errors : [];
             const sdkDetail = errorsArr.length > 0

package/src/providers/mock-provider.ts

@@ -92,13 +92,8 @@ export class MockProvider implements AIProvider {
     const abortController = new AbortController();
     this.activeAborts.set(sessionId, abortController);
 
-    // Simulate SDK system events (hooks, init) — real SDK emits these before content
-    yield { type: "system" as any, subtype: "hook_started" } as any;
-    await sleep(50);
-    yield { type: "system" as any, subtype: "init" } as any;
-
     // Simulate thinking delay
-    await sleep(250);
+    await sleep(300);
 
     // Pick a response
     const responseText =