agent-relay 2.0.19 → 2.0.20

package/packages/cli-tester/scripts/test-full-spawn.sh

@@ -0,0 +1,238 @@
+#!/bin/bash
+# Full spawn simulation - replicates the EXACT spawner flow
+# This is the most accurate test for debugging registration timeouts
+#
+# Usage: ./test-full-spawn.sh <cli> [interactive]
+# Example: ./test-full-spawn.sh cursor           # Non-interactive (with --force)
+#          ./test-full-spawn.sh cursor true      # Interactive (no --force, like setup terminal)
+#          DEBUG=1 ./test-full-spawn.sh cursor true
+
+set -e
+
+CLI=${1:-cursor}
+INTERACTIVE=${2:-false}
+
+# Map CLI to command
+CLI_CMD="$CLI"
+if [ "$CLI" = "cursor" ]; then
+    CLI_CMD="agent"
+fi
+
+# Generate name
+if [ "$INTERACTIVE" = "true" ]; then
+    TIMESTAMP=$(date +%s | tail -c 8)
+    RANDOM_SUFFIX=$(head /dev/urandom | tr -dc 'a-z0-9' | head -c 4)
+    NAME="__setup__${CLI}-${TIMESTAMP}${RANDOM_SUFFIX}"
+else
+    NAME="spawn-test-${CLI}"
+fi
+
+SOCKET="/tmp/relay-pty-${NAME}.sock"
+LOG_FILE="/tmp/relay-spawn-${NAME}.log"
+REGISTRATION_TIMEOUT=30
+
+echo "========================================"
+echo "  Full Spawn Simulation: $CLI"
+echo "========================================"
+echo ""
+echo "Mode: $([ "$INTERACTIVE" = "true" ] && echo "INTERACTIVE (setup terminal)" || echo "NON-INTERACTIVE (normal spawn)")"
+echo ""
+echo "Configuration:"
+echo "  CLI:                 $CLI_CMD"
+echo "  Agent name:          $NAME"
+echo "  Interactive:         $INTERACTIVE"
+echo "  Registration timeout: ${REGISTRATION_TIMEOUT}s"
+echo ""
+
+# Build CLI args based on mode
+CLI_ARGS=()
+if [ "$INTERACTIVE" != "true" ]; then
+    case $CLI in
+        cursor)
+            CLI_ARGS+=(--force)
+            echo "  Adding: --force (non-interactive cursor)"
+            ;;
+        claude)
+            CLI_ARGS+=(--dangerously-skip-permissions)
+            echo "  Adding: --dangerously-skip-permissions (non-interactive claude)"
+            ;;
+        copilot)
+            # Copilot uses device flow for headless auth
+            CLI_ARGS+=(auth login --device)
+            echo "  Adding: auth login --device (non-interactive copilot)"
+            ;;
+        codex)
+            # Codex supports device flow
+            CLI_ARGS+=(login --device-auth)
+            echo "  Adding: login --device-auth (non-interactive codex)"
+            ;;
+    esac
+else
+    echo "  No auto-accept flags (interactive mode)"
+    # For interactive copilot, still need auth login command
+    if [ "$CLI" = "copilot" ]; then
+        CLI_ARGS+=(auth login)
+        echo "  Adding: auth login (interactive copilot)"
+    fi
+fi
+echo ""
+
+# Cleanup
+cleanup() {
+    echo ""
+    echo "========================================"
+    echo "  Cleanup"
+    echo "========================================"
+    rm -f "$SOCKET"
+    if [ -n "$PTY_PID" ] && kill -0 $PTY_PID 2>/dev/null; then
+        echo "Stopping relay-pty (PID: $PTY_PID)..."
+        kill $PTY_PID 2>/dev/null || true
+    fi
+    echo ""
+    echo "Log file: $LOG_FILE"
+    if [ -f "$LOG_FILE" ]; then
+        echo ""
+        echo "Last 30 lines of output:"
+        echo "----------------------------------------"
+        tail -30 "$LOG_FILE"
+        echo "----------------------------------------"
+    fi
+}
+trap cleanup EXIT
+
+# Remove stale files
+rm -f "$SOCKET" "$LOG_FILE"
+
+echo "========================================"
+echo "  Phase 1: Start PTY"
+echo "========================================"
+echo ""
+
+# This is what spawner does: pty.start()
+RELAY_ARGS=(
+    --name "$NAME"
+    --socket "$SOCKET"
+    --idle-timeout 300
+)
+
+if [ -n "$DEBUG" ]; then
+    RELAY_ARGS+=(--json-output)
+fi
+
+echo "[$(date +%T)] Starting: relay-pty ${RELAY_ARGS[*]} -- $CLI_CMD ${CLI_ARGS[*]}"
+echo ""
+
+# Start relay-pty and capture output
+if [ ${#CLI_ARGS[@]} -gt 0 ]; then
+    relay-pty "${RELAY_ARGS[@]}" -- "$CLI_CMD" "${CLI_ARGS[@]}" > "$LOG_FILE" 2>&1 &
+else
+    relay-pty "${RELAY_ARGS[@]}" -- "$CLI_CMD" > "$LOG_FILE" 2>&1 &
+fi
+PTY_PID=$!
+
+echo "[$(date +%T)] PTY started (PID: $PTY_PID)"
+
+# Brief wait for startup
+sleep 1
+
+if ! kill -0 $PTY_PID 2>/dev/null; then
+    echo "[$(date +%T)] ERROR: PTY exited immediately!"
+    echo ""
+    cat "$LOG_FILE"
+    exit 1
+fi
+
+echo "[$(date +%T)] PTY is running"
+echo ""
+
+echo "========================================"
+echo "  Phase 2: Wait for Registration"
+echo "========================================"
+echo ""
+echo "In the real spawner, this calls waitForAgentRegistration()"
+echo "which polls connected-agents.json and agents.json for ${REGISTRATION_TIMEOUT}s"
+echo ""
+echo "Without a daemon, these files won't be updated, causing timeout."
+echo "But we can still see if the CLI starts successfully."
+echo ""
+
+# Simulate the spawner's registration polling
+START_TIME=$(date +%s)
+POLL_COUNT=0
+
+while true; do
+    POLL_COUNT=$((POLL_COUNT + 1))
+    CURRENT_TIME=$(date +%s)
+    ELAPSED=$((CURRENT_TIME - START_TIME))
+
+    # Check if PTY is still running
+    if ! kill -0 $PTY_PID 2>/dev/null; then
+        echo ""
+        echo "[$(date +%T)] PTY exited after ${ELAPSED}s (${POLL_COUNT} polls)"
+        echo ""
+        echo "Exit could be normal (auth complete) or an error."
+        echo "Check log output below."
+        exit 0
+    fi
+
+    # Check socket status
+    SOCKET_STATUS="no socket"
+    if [ -S "$SOCKET" ]; then
+        RESPONSE=$(echo '{"type":"status"}' | timeout 1 nc -U "$SOCKET" 2>/dev/null | head -1 || echo "")
+        if [ -n "$RESPONSE" ]; then
+            IDLE=$(echo "$RESPONSE" | grep -o '"agent_idle":[^,}]*' | cut -d: -f2 || echo "?")
+            QUEUE=$(echo "$RESPONSE" | grep -o '"queue_length":[^,}]*' | cut -d: -f2 || echo "?")
+            SOCKET_STATUS="idle=$IDLE queue=$QUEUE"
+        else
+            SOCKET_STATUS="no response"
+        fi
+    fi
+
+    # Log poll (like spawner does for first 3 and every 10th)
+    if [ $POLL_COUNT -le 3 ] || [ $((POLL_COUNT % 10)) -eq 0 ]; then
+        echo "[$(date +%T)] Poll #$POLL_COUNT: elapsed=${ELAPSED}s socket=[$SOCKET_STATUS]"
+    fi
+
+    # Check registration timeout (what spawner does)
+    if [ $ELAPSED -ge $REGISTRATION_TIMEOUT ]; then
+        echo ""
+        echo "========================================"
+        echo "  REGISTRATION TIMEOUT (${REGISTRATION_TIMEOUT}s)"
+        echo "========================================"
+        echo ""
+        echo "This is where the real spawner would fail with:"
+        echo "  'Agent registration timeout'"
+        echo ""
+        echo "The CLI started but didn't register with daemon."
+        echo ""
+        echo "Possible causes:"
+        if [ "$INTERACTIVE" = "true" ]; then
+            echo "  1. CLI is waiting for user input (trust prompt, auth)"
+            echo "  2. No daemon running to register with"
+            echo "  3. CLI crashed silently"
+        else
+            echo "  1. --force flag not working as expected"
+            echo "  2. No daemon running to register with"
+            echo "  3. CLI has other blocking prompts"
+        fi
+        echo ""
+        echo "To see what the CLI is doing, check the log file:"
+        echo "  cat $LOG_FILE"
+        echo ""
+        echo "Or tail it live in another terminal:"
+        echo "  tail -f $LOG_FILE"
+        echo ""
+
+        # Show recent output
+        echo "Recent output (last 20 lines):"
+        echo "----------------------------------------"
+        tail -20 "$LOG_FILE" 2>/dev/null || echo "(no output)"
+        echo "----------------------------------------"
+        echo ""
+        echo "Process is still running. Press Enter to stop."
+        read -r
+        exit 1
+    fi
+
+    sleep 0.5
+done

package/packages/cli-tester/scripts/test-registration.sh

@@ -0,0 +1,182 @@
+#!/bin/bash
+# Test agent registration flow - monitors what happens during spawn
+# This tests the critical registration step that's timing out for cursor
+#
+# Usage: ./test-registration.sh <cli> [timeout_seconds]
+# Example: ./test-registration.sh cursor 60
+#          DEBUG_SPAWN=1 ./test-registration.sh cursor
+
+set -e
+
+CLI=${1:-cursor}
+TIMEOUT_SEC=${2:-60}
+
+# Map CLI name to actual command
+CLI_CMD="$CLI"
+if [ "$CLI" = "cursor" ]; then
+    CLI_CMD="agent"
+fi
+
+NAME="reg-test-${CLI}"
+SOCKET="/tmp/relay-pty-${NAME}.sock"
+
+# Test data directories (mimic real daemon structure)
+TEST_DATA_DIR="/tmp/relay-registration-test"
+AGENTS_FILE="$TEST_DATA_DIR/agents.json"
+CONNECTED_FILE="$TEST_DATA_DIR/connected-agents.json"
+
+# Create directories
+mkdir -p "$TEST_DATA_DIR"
+
+# Initialize empty registration files
+echo '{"agents":[],"updatedAt":'$(date +%s000)'}' > "$CONNECTED_FILE"
+echo '{}' > "$AGENTS_FILE"
+
+# Cleanup function
+cleanup() {
+    echo ""
+    echo "========================================"
+    echo "  Final State"
+    echo "========================================"
+    echo ""
+    echo "Connected agents file ($CONNECTED_FILE):"
+    cat "$CONNECTED_FILE" 2>/dev/null | jq . 2>/dev/null || cat "$CONNECTED_FILE" 2>/dev/null || echo "  (empty/missing)"
+    echo ""
+    echo "Agents registry file ($AGENTS_FILE):"
+    cat "$AGENTS_FILE" 2>/dev/null | jq . 2>/dev/null || cat "$AGENTS_FILE" 2>/dev/null || echo "  (empty/missing)"
+    echo ""
+    echo "Cleaning up processes..."
+    rm -f "$SOCKET"
+    pkill -f "relay-pty.*${NAME}" 2>/dev/null || true
+}
+trap cleanup EXIT
+
+# Remove stale socket
+rm -f "$SOCKET"
+
+echo "========================================"
+echo "  Registration Flow Test: $CLI"
+echo "========================================"
+echo ""
+echo "This test monitors the registration files that the spawner polls."
+echo "When the spawner times out, it's because these files don't show"
+echo "the agent as registered within 30 seconds."
+echo ""
+echo "Test config:"
+echo "  CLI:            $CLI_CMD"
+echo "  Name:           $NAME"
+echo "  Socket:         $SOCKET"
+echo "  Timeout:        ${TIMEOUT_SEC}s"
+echo "  Connected file: $CONNECTED_FILE"
+echo "  Agents file:    $AGENTS_FILE"
+echo ""
+
+# Build CLI arguments
+CLI_ARGS=()
+case $CLI in
+    claude)
+        CLI_ARGS+=(--dangerously-skip-permissions)
+        ;;
+    cursor)
+        CLI_ARGS+=(--force)
+        ;;
+esac
+
+echo "Command: $CLI_CMD ${CLI_ARGS[*]}"
+echo ""
+echo "========================================"
+echo "  Phase 1: Starting CLI"
+echo "========================================"
+echo ""
+
+# Start relay-pty in background
+RELAY_ARGS=(
+    --name "$NAME"
+    --socket "$SOCKET"
+    --idle-timeout 300
+    --json-output
+)
+
+if [ ${#CLI_ARGS[@]} -gt 0 ]; then
+    relay-pty "${RELAY_ARGS[@]}" -- "$CLI_CMD" "${CLI_ARGS[@]}" 2>&1 &
+else
+    relay-pty "${RELAY_ARGS[@]}" -- "$CLI_CMD" 2>&1 &
+fi
+
+PTY_PID=$!
+echo "[$(date +%T)] Started relay-pty (PID: $PTY_PID)"
+
+# Give it a moment to start
+sleep 1
+
+# Check if process is still running
+if ! kill -0 $PTY_PID 2>/dev/null; then
+    echo "[$(date +%T)] ERROR: relay-pty exited immediately"
+    wait $PTY_PID || true
+    exit 1
+fi
+
+echo "[$(date +%T)] relay-pty is running"
+echo ""
+
+echo "========================================"
+echo "  Phase 2: Monitoring Registration"
+echo "========================================"
+echo ""
+echo "Polling for agent registration (timeout: ${TIMEOUT_SEC}s)..."
+echo "In a real spawn, the daemon would update these files."
+echo ""
+
+# Poll for registration (simulating what spawner does)
+START_TIME=$(date +%s)
+POLL_COUNT=0
+
+while true; do
+    POLL_COUNT=$((POLL_COUNT + 1))
+    CURRENT_TIME=$(date +%s)
+    ELAPSED=$((CURRENT_TIME - START_TIME))
+
+    # Check if process is still running
+    if ! kill -0 $PTY_PID 2>/dev/null; then
+        echo ""
+        echo "[$(date +%T)] ERROR: relay-pty exited after ${ELAPSED}s"
+        echo ""
+        echo "This is likely the issue - the CLI crashed or exited before registration."
+        echo "Check the output above for error messages."
+        wait $PTY_PID || true
+        exit 1
+    fi
+
+    # Check socket exists
+    if [ -S "$SOCKET" ]; then
+        # Try to get status from socket
+        STATUS=$(echo '{"type":"status"}' | timeout 2 nc -U "$SOCKET" 2>/dev/null || echo "timeout")
+        if [ "$STATUS" != "timeout" ]; then
+            echo "[$(date +%T)] Poll #$POLL_COUNT: Socket active, status: $(echo "$STATUS" | head -c 100)..."
+        else
+            echo "[$(date +%T)] Poll #$POLL_COUNT: Socket exists but no response"
+        fi
+    else
+        echo "[$(date +%T)] Poll #$POLL_COUNT: Socket not yet created"
+    fi
+
+    # Check timeout
+    if [ $ELAPSED -ge $TIMEOUT_SEC ]; then
+        echo ""
+        echo "========================================"
+        echo "  TIMEOUT after ${ELAPSED}s"
+        echo "========================================"
+        echo ""
+        echo "This simulates what happens when the spawner times out."
+        echo "In a real scenario, the daemon would need to see this agent"
+        echo "and update the registration files within 30 seconds."
+        echo ""
+        echo "To debug further:"
+        echo "  1. Check if the CLI started correctly (auth issues?)"
+        echo "  2. Check if relay-pty connected to the daemon"
+        echo "  3. Check daemon logs for this agent name"
+        exit 1
+    fi
+
+    sleep 1
+done

package/packages/cli-tester/scripts/test-setup-flow.sh

@@ -0,0 +1,202 @@
+#!/bin/bash
+# Test the EXACT setup terminal flow used for provider onboarding
+# This replicates what TerminalProviderSetup.tsx does when you click "Connect"
+#
+# The setup flow:
+# 1. Frontend calls POST /api/workspaces/{id}/agents with { name, provider, interactive: true }
+# 2. Spawner runs: relay-pty --name __setup__cursor-xxx -- agent (NO --force flag!)
+# 3. Spawner waits up to 30s for agent to register with daemon
+# 4. If timeout, spawn fails with "Agent registration timeout"
+#
+# Usage: ./test-setup-flow.sh <cli> [timeout_seconds]
+# Example: ./test-setup-flow.sh cursor 60
+#          DEBUG=1 ./test-setup-flow.sh cursor
+
+set -e
+
+CLI=${1:-cursor}
+TIMEOUT_SEC=${2:-60}
+
+# Generate setup agent name like frontend does
+TIMESTAMP=$(date +%s | tail -c 8)
+RANDOM_SUFFIX=$(head /dev/urandom | tr -dc 'a-z0-9' | head -c 4)
+NAME="__setup__${CLI}-${TIMESTAMP}${RANDOM_SUFFIX}"
+
+# Map CLI to command
+CLI_CMD="$CLI"
+if [ "$CLI" = "cursor" ]; then
+    CLI_CMD="agent"
+fi
+
+SOCKET="/tmp/relay-pty-${NAME}.sock"
+
+echo "========================================"
+echo "  Setup Terminal Flow Test: $CLI"
+echo "========================================"
+echo ""
+echo "This test replicates the EXACT flow from TerminalProviderSetup.tsx"
+echo ""
+echo "Key difference from test-spawn.sh:"
+echo "  - interactive=true means NO auto-accept flags (--force, etc.)"
+echo "  - This is how setup terminals work - user must respond to prompts"
+echo ""
+echo "Configuration:"
+echo "  CLI:          $CLI_CMD"
+echo "  Agent name:   $NAME"
+echo "  Socket:       $SOCKET"
+echo "  Interactive:  true (no --force, no --dangerously-skip-permissions)"
+echo "  Timeout:      ${TIMEOUT_SEC}s"
+echo ""
+
+# Interactive mode - no auto-accept flags, but some CLIs need auth subcommands
+CLI_ARGS=()
+case $CLI in
+    copilot)
+        CLI_ARGS+=(auth login)
+        ;;
+    opencode)
+        CLI_ARGS+=(auth login)
+        ;;
+    droid)
+        CLI_ARGS+=(--login)
+        ;;
+    codex)
+        CLI_ARGS+=(login)
+        ;;
+esac
+
+if [ ${#CLI_ARGS[@]} -gt 0 ]; then
+    echo "Command: relay-pty --name $NAME -- $CLI_CMD ${CLI_ARGS[*]}"
+else
+    echo "Command: relay-pty --name $NAME -- $CLI_CMD"
+fi
+echo "         (no auto-accept flags because interactive=true)"
+echo ""
+
+# Cleanup
+cleanup() {
+    echo ""
+    echo "Cleaning up..."
+    rm -f "$SOCKET"
+    if [ -n "$PTY_PID" ]; then
+        kill $PTY_PID 2>/dev/null || true
+    fi
+}
+trap cleanup EXIT
+
+# Remove stale socket
+rm -f "$SOCKET"
+
+echo "========================================"
+echo "  Starting Setup Terminal"
+echo "========================================"
+echo ""
+
+# Build relay-pty args
+RELAY_ARGS=(
+    --name "$NAME"
+    --socket "$SOCKET"
+    --idle-timeout 300
+)
+
+# Add verbose output
+if [ -n "$DEBUG" ]; then
+    RELAY_ARGS+=(--json-output)
+    echo "[DEBUG] JSON output enabled"
+    echo ""
+fi
+
+# Start in background so we can monitor
+if [ ${#CLI_ARGS[@]} -gt 0 ]; then
+    relay-pty "${RELAY_ARGS[@]}" -- "$CLI_CMD" "${CLI_ARGS[@]}" 2>&1 &
+else
+    relay-pty "${RELAY_ARGS[@]}" -- "$CLI_CMD" 2>&1 &
+fi
+PTY_PID=$!
+
+echo "[$(date +%T)] Started relay-pty (PID: $PTY_PID)"
+echo ""
+echo "The CLI is now running. In a real setup flow, you would see"
+echo "the terminal output in the browser and respond to prompts there."
+echo ""
+echo "Watch for:"
+echo "  - Trust prompts (workspace trust, etc.)"
+echo "  - Auth URL being printed"
+echo "  - Any errors or crashes"
+echo ""
+
+# Give it a moment
+sleep 2
+
+# Check if still running
+if ! kill -0 $PTY_PID 2>/dev/null; then
+    echo "[$(date +%T)] ERROR: relay-pty exited immediately!"
+    echo ""
+    echo "This means the CLI crashed on startup."
+    echo "Common causes:"
+    echo "  - CLI not installed"
+    echo "  - Missing dependencies"
+    echo "  - Permission issues"
+    exit 1
+fi
+
+echo "========================================"
+echo "  Monitoring (${TIMEOUT_SEC}s timeout)"
+echo "========================================"
+echo ""
+echo "In a real spawn, the spawner would poll registration files here."
+echo "The timeout you're seeing is because this polling fails."
+echo ""
+echo "Press Ctrl+C to stop, or wait for timeout."
+echo ""
+
+START_TIME=$(date +%s)
+
+while true; do
+    CURRENT_TIME=$(date +%s)
+    ELAPSED=$((CURRENT_TIME - START_TIME))
+
+    # Check process
+    if ! kill -0 $PTY_PID 2>/dev/null; then
+        echo ""
+        echo "[$(date +%T)] CLI exited after ${ELAPSED}s"
+        echo ""
+        echo "If this was unexpected, check the output above for errors."
+        echo "If the CLI authenticated and exited normally, that's OK."
+        exit 0
+    fi
+
+    # Check socket
+    if [ -S "$SOCKET" ]; then
+        STATUS=$(echo '{"type":"status"}' | timeout 2 nc -U "$SOCKET" 2>/dev/null | head -1 || echo "")
+        if [ -n "$STATUS" ]; then
+            IDLE=$(echo "$STATUS" | grep -o '"agent_idle":[^,}]*' | cut -d: -f2 || echo "unknown")
+            echo "[$(date +%T)] +${ELAPSED}s: Socket OK, idle=$IDLE"
+        else
+            echo "[$(date +%T)] +${ELAPSED}s: Socket exists, no status response"
+        fi
+    else
+        echo "[$(date +%T)] +${ELAPSED}s: Waiting for socket..."
+    fi
+
+    # Timeout check
+    if [ $ELAPSED -ge $TIMEOUT_SEC ]; then
+        echo ""
+        echo "========================================"
+        echo "  Test Complete (${TIMEOUT_SEC}s elapsed)"
+        echo "========================================"
+        echo ""
+        echo "The CLI ran for ${TIMEOUT_SEC}s without exiting."
+        echo ""
+        echo "If this were a real spawn:"
+        echo "  - The spawner would have timed out at 30s"
+        echo "  - You'd see 'Agent registration timeout' error"
+        echo ""
+        echo "To continue interacting with the CLI, the process"
+        echo "is still running. Press Enter to stop it."
+        read -r
+        exit 0
+    fi
+
+    sleep 3
+done