Merge pull request #728 from gabsprogrammer/fix/preflight-llm-port-default-and-pipefail

fix(preflight): align llama-server default port to 8080 and add set -euo pipefail

Author: Lightheartdevs

dream-server/dream-preflight.sh

@@ -4,7 +4,7 @@
 # Backend-aware: detects AMD vs NVIDIA (both use llama-server)
 # Usage: ./dream-preflight.sh
 
-set -e
+set -euo pipefail
 
 SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
 DREAM_DIR="$SCRIPT_DIR"
@@ -159,22 +159,19 @@ else
 fi
 log ""
 
-# 4. LLM Endpoint check — backend-aware
+# 4. LLM Endpoint check
+# OLLAMA_PORT controls the external port for llama-server.
+# Canonical default is 8080 (config/ports.json, docker-compose.base.yml).
+# 11434 is only used on Strix Halo AMD installs where phase 06 writes
+# OLLAMA_PORT=11434 to .env automatically — it will be picked up via the
+# ${OLLAMA_PORT:-...} expansion below, so the fallback should be 8080.
 log "[4/8] Checking LLM endpoint..."
-if [[ "$BACKEND" == "amd" ]]; then
-    LLM_PORT="${OLLAMA_PORT:-${LLAMA_SERVER_PORT:-11434}}"
-    # llama-server may be mapped to a different external port
-    EXTERNAL_PORT=$(docker port dream-llama-server 8080/tcp 2>/dev/null | head -1 | cut -d: -f2 || echo "$LLM_PORT")
-    LLM_ENDPOINTS=("http://${SERVICE_HOST}:${EXTERNAL_PORT}" "http://localhost:${EXTERNAL_PORT}" "http://localhost:${LLM_PORT}")
-    LLM_SERVICE_NAME="llama-server"
-    LLM_START_CMD="docker compose up -d llama-server"
-else
-    LLM_PORT="${OLLAMA_PORT:-${LLAMA_SERVER_PORT:-11434}}"
-    EXTERNAL_PORT=$(docker port dream-llama-server 8080/tcp 2>/dev/null | head -1 | cut -d: -f2 || echo "$LLM_PORT")
-    LLM_ENDPOINTS=("http://${SERVICE_HOST}:${EXTERNAL_PORT}" "http://localhost:${EXTERNAL_PORT}" "http://localhost:${LLM_PORT}")
-    LLM_SERVICE_NAME="llama-server"
-    LLM_START_CMD="docker compose up -d llama-server"
-fi
+LLM_PORT="${OLLAMA_PORT:-${LLAMA_SERVER_PORT:-8080}}"
+# Also probe the actual mapped port in case docker remapped it
+EXTERNAL_PORT=$(docker port dream-llama-server 8080/tcp 2>/dev/null | head -1 | cut -d: -f2 || echo "$LLM_PORT")
+LLM_ENDPOINTS=("http://${SERVICE_HOST}:${EXTERNAL_PORT}" "http://localhost:${EXTERNAL_PORT}" "http://localhost:${LLM_PORT}")
+LLM_SERVICE_NAME="llama-server"
+LLM_START_CMD="docker compose up -d llama-server"
 
 LLM_FOUND=false
 for ENDPOINT in "${LLM_ENDPOINTS[@]}"; do

dream-server/tests/test-preflight.sh

@@ -0,0 +1,156 @@
+#!/bin/bash
+# ============================================================================
+# Dream Server dream-preflight.sh Test Suite
+# ============================================================================
+# Validates that dream-preflight.sh is syntactically correct, follows the
+# project's shell style requirements, and encodes the correct LLM port
+# default (8080, not 11434).
+#
+# These tests are static (no running Docker) — they inspect the script itself.
+# Integration-level tests (actual service probing) are covered in
+# tests/test-health-check.sh and tests/test-integration.sh.
+#
+# Usage: bash tests/test-preflight.sh
+# Exit codes: 0 = all pass, 1 = one or more failures
+# ============================================================================
+
+set -euo pipefail
+
+SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
+ROOT_DIR="$(cd "$SCRIPT_DIR/.." && pwd)"
+PREFLIGHT="$ROOT_DIR/dream-preflight.sh"
+
+GREEN='\033[0;32m'
+RED='\033[0;31m'
+YELLOW='\033[1;33m'
+NC='\033[0m'
+
+PASSED=0
+FAILED=0
+
+pass() { printf "  ${GREEN}✓ PASS${NC} %s\n" "$1"; PASSED=$((PASSED + 1)); }
+fail() { printf "  ${RED}✗ FAIL${NC} %s\n" "$1"; FAILED=$((FAILED + 1)); }
+skip() { printf "  ${YELLOW}⊘ SKIP${NC} %s\n" "$1"; }
+
+echo ""
+echo "╔════════════════════════════════════════════════╗"
+echo "║   dream-preflight.sh Test Suite               ║"
+echo "╚════════════════════════════════════════════════╝"
+echo ""
+
+# ── Static checks ──────────────────────────────────────────────────────────
+
+# 1. Script exists
+if [[ ! -f "$PREFLIGHT" ]]; then
+    fail "dream-preflight.sh not found at $PREFLIGHT"
+    echo ""
+    echo "Result: $PASSED passed, $FAILED failed"
+    exit 1
+fi
+pass "dream-preflight.sh exists"
+
+# 2. Script passes bash syntax check
+if bash -n "$PREFLIGHT" 2>/dev/null; then
+    pass "bash -n syntax check passes"
+else
+    fail "bash -n syntax check failed"
+fi
+
+# 3. Uses set -euo pipefail (project style requirement per CONTRIBUTING.md)
+if grep -q 'set -euo pipefail' "$PREFLIGHT"; then
+    pass "set -euo pipefail present"
+else
+    fail "set -euo pipefail missing — CONTRIBUTING.md requires this in all bash files"
+fi
+
+# 4. LLM port default is 8080, NOT 11434
+# config/ports.json and docker-compose.base.yml define the canonical default as 8080.
+# 11434 is only the Strix Halo AMD override (written to .env by phase 06).
+# The fallback in the script must be 8080 so plain installs without OLLAMA_PORT in
+# .env get the correct port probed.
+if grep -q 'LLAMA_SERVER_PORT:-8080' "$PREFLIGHT"; then
+    pass "LLM port fallback is 8080 (aligns with config/ports.json)"
+else
+    fail "LLM port fallback is not 8080 — check the OLLAMA_PORT expansion in dream-preflight.sh"
+fi
+
+# 5. Does NOT contain the old wrong fallback of 11434 in port resolution
+if grep -q 'LLAMA_SERVER_PORT:-11434' "$PREFLIGHT"; then
+    fail "Old wrong LLM port fallback 11434 still present — should be 8080"
+else
+    pass "Old wrong LLM port fallback 11434 is gone"
+fi
+
+# 6. detect_backend function is present
+if grep -q 'detect_backend()' "$PREFLIGHT"; then
+    pass "detect_backend function present"
+else
+    fail "detect_backend function missing"
+fi
+
+# 7. AMD sysfs scan iterates all DRM cards (not just card1)
+if grep -q '/sys/class/drm/card\*/device' "$PREFLIGHT"; then
+    pass "AMD sysfs scan uses glob (all DRM cards)"
+else
+    fail "AMD sysfs scan missing wildcard — may miss some AMD GPUs"
+fi
+
+# 8. Script uses BASH_SOURCE for portability (not $0)
+if grep -q 'BASH_SOURCE' "$PREFLIGHT"; then
+    pass "Uses BASH_SOURCE for script dir resolution"
+else
+    fail "Missing BASH_SOURCE — \$0 breaks when script is sourced"
+fi
+
+# 9. Summary section has all expected check sections
+for check_label in "Checking Docker" "Checking Docker Compose" "Checking GPU" \
+                   "Checking LLM endpoint" "Checking Whisper" "Checking TTS" \
+                   "Checking Embeddings" "Checking Dashboard"; do
+    if grep -q "$check_label" "$PREFLIGHT"; then
+        pass "Check section present: $check_label"
+    else
+        fail "Check section missing: $check_label"
+    fi
+done
+
+# 10. Uses docker port to probe actual external port mapping (not just hardcoded)
+if grep -q 'docker port dream-llama-server' "$PREFLIGHT"; then
+    pass "Probes actual Docker port mapping via 'docker port'"
+else
+    fail "Does not probe actual Docker port mapping"
+fi
+
+# ── Runtime smoke test (no Docker required) ─────────────────────────────────
+
+# 11. Script runs to completion without unbound variable or syntax errors
+#     (Services won't be up, so we expect exit 1 — that is correct behavior)
+set +e
+err_output=$(
+    SERVICE_HOST=localhost \
+    OLLAMA_PORT="" \
+    LLAMA_SERVER_PORT="" \
+    GPU_BACKEND="" \
+    bash "$PREFLIGHT" 2>&1
+)
+run_exit=$?
+set -e
+
+if echo "$err_output" | grep -qiE 'unbound variable|syntax error|command not found: \['; then
+    fail "Script produced shell error: $(echo "$err_output" | grep -iE 'unbound variable|syntax error' | head -1)"
+else
+    pass "Script runs without shell errors (exit $run_exit is expected)"
+fi
+
+# 12. Exit code is 0 or 1; never an unexpected crash code
+if [[ "$run_exit" -eq 0 ]] || [[ "$run_exit" -eq 1 ]]; then
+    pass "Exit code is valid (0=pass, 1=fail): $run_exit"
+else
+    fail "Unexpected exit code $run_exit — script may have crashed"
+fi
+
+# ── Summary ────────────────────────────────────────────────────────────────
+
+echo ""
+echo "Result: $PASSED passed, $FAILED failed"
+echo ""
+[[ $FAILED -eq 0 ]]