llama-sidecar v0.1.0: daemon + benchmarks + eval suite

Go daemon (cmd/llama-sidecar): per-agent llama-server process pool with
LRU eviction, OpenAI-compatible proxy, flag validation (Unsloth port),
deterministic hash-keyed sidecar reuse. Windows service support via
schtasks/NSSM with DETACHED_PROCESS, stdout pipe drain, and request-ctx
decoupled child lifetime.

Bug fixes (3b.1–3b5): -c flag drop from StripShadowingFlags, UTF-8 BOM
in JSON config, -fa → --flash-attn on default, child process exit after
one request (stdin devnull, stdout pipe, CREATE_NO_WINDOW → DETACHED,
context.Background for child lifetime, background reaper goroutine).

bench/: MTP on/off throughput sweep across 8 GGUFs via SSH+schtasks
automation to sam-desktop. Per-GGUF production flags from llama-swap
config with --ctx-size 32768 override.

eval/: accuracy benchmarks (MMLU 100q, GSM8K 50q, HumanEval 164) +
A/B model comparison (14 agent-typed prompts × 8 models). All scripts
resumable at individual question level.

94 Go tests, race detector clean.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

internal/config/config_test.go

@@ -0,0 +1,79 @@
+package config
+
+import (
+	"os"
+	"path/filepath"
+	"testing"
+)
+
+func TestLoad_MissingRequired(t *testing.T) {
+	os.Unsetenv("LLAMA_SERVER_BIN")
+	os.Unsetenv("MODEL_DIR_MAP_FILE")
+	_, err := Load()
+	if err == nil {
+		t.Fatal("expected error for missing LLAMA_SERVER_BIN")
+	}
+}
+
+func TestParsePortRange(t *testing.T) {
+	lo, hi, err := parsePortRange("8500-8599")
+	if err != nil {
+		t.Fatal(err)
+	}
+	if lo != 8500 || hi != 8599 {
+		t.Fatalf("got %d-%d", lo, hi)
+	}
+}
+
+func TestParsePortRange_Bad(t *testing.T) {
+	_, _, err := parsePortRange("abc")
+	if err == nil {
+		t.Fatal("expected error")
+	}
+	_, _, err = parsePortRange("100-50")
+	if err == nil {
+		t.Fatal("expected error for hi <= lo")
+	}
+}
+
+func TestLoadModelMap_BOM(t *testing.T) {
+	dir := t.TempDir()
+	path := filepath.Join(dir, "model_map.json")
+	content := append([]byte{0xEF, 0xBB, 0xBF}, []byte(`{"test-model": "/fake/path.gguf"}`)...)
+	if err := os.WriteFile(path, content, 0644); err != nil {
+		t.Fatal(err)
+	}
+	m, err := loadModelMap(path)
+	if err != nil {
+		t.Fatalf("BOM-prefixed JSON should parse: %v", err)
+	}
+	if m["test-model"] != "/fake/path.gguf" {
+		t.Fatalf("unexpected map: %v", m)
+	}
+}
+
+func TestDefaultBaseArgs_FlashAttn(t *testing.T) {
+	args := defaultBaseArgs()
+	for i, a := range args {
+		if a == "--flash-attn" && i+1 < len(args) && args[i+1] == "on" {
+			return
+		}
+	}
+	t.Fatal("expected --flash-attn on in default args")
+}
+
+func TestDefaultBaseArgs(t *testing.T) {
+	args := defaultBaseArgs()
+	if len(args) == 0 {
+		t.Fatal("expected non-empty default args")
+	}
+	found := false
+	for _, a := range args {
+		if a == "--no-mmap" {
+			found = true
+		}
+	}
+	if !found {
+		t.Fatal("expected --no-mmap in default args")
+	}
+}