test: add repeatable E2E model tests for deepseek-chat, v4-flash, v4-pro, and gpt-4o-mini

Creates cmd/kode/model_e2e_test.go with integration-tagged Go tests that
exercise the full kode agent loop against real LLM APIs.

- TestModelE2E_DeepSeek_Chat
- TestModelE2E_DeepSeek_V4Flash
- TestModelE2E_DeepSeek_V4Pro
- TestModelE2E_OpenAI_GPT4oMini

Each test:
  - Creates a real kode.Agent with the model config
  - Runs the prompt 'Respond with exactly the word ALIVE'
  - Verifies response contains ALIVE, non-zero token tracking
  - Uses 120s timeout per test, 15 max iterations
  - Gated by //go:build integration tag; skipped in -short mode

Run: go test -tags=integration -v -count=1 -run TestModelE2E ./cmd/kode/
Keys: KODE_API_KEY_DEEPSEEK or KODE_API_KEY_OPENAI env vars

Author: jkyberneees

cmd/kode/model_e2e_test.go

@@ -0,0 +1,115 @@
+//go:build integration
+
+// E2E tests that exercise the full kode agent loop against real LLM APIs.
+//
+// These tests are gated by the "integration" build tag and require API keys
+// to be set via environment variables:
+//
+//	export KODE_API_KEY_DEEPSEEK="sk-..."   # required for DeepSeek models
+//	export KODE_API_KEY_OPENAI="sk-..."     # required for OpenAI models
+//	go test -tags=integration -v -count=1 -run 'TestModelE2E' ./cmd/kode/
+//
+// Each test case creates a real kode.Agent with the given model config,
+// runs a simple task ("respond with ALIVE"), and verifies:
+//   - The agent loop completes without error
+//   - The response contains the expected substring
+//   - Token tracking produces non-zero values
+//   - The loop runs within a 120s timeout
+
+package main
+
+import (
+	"context"
+	"os"
+	"strings"
+	"testing"
+	"time"
+
+	"github.com/BackendStack21/kode"
+	"github.com/BackendStack21/kode/internal/render"
+)
+
+type modelCase struct {
+	Name    string
+	Model   string
+	BaseURL string
+	KeyEnv  string // env var name for the API key
+}
+
+// ── DeepSeek models ────────────────────────────────────────────────────
+
+func TestModelE2E_DeepSeek_Chat(t *testing.T)   { testModel(t, modelCase{Name: "deepseek-chat", Model: "deepseek-chat", KeyEnv: "KODE_API_KEY_DEEPSEEK"}) }
+func TestModelE2E_DeepSeek_V4Flash(t *testing.T) { testModel(t, modelCase{Name: "deepseek-v4-flash", Model: "deepseek-v4-flash", KeyEnv: "KODE_API_KEY_DEEPSEEK"}) }
+func TestModelE2E_DeepSeek_V4Pro(t *testing.T)   { testModel(t, modelCase{Name: "deepseek-v4-pro", Model: "deepseek-v4-pro", KeyEnv: "KODE_API_KEY_DEEPSEEK"}) }
+
+// ── OpenAI models ──────────────────────────────────────────────────────
+
+func TestModelE2E_OpenAI_GPT4oMini(t *testing.T) {
+	testModel(t, modelCase{
+		Name:    "gpt-4o-mini",
+		Model:   "gpt-4o-mini",
+		BaseURL: "https://api.openai.com/v1",
+		KeyEnv:  "KODE_API_KEY_OPENAI",
+	})
+}
+
+// testModel runs a single E2E model test: creates an agent, runs a prompt,
+// and verifies the response contains "ALIVE".
+func testModel(t *testing.T, mc modelCase) {
+	t.Helper()
+
+	if testing.Short() {
+		t.Skip("skipping E2E model test in short mode")
+	}
+
+	apiKey := os.Getenv(mc.KeyEnv)
+	if apiKey == "" {
+		t.Skipf("env %s not set — skipping %s", mc.KeyEnv, mc.Name)
+	}
+
+	baseURL := mc.BaseURL
+	if baseURL == "" {
+		baseURL = "https://api.deepseek.com/v1"
+	}
+
+	agent, err := kode.New(kode.Config{
+		Model:         mc.Model,
+		BaseURL:       baseURL,
+		APIKey:        apiKey,
+		MaxIterations: 15,
+		SystemMessage: "You are a helpful assistant. Follow instructions precisely.",
+		NoProjectFile: true,
+		Renderer:      render.New(os.Stderr, false),
+	})
+	if err != nil {
+		t.Fatalf("kode.New: %v", err)
+	}
+	defer agent.Close()
+
+	ctx, cancel := context.WithTimeout(context.Background(), 120*time.Second)
+	defer cancel()
+
+	start := time.Now()
+	answer, err := agent.Run(ctx, "Respond with exactly the word 'ALIVE' and nothing else.")
+	elapsed := time.Since(start)
+
+	if err != nil {
+		t.Fatalf("Agent.Run: %v (after %v)", err, elapsed)
+	}
+
+	inputTok := agent.TotalInputTokens()
+	outputTok := agent.TotalOutputTokens()
+
+	t.Logf("✅ %-20s  latency=%v  input=%-5d  output=%-5d  total=%-5d",
+		mc.Name, elapsed.Round(time.Millisecond), inputTok, outputTok, inputTok+outputTok)
+
+	if !strings.Contains(answer, "ALIVE") {
+		t.Errorf("response = %q, want substring %q", strings.TrimSpace(answer), "ALIVE")
+	}
+	if inputTok <= 0 {
+		t.Error("TotalInputTokens = 0, expected > 0")
+	}
+	if outputTok <= 0 {
+		t.Error("TotalOutputTokens = 0, expected > 0")
+	}
+}