|
| 1 | +package e2e |
| 2 | + |
| 3 | +import ( |
| 4 | + "context" |
| 5 | + "encoding/base64" |
| 6 | + "net/http" |
| 7 | + "os/exec" |
| 8 | + "strings" |
| 9 | + "testing" |
| 10 | + "time" |
| 11 | + |
| 12 | + instanceoapi "github.com/onkernel/kernel-images/server/lib/oapi" |
| 13 | + "github.com/stretchr/testify/require" |
| 14 | +) |
| 15 | + |
| 16 | +// TestAgentBrowserCDPProxy tests that agent-browser can connect to Chrome via the CDP proxy on port 9222. |
| 17 | +// This validates the /json and /json/list endpoints that the proxy exposes for target discovery, |
| 18 | +// which is required for tools like agent-browser and Playwright's connectOverCDP. |
| 19 | +func TestAgentBrowserCDPProxy(t *testing.T) { |
| 20 | + t.Parallel() |
| 21 | + |
| 22 | + if _, err := exec.LookPath("docker"); err != nil { |
| 23 | + t.Skipf("docker not available: %v", err) |
| 24 | + } |
| 25 | + |
| 26 | + ctx, cancel := context.WithTimeout(context.Background(), 5*time.Minute) |
| 27 | + defer cancel() |
| 28 | + |
| 29 | + c := NewTestContainer(t, headlessImage) |
| 30 | + require.NoError(t, c.Start(ctx, ContainerConfig{}), "failed to start container") |
| 31 | + defer c.Stop(ctx) |
| 32 | + |
| 33 | + require.NoError(t, c.WaitReady(ctx), "api not ready") |
| 34 | + require.NoError(t, c.WaitDevTools(ctx), "devtools not ready") |
| 35 | + |
| 36 | + client, err := c.APIClient() |
| 37 | + require.NoError(t, err) |
| 38 | + |
| 39 | + // Install agent-browser globally inside the container |
| 40 | + t.Log("Installing agent-browser...") |
| 41 | + timeoutSec := 120 // npm install can take a while |
| 42 | + installResult := execCommandWithTimeout(t, ctx, client, "npm", []string{"install", "-g", "agent-browser"}, &timeoutSec) |
| 43 | + require.Zero(t, installResult.exitCode, "failed to install agent-browser: %s", installResult.output) |
| 44 | + t.Log("agent-browser installed successfully") |
| 45 | + |
| 46 | + // First test the /json endpoints via curl to verify the proxy is working correctly |
| 47 | + // before we test agent-browser |
| 48 | + |
| 49 | + // Test that /json endpoint returns proper target list with rewritten URLs |
| 50 | + t.Run("json endpoint returns targets with rewritten URLs", func(t *testing.T) { |
| 51 | + t.Log("Testing /json endpoint via curl") |
| 52 | + |
| 53 | + result := execCommand(t, ctx, client, "curl", []string{"-s", "http://127.0.0.1:9222/json"}) |
| 54 | + require.Zero(t, result.exitCode, "curl /json failed: %s", result.output) |
| 55 | + |
| 56 | + // The response should be a JSON array containing targets |
| 57 | + require.True(t, strings.HasPrefix(strings.TrimSpace(result.output), "["), |
| 58 | + "expected JSON array from /json, got: %s", result.output) |
| 59 | + |
| 60 | + // The URLs should point to the proxy (port 9222), not Chrome directly (port 9223) |
| 61 | + require.Contains(t, result.output, "9222", |
| 62 | + "expected target URLs to be rewritten to proxy port 9222, got: %s", result.output) |
| 63 | + require.NotContains(t, result.output, "9223", |
| 64 | + "target URLs should not contain Chrome port 9223, got: %s", result.output) |
| 65 | + }) |
| 66 | + |
| 67 | + // Test that /json/list endpoint also works |
| 68 | + t.Run("json/list endpoint returns targets with rewritten URLs", func(t *testing.T) { |
| 69 | + t.Log("Testing /json/list endpoint via curl") |
| 70 | + |
| 71 | + result := execCommand(t, ctx, client, "curl", []string{"-s", "http://127.0.0.1:9222/json/list"}) |
| 72 | + require.Zero(t, result.exitCode, "curl /json/list failed: %s", result.output) |
| 73 | + |
| 74 | + // The response should be a JSON array containing targets |
| 75 | + require.True(t, strings.HasPrefix(strings.TrimSpace(result.output), "["), |
| 76 | + "expected JSON array from /json/list, got: %s", result.output) |
| 77 | + |
| 78 | + // The URLs should point to the proxy (port 9222), not Chrome directly (port 9223) |
| 79 | + require.Contains(t, result.output, "9222", |
| 80 | + "expected target URLs to be rewritten to proxy port 9222, got: %s", result.output) |
| 81 | + require.NotContains(t, result.output, "9223", |
| 82 | + "target URLs should not contain Chrome port 9223, got: %s", result.output) |
| 83 | + }) |
| 84 | + |
| 85 | + // Test that /json/version endpoint works (this was already there) |
| 86 | + t.Run("json/version endpoint works", func(t *testing.T) { |
| 87 | + t.Log("Testing /json/version endpoint via curl") |
| 88 | + |
| 89 | + result := execCommand(t, ctx, client, "curl", []string{"-s", "http://127.0.0.1:9222/json/version"}) |
| 90 | + require.Zero(t, result.exitCode, "curl /json/version failed: %s", result.output) |
| 91 | + |
| 92 | + // The response should be a JSON object with browser info |
| 93 | + require.True(t, strings.HasPrefix(strings.TrimSpace(result.output), "{"), |
| 94 | + "expected JSON object from /json/version, got: %s", result.output) |
| 95 | + |
| 96 | + // Should contain webSocketDebuggerUrl pointing to proxy |
| 97 | + require.Contains(t, result.output, "webSocketDebuggerUrl", |
| 98 | + "expected webSocketDebuggerUrl in response, got: %s", result.output) |
| 99 | + require.Contains(t, result.output, "9222", |
| 100 | + "expected webSocketDebuggerUrl to point to proxy port 9222, got: %s", result.output) |
| 101 | + }) |
| 102 | + |
| 103 | + // Now test agent-browser with different CDP connection variations |
| 104 | + // Each test connects to the browser, navigates, and gets the URL to verify connectivity |
| 105 | + |
| 106 | + testCases := []struct { |
| 107 | + name string |
| 108 | + cdpArg string |
| 109 | + }{ |
| 110 | + { |
| 111 | + name: "port only (9222)", |
| 112 | + cdpArg: "9222", |
| 113 | + }, |
| 114 | + { |
| 115 | + name: "http URL", |
| 116 | + cdpArg: "http://127.0.0.1:9222", |
| 117 | + }, |
| 118 | + { |
| 119 | + name: "localhost:port", |
| 120 | + cdpArg: "localhost:9222", |
| 121 | + }, |
| 122 | + { |
| 123 | + name: "127.0.0.1:port", |
| 124 | + cdpArg: "127.0.0.1:9222", |
| 125 | + }, |
| 126 | + } |
| 127 | + |
| 128 | + for _, tc := range testCases { |
| 129 | + tc := tc // capture range variable |
| 130 | + t.Run("agent-browser connect "+tc.name, func(t *testing.T) { |
| 131 | + t.Logf("Testing agent-browser with --cdp %s", tc.cdpArg) |
| 132 | + |
| 133 | + // Navigate to example.com - this is the key test that verifies: |
| 134 | + // 1. agent-browser can connect to the CDP proxy on port 9222 |
| 135 | + // 2. The proxy's /json endpoint works for target discovery |
| 136 | + // 3. The WebSocket connection through the proxy works |
| 137 | + navResult := execCommand(t, ctx, client, "agent-browser", []string{"--cdp", tc.cdpArg, "open", "https://example.com"}) |
| 138 | + t.Logf("Navigate result: exit=%d, output=%s", navResult.exitCode, navResult.output) |
| 139 | + require.Zero(t, navResult.exitCode, "agent-browser --cdp %s open failed: %s", tc.cdpArg, navResult.output) |
| 140 | + |
| 141 | + // Get the current URL to verify we're connected and navigation worked |
| 142 | + urlResult := execCommand(t, ctx, client, "agent-browser", []string{"--cdp", tc.cdpArg, "get", "url", "--json"}) |
| 143 | + t.Logf("Get URL result: exit=%d, output=%s", urlResult.exitCode, urlResult.output) |
| 144 | + require.Zero(t, urlResult.exitCode, "agent-browser --cdp %s get url failed: %s", tc.cdpArg, urlResult.output) |
| 145 | + |
| 146 | + // Verify we got a valid response containing example.com |
| 147 | + require.Contains(t, urlResult.output, "example.com", |
| 148 | + "expected URL to contain example.com, got: %s", urlResult.output) |
| 149 | + }) |
| 150 | + } |
| 151 | + |
| 152 | + // Test agent-browser snapshot command which uses /json for target discovery |
| 153 | + t.Run("agent-browser snapshot via proxy", func(t *testing.T) { |
| 154 | + t.Log("Testing agent-browser snapshot via CDP proxy") |
| 155 | + |
| 156 | + // Get snapshot - this exercises the /json endpoint to discover targets |
| 157 | + result := execCommand(t, ctx, client, "agent-browser", []string{"--cdp", "9222", "snapshot", "-i", "--json"}) |
| 158 | + t.Logf("Snapshot exit code: %d, Output length: %d", result.exitCode, len(result.output)) |
| 159 | + |
| 160 | + require.Zero(t, result.exitCode, "agent-browser snapshot failed: %s", result.output) |
| 161 | + // Verify we got a valid snapshot response |
| 162 | + require.True(t, strings.Contains(result.output, "success") || strings.Contains(result.output, "snapshot") || strings.Contains(result.output, "data"), |
| 163 | + "expected valid snapshot response, got: %s", result.output) |
| 164 | + }) |
| 165 | + |
| 166 | + // Test agent-browser get title to further verify connectivity |
| 167 | + t.Run("agent-browser get title via proxy", func(t *testing.T) { |
| 168 | + t.Log("Testing agent-browser get title via CDP proxy") |
| 169 | + |
| 170 | + result := execCommand(t, ctx, client, "agent-browser", []string{"--cdp", "9222", "get", "title", "--json"}) |
| 171 | + t.Logf("Get title result: exit=%d, output=%s", result.exitCode, result.output) |
| 172 | + |
| 173 | + require.Zero(t, result.exitCode, "agent-browser get title failed: %s", result.output) |
| 174 | + // Should contain "Example Domain" from previous navigation |
| 175 | + require.Contains(t, result.output, "Example", |
| 176 | + "expected title to contain 'Example', got: %s", result.output) |
| 177 | + }) |
| 178 | + |
| 179 | + t.Log("All agent-browser CDP proxy tests passed") |
| 180 | +} |
| 181 | + |
| 182 | +// execResult holds the result of a command execution |
| 183 | +type execResult struct { |
| 184 | + exitCode int |
| 185 | + output string |
| 186 | +} |
| 187 | + |
| 188 | +// execCommand runs a command via the container's process exec API and returns the result |
| 189 | +func execCommand(t *testing.T, ctx context.Context, client *instanceoapi.ClientWithResponses, command string, args []string) execResult { |
| 190 | + t.Helper() |
| 191 | + return execCommandWithTimeout(t, ctx, client, command, args, nil) |
| 192 | +} |
| 193 | + |
| 194 | +// execCommandWithTimeout runs a command with an optional timeout |
| 195 | +func execCommandWithTimeout(t *testing.T, ctx context.Context, client *instanceoapi.ClientWithResponses, command string, args []string, timeoutSec *int) execResult { |
| 196 | + t.Helper() |
| 197 | + |
| 198 | + req := instanceoapi.ProcessExecJSONRequestBody{ |
| 199 | + Command: command, |
| 200 | + Args: &args, |
| 201 | + TimeoutSec: timeoutSec, |
| 202 | + } |
| 203 | + |
| 204 | + rsp, err := client.ProcessExecWithResponse(ctx, req) |
| 205 | + require.NoError(t, err, "process exec request error for %s", command) |
| 206 | + require.Equal(t, http.StatusOK, rsp.StatusCode(), "unexpected status for %s: %s body=%s", |
| 207 | + command, rsp.Status(), string(rsp.Body)) |
| 208 | + require.NotNil(t, rsp.JSON200, "expected JSON200 response for %s", command) |
| 209 | + |
| 210 | + var stdout, stderr string |
| 211 | + if rsp.JSON200.StdoutB64 != nil && *rsp.JSON200.StdoutB64 != "" { |
| 212 | + if b, decErr := base64.StdEncoding.DecodeString(*rsp.JSON200.StdoutB64); decErr == nil { |
| 213 | + stdout = string(b) |
| 214 | + } |
| 215 | + } |
| 216 | + if rsp.JSON200.StderrB64 != nil && *rsp.JSON200.StderrB64 != "" { |
| 217 | + if b, decErr := base64.StdEncoding.DecodeString(*rsp.JSON200.StderrB64); decErr == nil { |
| 218 | + stderr = string(b) |
| 219 | + } |
| 220 | + } |
| 221 | + |
| 222 | + exitCode := 0 |
| 223 | + if rsp.JSON200.ExitCode != nil { |
| 224 | + exitCode = *rsp.JSON200.ExitCode |
| 225 | + } |
| 226 | + |
| 227 | + return execResult{ |
| 228 | + exitCode: exitCode, |
| 229 | + output: stdout + stderr, |
| 230 | + } |
| 231 | +} |
0 commit comments