feat: add messagelist test infrastructure and tests

- Add domaintest factories (NewMessage, NewTextBlock, etc.)
- Add messagelisttest helper with real SQLite, mock chat client
- Add mouse_test.go: selection math tests
- Add update_test.go: message dispatch ordering
- Add turn_test.go: state machine transitions
- Add user_test.go: rendering width contracts
- Add tool_test.go: status rendering for all states
- Update TESTING.md with real vs mock guidance
- Fix: thinking animation not stopping after stream completed
- Fix: copy/highlight including border characters

Author: binarylogic

Taskfile.yml

@@ -53,7 +53,7 @@ tasks:
     cmds:
       - go generate ./internal/powersync
       - go generate ./internal/sqlite
-      - cp ../control-plane/internal/powersync/schema.sql internal/chat/tools/query_schema.sql
+      - sed '/^-- Auto-generated/d; /^-- Run .task generate/d' ../control-plane/internal/powersync/schema.sql > internal/chat/tools/query_schema.sql
 
   # ===========================================================================
   # Build

docs/TESTING.md

@@ -4,7 +4,7 @@ Test behavior, not implementation. A good test breaks when something stops worki
 
 ## Three Types of Tests
 
-**Unit tests** run fast with no external dependencies. They mock interfaces and test logic in isolation.
+**Unit tests** run fast with no external services. Use real local deps (SQLite, theme), mock remote ones (APIs).
 
 ```bash
 task test
@@ -47,9 +47,23 @@ func TestSync_Connect(t *testing.T) {
 }
 ```
 
+### Real vs Mock
+
+Use real things when they're cheap. Mock only what's expensive or external.
+
+| Dependency | Strategy | Why |
+|------------|----------|-----|
+| SQLite | Real — `sqlitetest.OpenBareDB(t)` or `dbtest.OpenTestDB(t)` | Local, fast, tmpdir-based, auto-cleanup |
+| Theme | Real — `styles.NewTheme(true)` | Just a struct with colors |
+| Logger | Real — `logtest.NewScope(t)` | Writes to `testing.T`, shows on failure |
+| Chat API | Mock — `chattest.MockClient{}` | Hits remote server |
+| Tool registry | `nil` or partial — `&chattools.Registry{Query: ...}` | Only construct what you're testing |
+
+The rule: if you can construct it in a test without network, credentials, or slow setup, use the real thing. Mocks hide bugs.
+
 ### Mock with Structs
 
-Don't use mocking frameworks. Write simple structs:
+When you do need mocks, don't use mocking frameworks. Write simple structs:
 
 ```go
 type mockClient struct {

internal/api/datadog_account_service.go

@@ -85,7 +85,11 @@ type DatadogAccountStatus struct {
 	BrokenServices      int
 	DisabledServices    int
 	InactiveServices    int
-	SavedCount          int  // Number of log events saved (for progress display)
+	AnalyzedCount       int  // Number of log events analyzed (for progress display)
+	ResolvedCount       int  // Log events with all policies acted on
+	CleanCount          int  // Log events analyzed with no issues
+	PendingCount        int  // Log events with policies awaiting action
+	PendingPolicyCount  int  // Policies awaiting user action
 	ReadyForUse         bool // Whether the account has enough data to proceed
 }
 
@@ -232,7 +236,11 @@ func (s *DatadogAccountService) GetStatus(ctx context.Context, datadogAccountID
 		BrokenServices:      statusNode.LogBrokenServices,
 		DisabledServices:    statusNode.LogDisabledServices,
 		InactiveServices:    statusNode.LogInactiveServices,
-		SavedCount:          statusNode.LogSavedCount,
+		AnalyzedCount:       statusNode.LogAnalyzedCount,
+		ResolvedCount:       statusNode.LogResolvedCount,
+		CleanCount:          statusNode.LogCleanCount,
+		PendingCount:        statusNode.LogPendingCount,
+		PendingPolicyCount:  statusNode.LogPendingPolicyCount,
 		ReadyForUse:         statusNode.ReadyForUse,
 	}
 

internal/api/gen/generated.go

@@ -51,7 +51,11 @@ query GetDatadogAccountStatus($id: ID!) {
                     logBrokenServices
                     logDisabledServices
                     logInactiveServices
-                    logSavedCount
+                    logAnalyzedCount
+                    logResolvedCount
+                    logCleanCount
+                    logPendingCount
+                    logPendingPolicyCount
                     readyForUse
                 }
             }