feat: add diff filtering for large staged changes

README.md

@@ -73,6 +73,9 @@ Supported settings:
 - `prompt` Bundled prompt preset: `default`, `conventional`, `gitmoji`, `karma`
 - `prompt_file` Path to a custom prompt file (relative to the config file)
 - `engines.<name>.args` Argument list for the engine command (array of strings)
+- `filter.max_file_lines` Maximum lines per file in diff (default: 100)
+- `filter.exclude_patterns` Additional glob patterns to exclude from diff
+- `filter.default_exclude_patterns` Override built-in exclude patterns
 
 ### Engines
 
@@ -117,6 +120,38 @@ prompt_file = "prompts/commit.md"
 
 Note: `prompt` and `prompt_file` are mutually exclusive within the same config file. If both are set, an error is returned. When settings come from different layers (user config vs repo config), the later layer wins.
 
+### Diff Filtering
+
+When the staged diff is large, it can exceed LLM context limits or degrade commit message quality. git-ai-commit automatically filters the diff to help LLMs focus on meaningful changes.
+
+**Default behavior:**
+
+- Each file is limited to 100 lines (configurable via `filter.max_file_lines`)
+- Lock files and generated files are excluded by default
+
+**Default exclude patterns:**
+
+- `**/*.lock`, `**/*-lock.json`, `**/*.lock.yaml`, `**/*-lock.yaml`, `**/*.lockfile`
+- `**/*.min.js`, `**/*.min.css`, `**/*.map`
+- `**/go.sum`
+
+Example: Customize filtering
+
+```toml
+[filter]
+max_file_lines = 300
+
+# Add patterns to exclude (merged with defaults)
+exclude_patterns = [
+    "**/vendor/**",
+    "**/*.generated.go",
+    "**/dist/**"
+]
+
+# Or replace defaults entirely
+# default_exclude_patterns = ["**/my-lock.json"]
+```
+
 ## Claude Code Plugin
 
 If you use [Claude Code](https://docs.anthropic.com/en/docs/claude-code), you can integrate git-ai-commit as a plugin for a more convenient workflow.

internal/app/app.go

@@ -45,7 +45,7 @@ func Run(context, contextFile, promptName, promptFile, engineName string, amend,
 		}
 	}
 
-	diff, err := commitDiff(amend)
+	diff, err := commitDiff(amend, cfg)
 	if err != nil {
 		return err
 	}
@@ -141,11 +141,55 @@ func sanitizeMessage(message string) string {
 	return clean
 }
 
-func commitDiff(amend bool) (string, error) {
+func commitDiff(amend bool, cfg config.Config) (string, error) {
+	var diff string
+	var err error
 	if amend {
-		return git.LastCommitDiff()
+		diff, err = git.LastCommitDiff()
+	} else {
+		diff, err = git.StagedDiff()
 	}
-	return git.StagedDiff()
+	if err != nil {
+		return "", err
+	}
+
+	// Determine exclude patterns
+	patterns := git.DefaultExcludePatterns()
+	if len(cfg.Filter.DefaultExcludePatterns) > 0 {
+		patterns = cfg.Filter.DefaultExcludePatterns
+	}
+	patterns = append(patterns, cfg.Filter.ExcludePatterns...)
+
+	// Determine max file lines
+	maxLines := cfg.Filter.MaxFileLines
+	if maxLines == 0 {
+		maxLines = git.DefaultMaxFileLines
+	}
+
+	opts := git.Options{
+		MaxFileLines:    maxLines,
+		ExcludePatterns: patterns,
+	}
+	result := git.Filter(diff, opts)
+
+	if result.Truncated || len(result.ExcludedFiles) > 0 {
+		return result.Diff + formatFilterNotice(result), nil
+	}
+	return result.Diff, nil
+}
+
+func formatFilterNotice(result git.Result) string {
+	var parts []string
+	if len(result.ExcludedFiles) > 0 {
+		parts = append(parts, fmt.Sprintf("Excluded files: %s", strings.Join(result.ExcludedFiles, ", ")))
+	}
+	if len(result.TruncatedFiles) > 0 {
+		parts = append(parts, fmt.Sprintf("Truncated files: %s", strings.Join(result.TruncatedFiles, ", ")))
+	}
+	if len(parts) == 0 {
+		return ""
+	}
+	return "\n\n[Filter notice: " + strings.Join(parts, "; ") + "]"
 }
 
 func stageChanges(addAll bool, includeFiles []string) (func(), error) {

internal/config/config.go

@@ -17,18 +17,27 @@ type Config struct {
 	Prompt        string                  `toml:"prompt"`
 	PromptFile    string                  `toml:"prompt_file"`
 	Engines       map[string]EngineConfig `toml:"engines"`
+	Filter        FilterConfig            `toml:"filter"`
 
 	// ResolvedPrompt holds the final prompt text after loading from preset or file.
 	// This is not read from config files directly.
 	ResolvedPrompt string `toml:"-"`
 }
 
+// FilterConfig holds diff filtering configuration.
+type FilterConfig struct {
+	MaxFileLines           int      `toml:"max_file_lines"`            // Max lines per file (0 = use default)
+	DefaultExcludePatterns []string `toml:"default_exclude_patterns"`  // Override built-in defaults
+	ExcludePatterns        []string `toml:"exclude_patterns"`          // Additional patterns to exclude
+}
+
 // rawConfig is the TOML structure used to detect mutual exclusivity in a single layer.
 type rawConfig struct {
 	DefaultEngine string                  `toml:"engine"`
 	Prompt        string                  `toml:"prompt"`
 	PromptFile    string                  `toml:"prompt_file"`
 	Engines       map[string]EngineConfig `toml:"engines"`
+	Filter        FilterConfig            `toml:"filter"`
 }
 
 type EngineConfig struct {
@@ -129,6 +138,16 @@ func Load() (Config, error) {
 					cfg.Engines[name] = ec
 				}
 			}
+			// Merge filter config from repo
+			if repoCfg.Filter.MaxFileLines != 0 {
+				cfg.Filter.MaxFileLines = repoCfg.Filter.MaxFileLines
+			}
+			if len(repoCfg.Filter.DefaultExcludePatterns) > 0 {
+				cfg.Filter.DefaultExcludePatterns = repoCfg.Filter.DefaultExcludePatterns
+			}
+			if len(repoCfg.Filter.ExcludePatterns) > 0 {
+				cfg.Filter.ExcludePatterns = append(cfg.Filter.ExcludePatterns, repoCfg.Filter.ExcludePatterns...)
+			}
 		}
 	}
 
@@ -181,6 +200,16 @@ func loadConfigLayer(data []byte, cfg *Config, source string) error {
 			cfg.Engines[name] = ec
 		}
 	}
+	// Merge filter config
+	if raw.Filter.MaxFileLines != 0 {
+		cfg.Filter.MaxFileLines = raw.Filter.MaxFileLines
+	}
+	if len(raw.Filter.DefaultExcludePatterns) > 0 {
+		cfg.Filter.DefaultExcludePatterns = raw.Filter.DefaultExcludePatterns
+	}
+	if len(raw.Filter.ExcludePatterns) > 0 {
+		cfg.Filter.ExcludePatterns = append(cfg.Filter.ExcludePatterns, raw.Filter.ExcludePatterns...)
+	}
 	return nil
 }