Usability improvements

gpt_engineer/applications/cli/file_selector.py

@@ -76,7 +76,7 @@ def __init__(self, project_path: Union[str, Path]):
         self.metadata_db = DiskMemory(metadata_path(self.project_path))
         self.toml_path = self.metadata_db.path / self.FILE_LIST_NAME
 
-    def ask_for_files(self) -> tuple[FilesDict, bool]:
+    def ask_for_files(self, skip_file_selection=False) -> tuple[FilesDict, bool]:
         """
         Prompts the user to select files for context improvement.
 
@@ -89,8 +89,9 @@ def ask_for_files(self) -> tuple[FilesDict, bool]:
             A dictionary with file paths as keys and file contents as values.
         """
 
-        if os.getenv("GPTE_TEST_MODE"):
+        if os.getenv("GPTE_TEST_MODE") or skip_file_selection:
             # In test mode, retrieve files from a predefined TOML configuration
+            # also get from toml if skip_file_selector is active
             assert self.FILE_LIST_NAME in self.metadata_db
             selected_files = self.get_files_from_toml(self.project_path, self.toml_path)
         else:
@@ -412,7 +413,7 @@ def get_current_files(self, project_path: Union[str, Path]) -> List[str]:
         if is_git_repo(project_path) and "projects" not in project_path.parts:
             all_files = filter_by_gitignore(project_path, all_files)
 
-        return all_files
+        return sorted(all_files, key=lambda x: Path(x).as_posix())
 
 
 class DisplayablePath(object):

gpt_engineer/applications/cli/main.py

@@ -357,6 +357,12 @@ def main(
         "--use_cache",
         help="Speeds up computations and saves tokens when running the same prompt multiple times by caching the LLM response.",
     ),
+    skip_file_selection: bool = typer.Option(
+        False,
+        "--skip-file-selection",
+        "-s",
+        help="Skip interactive file selection in improve mode and use the generated TOML file directly.",
+    ),
     no_execution: bool = typer.Option(
         False,
         "--no_execution",
@@ -405,6 +411,8 @@ def main(
         Speeds up computations and saves tokens when running the same prompt multiple times by caching the LLM response.
     verbose : bool
         Flag indicating whether to enable verbose logging.
+    skip_file_selection: bool
+        Skip interactive file selection in improve mode and use the generated TOML file directly
     no_execution: bool
         Run setup but to not call LLM or write any code. For testing purposes.
     sysinfo: bool
@@ -501,7 +509,9 @@ def main(
     files = FileStore(project_path)
     if not no_execution:
         if improve_mode:
-            files_dict_before, is_linting = FileSelector(project_path).ask_for_files()
+            files_dict_before, is_linting = FileSelector(project_path).ask_for_files(
+                skip_file_selection=skip_file_selection
+            )
 
             # lint the code
             if is_linting:

gpt_engineer/core/chat_to_files.py

@@ -138,7 +138,7 @@ def parse_diffs(diff_string: str) -> dict:
 
     diffs = {}
     try:
-        for block in diff_block_pattern.finditer(diff_string, timeout=1):
+        for block in diff_block_pattern.finditer(diff_string, timeout=10):
             diff_block = block.group()
 
             # Parse individual diff blocks and update the diffs dictionary

gpt_engineer/tools/supported_languages.py

@@ -46,7 +46,8 @@
         "extensions": [".cpp", ".cc", ".cxx", ".h", ".hpp", ".hxx"],
         "tree_sitter_name": "cpp",
     },
-    {"name": "C", "extensions": [".c", ".h"], "tree_sitter_name": "c"}
+    {"name": "C", "extensions": [".c", ".h"], "tree_sitter_name": "c"},
+    {"name": "Markdown", "extensions": [".md"], "tree_sitter_name": "md"}
     # ---- the following are not supported by the current code chunker implementation ----
     # {
     #     "name": "Swift",