Remove delete_existing option; Introduce archive (#409)

* Remove `delete_existing` option; Introduce archive

* Update gpt_engineer/db.py

* Update gpt_engineer/main.py

* Update gpt_engineer/main.py

* Update gpt_engineer/steps.py

* Update gpt_engineer/steps.py

---------

Co-authored-by: Anton Osika <anton.osika@gmail.com>

gpt_engineer/db.py

@@ -1,5 +1,6 @@
 from dataclasses import dataclass
 from pathlib import Path
+from typing import Optional
 
 
 # This class represents a simple database that stores its data as files in a directory.
@@ -47,3 +48,4 @@ class DBs:
     preprompts: DB
     input: DB
     workspace: DB
+    archive: DB

gpt_engineer/main.py

@@ -1,6 +1,6 @@
 import json
 import logging
-import shutil
+import os
 
 from pathlib import Path
 
@@ -18,7 +18,6 @@ app = typer.Typer()
 @app.command()
 def main(
     project_path: str = typer.Argument("example", help="path"),
-    delete_existing: bool = typer.Argument(False, help="delete existing files"),
     model: str = typer.Argument("gpt-4", help="model id string"),
     temperature: float = 0.1,
     steps_config: steps.Config = typer.Option(
@@ -35,28 +34,25 @@ def main(
 ):
     logging.basicConfig(level=logging.DEBUG if verbose else logging.INFO)
 
-    input_path = Path(project_path).absolute()
-    memory_path = input_path / f"{run_prefix}memory"
-    workspace_path = input_path / f"{run_prefix}workspace"
-
-    if delete_existing:
-        # Delete files and subdirectories in paths
-        shutil.rmtree(memory_path, ignore_errors=True)
-        shutil.rmtree(workspace_path, ignore_errors=True)
-
     model = fallback_model(model)
-
     ai = AI(
         model=model,
         temperature=temperature,
     )
 
+    input_path = Path(project_path).absolute()
+    memory_path = input_path / f"{run_prefix}memory"
+    workspace_path = input_path / f"{run_prefix}workspace"
+    archive_path = input_path / f"{run_prefix}archive"
+
+    initial_run = not os.path.exists(memory_path) and not os.path.exists(workspace_path)
     dbs = DBs(
         memory=DB(memory_path),
         logs=DB(memory_path / "logs"),
         input=DB(input_path),
         workspace=DB(workspace_path),
         preprompts=DB(Path(__file__).parent / "preprompts"),
+        archive=None if initial_run else DB(archive_path),
     )
 
     steps = STEPS[steps_config]

gpt_engineer/steps.py

@@ -1,5 +1,8 @@
+import datetime
 import json
+import os
 import re
+import shutil
 import subprocess
 
 from enum import Enum
@@ -258,6 +261,16 @@ def fix_code(ai: AI, dbs: DBs):
     return messages
 
 
+def archive(ai: AI, dbs: DBs):
+    os.makedirs(dbs.archive.path, exist_ok=True)
+    timestamp = datetime.datetime.now().strftime("%Y%m%d_%H%M%S")
+    shutil.move(dbs.memory.path, dbs.archive.path / timestamp / dbs.memory.path.name)
+    shutil.move(
+        dbs.workspace.path, dbs.archive.path / timestamp / dbs.workspace.path.name
+    )
+    return []
+
+
 def human_review(ai: AI, dbs: DBs):
     review = human_input()
     dbs.memory["review"] = review.to_json()  # type: ignore
@@ -280,15 +293,17 @@ class Config(str, Enum):
 # Different configs of what steps to run
 STEPS = {
     Config.DEFAULT: [
+        archive,
         clarify,
         gen_clarified_code,
         gen_entrypoint,
         execute_entrypoint,
         human_review,
     ],
-    Config.BENCHMARK: [simple_gen, gen_entrypoint],
-    Config.SIMPLE: [simple_gen, gen_entrypoint, execute_entrypoint],
+    Config.BENCHMARK: [archive, simple_gen, gen_entrypoint],
+    Config.SIMPLE: [archive, simple_gen, gen_entrypoint, execute_entrypoint],
     Config.TDD: [
+        archive,
         gen_spec,
         gen_unit_tests,
         gen_code,
@@ -297,6 +312,7 @@ STEPS = {
         human_review,
     ],
     Config.TDD_PLUS: [
+        archive,
         gen_spec,
         gen_unit_tests,
         gen_code,
@@ -306,6 +322,7 @@ STEPS = {
         human_review,
     ],
     Config.CLARIFY: [
+        archive,
         clarify,
         gen_clarified_code,
         gen_entrypoint,
@@ -313,6 +330,7 @@ STEPS = {
         human_review,
     ],
     Config.RESPEC: [
+        archive,
         gen_spec,
         respec,
         gen_unit_tests,

tests/steps/test_archive.py

@@ -0,0 +1,44 @@
+import datetime
+import os
+
+from unittest.mock import MagicMock
+
+from gpt_engineer.db import DB, DBs
+from gpt_engineer.steps import archive
+
+
+def freeze_at(monkeypatch, time):
+    datetime_mock = MagicMock(wraps=datetime.datetime)
+    datetime_mock.now.return_value = time
+    monkeypatch.setattr(datetime, "datetime", datetime_mock)
+
+
+def setup_dbs(tmp_path, dir_names):
+    directories = [tmp_path / name for name in dir_names]
+
+    # Create DB objects
+    dbs = [DB(dir) for dir in directories]
+
+    # Create DBs instance
+    return DBs(*dbs)
+
+
+def test_archive(tmp_path, monkeypatch):
+    dbs = setup_dbs(
+        tmp_path, ["memory", "logs", "preprompts", "input", "workspace", "archive"]
+    )
+    freeze_at(monkeypatch, datetime.datetime(2020, 12, 25, 17, 5, 55))
+    archive(None, dbs)
+    assert not os.path.exists(tmp_path / "memory")
+    assert not os.path.exists(tmp_path / "workspace")
+    assert os.path.isdir(tmp_path / "archive" / "20201225_170555")
+
+    dbs = setup_dbs(
+        tmp_path, ["memory", "logs", "preprompts", "input", "workspace", "archive"]
+    )
+    freeze_at(monkeypatch, datetime.datetime(2022, 8, 14, 8, 5, 12))
+    archive(None, dbs)
+    assert not os.path.exists(tmp_path / "memory")
+    assert not os.path.exists(tmp_path / "workspace")
+    assert os.path.isdir(tmp_path / "archive" / "20201225_170555")
+    assert os.path.isdir(tmp_path / "archive" / "20220814_080512")

tests/test_collect.py

@@ -19,7 +19,7 @@ def test_collect_learnings(monkeypatch):
     model = "test_model"
     temperature = 0.5
     steps = [gen_code]
-    dbs = DBs(DB("/tmp"), DB("/tmp"), DB("/tmp"), DB("/tmp"), DB("/tmp"))
+    dbs = DBs(DB("/tmp"), DB("/tmp"), DB("/tmp"), DB("/tmp"), DB("/tmp"), DB("/tmp"))
     dbs.input = {
         "prompt": "test prompt\n with newlines",
         "feedback": "test feedback",

tests/test_db.py

@@ -27,7 +27,7 @@ def test_DB_operations(tmp_path):
 
 
 def test_DBs_initialization(tmp_path):
-    dir_names = ["memory", "logs", "preprompts", "input", "workspace"]
+    dir_names = ["memory", "logs", "preprompts", "input", "workspace", "archive"]
     directories = [tmp_path / name for name in dir_names]
 
     # Create DB objects
@@ -41,6 +41,7 @@ def test_DBs_initialization(tmp_path):
     assert isinstance(dbs_instance.preprompts, DB)
     assert isinstance(dbs_instance.input, DB)
     assert isinstance(dbs_instance.workspace, DB)
+    assert isinstance(dbs_instance.archive, DB)
 
 
 def test_invalid_path():
@@ -106,11 +107,11 @@ def test_DBs_instantiation_with_wrong_number_of_arguments(tmp_path):
         DBs(db, db, db)
 
     with pytest.raises(TypeError):
-        DBs(db, db, db, db, db, db)
+        DBs(db, db, db, db, db, db, db)
 
 
 def test_DBs_dataclass_attributes(tmp_path):
-    dir_names = ["memory", "logs", "preprompts", "input", "workspace"]
+    dir_names = ["memory", "logs", "preprompts", "input", "workspace", "archive"]
     directories = [tmp_path / name for name in dir_names]
 
     # Create DB objects