Simplify archiving process (AntonOsika#469)

* simplify args

* Fix tests

* Black format

Author: AntonOsika

gpt_engineer/ai.py

@@ -43,7 +43,7 @@ def next(self, messages: list[dict[str, str]], prompt=None):
 
         chat = []
         for chunk in response:
-            delta = chunk["choices"][0]["delta"]
+            delta = chunk["choices"][0]["delta"]  # type: ignore
             msg = delta.get("content", "")
             print(msg, end="")
             chat.append(msg)

gpt_engineer/db.py

@@ -1,6 +1,8 @@
+import datetime
+import shutil
+
 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.
@@ -49,3 +51,15 @@ class DBs:
     input: DB
     workspace: DB
     archive: DB
+
+
+def archive(dbs: DBs):
+    timestamp = datetime.datetime.now().strftime("%Y%m%d_%H%M%S")
+    shutil.move(
+        str(dbs.memory.path), str(dbs.archive.path / timestamp / dbs.memory.path.name)
+    )
+    shutil.move(
+        str(dbs.workspace.path),
+        str(dbs.archive.path / timestamp / dbs.workspace.path.name),
+    )
+    return []

gpt_engineer/main.py

@@ -1,16 +1,14 @@
 import json
 import logging
-import os
 
 from pathlib import Path
 
 import typer
 
-from gpt_engineer import steps
 from gpt_engineer.ai import AI, fallback_model
 from gpt_engineer.collect import collect_learnings
-from gpt_engineer.db import DB, DBs
-from gpt_engineer.steps import STEPS
+from gpt_engineer.db import DB, DBs, archive
+from gpt_engineer.steps import STEPS, Config as StepsConfig
 
 app = typer.Typer()
 
@@ -20,17 +18,10 @@ def main(
     project_path: str = typer.Argument("example", help="path"),
     model: str = typer.Argument("gpt-4", help="model id string"),
     temperature: float = 0.1,
-    steps_config: steps.Config = typer.Option(
-        steps.Config.DEFAULT, "--steps", "-s", help="decide which steps to run"
+    steps_config: StepsConfig = typer.Option(
+        StepsConfig.DEFAULT, "--steps", "-s", help="decide which steps to run"
     ),
     verbose: bool = typer.Option(False, "--verbose", "-v"),
-    run_prefix: str = typer.Option(
-        "",
-        help=(
-            "run prefix, if you want to run multiple variants of the same project and "
-            "later compare them"
-        ),
-    ),
 ):
     logging.basicConfig(level=logging.DEBUG if verbose else logging.INFO)
 
@@ -41,20 +32,26 @@ def main(
     )
 
     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"
+    memory_path = input_path / "memory"
+    workspace_path = input_path / "workspace"
+    archive_path = input_path / "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),
+        archive=DB(archive_path),
     )
 
+    if steps_config not in [
+        StepsConfig.EXECUTE_ONLY,
+        StepsConfig.USE_FEEDBACK,
+        StepsConfig.EVALUATE,
+    ]:
+        archive(dbs)
+
     steps = STEPS[steps_config]
     for step in steps:
         messages = step(ai, dbs)

gpt_engineer/steps.py

@@ -1,8 +1,5 @@
-import datetime
 import json
-import os
 import re
-import shutil
 import subprocess
 
 from enum import Enum
@@ -261,16 +258,6 @@ 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
@@ -293,17 +280,15 @@ 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: [archive, simple_gen, gen_entrypoint],
-    Config.SIMPLE: [archive, simple_gen, gen_entrypoint, execute_entrypoint],
+    Config.BENCHMARK: [simple_gen, gen_entrypoint],
+    Config.SIMPLE: [simple_gen, gen_entrypoint, execute_entrypoint],
     Config.TDD: [
-        archive,
         gen_spec,
         gen_unit_tests,
         gen_code,
@@ -312,7 +297,6 @@ class Config(str, Enum):
         human_review,
     ],
     Config.TDD_PLUS: [
-        archive,
         gen_spec,
         gen_unit_tests,
         gen_code,
@@ -322,15 +306,13 @@ class Config(str, Enum):
         human_review,
     ],
     Config.CLARIFY: [
-        archive,
         clarify,
         gen_clarified_code,
         gen_entrypoint,
         execute_entrypoint,
         human_review,
     ],
     Config.RESPEC: [
-        archive,
         gen_spec,
         respec,
         gen_unit_tests,

scripts/benchmark.py

@@ -125,7 +125,10 @@ def insert_markdown_section(file_path, section_title, section_text, level):
     if line_number != -1:
         lines.insert(line_number, new_section)
     else:
-        print(f"Markdown file was of unexpected format. No section of level {level} found. Did not write results.")
+        print(
+            f"Markdown file was of unexpected format. No section of level {level} found. "
+            "Did not write results."
+        )
         return
 
     # Write the file

tests/steps/test_archive.py

@@ -3,8 +3,7 @@
 
 from unittest.mock import MagicMock
 
-from gpt_engineer.db import DB, DBs
-from gpt_engineer.steps import archive
+from gpt_engineer.db import DB, DBs, archive
 
 
 def freeze_at(monkeypatch, time):
@@ -28,7 +27,7 @@ def test_archive(tmp_path, monkeypatch):
         tmp_path, ["memory", "logs", "preprompts", "input", "workspace", "archive"]
     )
     freeze_at(monkeypatch, datetime.datetime(2020, 12, 25, 17, 5, 55))
-    archive(None, dbs)
+    archive(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")
@@ -37,7 +36,7 @@ def test_archive(tmp_path, monkeypatch):
         tmp_path, ["memory", "logs", "preprompts", "input", "workspace", "archive"]
     )
     freeze_at(monkeypatch, datetime.datetime(2022, 8, 14, 8, 5, 12))
-    archive(None, dbs)
+    archive(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")