fix: tweaks to codegen docker runner (#621)

Author: caroljung-cg

Dockerfile-runner

@@ -10,6 +10,7 @@ ENV NVM_DIR=/root/.nvm \
     PYTHONPATH="/usr/local/lib/python3.13/site-packages" \
     IS_SANDBOX=True \
     HATCH_BUILD_HOOKS_ENABLE=1
+
 # Update packages lists and install git and curl
 RUN apt-get update && apt-get install -y \
     git \
@@ -41,17 +42,14 @@ RUN node --version \
     && pnpm --version \
     && python --version
 
-# Install codegen from source instead of PyPI
+# Build codegen
 WORKDIR /codegen-sdk
 COPY . .
-
-# Install dependencies and build codegen with entry points
 RUN --mount=type=cache,target=/root/.cache/uv \
     uv venv && source .venv/bin/activate \
     && uv sync --frozen --no-dev --all-extras \
     && uv pip install --system -e . --no-deps \
     && uv pip install --system .
-
 RUN codegen --version
 
 # Create a non-root user for local development + debugging

src/codegen/cli/commands/create/main.py

@@ -68,17 +68,17 @@ def make_relative(path: Path) -> str:
 @click.command(name="create")
 @requires_init
 @click.argument("name", type=str)
-@click.argument("path", type=click.Path(path_type=Path), default=Path.cwd())
+@click.argument("path", type=click.Path(path_type=Path), default=None)
 @click.option("--description", "-d", default=None, help="Description of what this codemod does.")
 @click.option("--overwrite", is_flag=True, help="Overwrites function if it already exists.")
-def create_command(session: CodegenSession, name: str, path: Path, description: str | None = None, overwrite: bool = False):
+def create_command(session: CodegenSession, name: str, path: Path | None, description: str | None = None, overwrite: bool = False):
     """Create a new codegen function.
 
     NAME is the name/label for the function
     PATH is where to create the function (default: current directory)
     """
     # Get the target path for the function
-    codemod_path, prompt_path = get_target_paths(name, path)
+    codemod_path, prompt_path = get_target_paths(name, path or Path.cwd())
 
     # Check if file exists
     if codemod_path.exists() and not overwrite:

src/codegen/cli/commands/start/main.py

@@ -8,24 +8,28 @@
 from rich.panel import Panel
 
 from codegen.configs.models.secrets import SecretsConfig
+from codegen.git.repo_operator.local_git_repo import LocalGitRepo
 from codegen.git.schemas.repo_config import RepoConfig
+from codegen.shared.network.port import get_free_port
 
 
 @click.command(name="start")
 @click.option("--platform", "-t", type=click.Choice(["linux/amd64", "linux/arm64", "linux/amd64,linux/arm64"]), default="linux/amd64,linux/arm64", help="Target platform(s) for the Docker image")
-@click.option("--port", "-p", type=int, default=8000)
+@click.option("--port", "-p", type=int, default=None, help="Port to run the server on")
 @click.option("--detached", "-d", is_flag=True, default=False, help="Starts up the server as detached background process")
-def start_command(port: int, platform: str, detached: bool):
+def start_command(port: int | None, platform: str, detached: bool):
     """Starts a local codegen server"""
     codegen_version = version("codegen")
     rich.print(f"[bold green]Codegen version:[/bold green] {codegen_version}")
     codegen_root = Path(__file__).parent.parent.parent.parent.parent.parent
+    if port is None:
+        port = get_free_port()
 
     try:
         rich.print("[bold blue]Building Docker image...[/bold blue]")
         _build_docker_image(codegen_root, platform)
         rich.print("[bold blue]Starting Docker container...[/bold blue]")
-        _run_docker_container(port, platform, detached)
+        _run_docker_container(port, detached)
         rich.print(Panel(f"[green]Server started successfully![/green]\nAccess the server at: [bold]http://0.0.0.0:{port}[/bold]", box=ROUNDED, title="Codegen Server"))
     except subprocess.CalledProcessError as e:
         rich.print(f"[bold red]Error:[/bold red] Failed to {e.cmd[0]} Docker container")
@@ -59,16 +63,15 @@ def _run_docker_container(port: int, detached: bool):
     container_repo_path = f"/app/git/{repo_config.name}"
     envvars = {
         "REPOSITORY_LANGUAGE": repo_config.language.value,
-        "REPOSITORY_OWNER": repo_config.organization_name,
+        "REPOSITORY_OWNER": LocalGitRepo(repo_path).owner,
         "REPOSITORY_PATH": container_repo_path,
         "GITHUB_TOKEN": SecretsConfig().github_token,
     }
     envvars_args = [arg for k, v in envvars.items() for arg in ("--env", f"{k}={v}")]
-
     mount_args = ["-v", f"{repo_path}:{container_repo_path}"]
     run_mode = "-d" if detached else "-it"
     entry_point = f"uv run --frozen uvicorn codegen.runner.sandbox.server:app --host 0.0.0.0 --port {port}"
-    run_cmd = ["docker", "run", run_mode, "-p", f"8000:{port}", *mount_args, *envvars_args, "codegen-runner", entry_point]
+    run_cmd = ["docker", "run", run_mode, "-p", f"{port}:{port}", *mount_args, *envvars_args, "codegen-runner", entry_point]
 
     rich.print(f"run_cmd: {str.join(' ', run_cmd)}")
     subprocess.run(run_cmd, check=True)

src/codegen/git/repo_operator/local_git_repo.py

@@ -47,6 +47,7 @@ def origin_remote(self) -> Remote | None:
         """Returns the url of the first remote found on the repo, or None if no remotes are set"""
         if self.has_remote():
             return self.git_cli.remote("origin")
+        return None
 
     @cached_property
     def base_url(self) -> str | None:

src/codegen/git/schemas/repo_config.py

@@ -38,4 +38,5 @@ def repo_path(self) -> Path:
     def organization_name(self) -> str | None:
         if self.full_name is not None:
             return self.full_name.split("/")[0]
+
         return None

src/codegen/runner/sandbox/runner.py

@@ -31,7 +31,7 @@ class SandboxRunner:
 
     def __init__(self, repo_config: RepoConfig) -> None:
         self.repo = repo_config
-        self.op = RepoOperator(repo_config=self.repo, setup_option=SetupOption.PULL_OR_CLONE)
+        self.op = RepoOperator(repo_config=self.repo, setup_option=SetupOption.PULL_OR_CLONE, bot_commit=True)
         self.commit = self.op.git_cli.head.commit
 
     async def warmup(self) -> None:

src/codegen/shared/network/port.py

@@ -0,0 +1,17 @@
+import socket
+from contextlib import closing
+
+
+def get_free_port() -> int:
+    """Find and return a free port on localhost"""
+    with closing(socket.socket(socket.AF_INET, socket.SOCK_STREAM)) as s:
+        s.bind(("", 0))
+        s.listen(1)
+        port = s.getsockname()[1]
+        return int(port)
+
+
+def is_port_free(port: int, host: str = "localhost") -> bool:
+    """Check if a port is free on localhost"""
+    with closing(socket.socket(socket.AF_INET, socket.SOCK_STREAM)) as s:
+        return s.connect_ex((host, port)) != 0

tests/integration/codegen/runner/conftest.py

@@ -1,6 +1,4 @@
-import socket
 from collections.abc import Generator
-from contextlib import closing
 from unittest.mock import Mock
 
 import pytest
@@ -11,16 +9,7 @@
 from codegen.git.schemas.repo_config import RepoConfig
 from codegen.runner.clients.codebase_client import CodebaseClient
 from codegen.shared.enums.programming_language import ProgrammingLanguage
-
-
-@pytest.fixture
-def get_free_port():
-    """Find and return a free port on localhost"""
-    with closing(socket.socket(socket.AF_INET, socket.SOCK_STREAM)) as s:
-        s.bind(("", 0))
-        s.listen(1)
-        port = s.getsockname()[1]
-        return port
+from codegen.shared.network.port import get_free_port
 
 
 @pytest.fixture()
@@ -44,7 +33,7 @@ def git_repo_client(op: RepoOperator, repo_config: RepoConfig) -> Generator[GitR
 
 
 @pytest.fixture
-def codebase_client(repo_config: RepoConfig, get_free_port) -> Generator[CodebaseClient, None, None]:
-    sb_client = CodebaseClient(repo_config=repo_config, port=get_free_port)
+def codebase_client(repo_config: RepoConfig) -> Generator[CodebaseClient, None, None]:
+    sb_client = CodebaseClient(repo_config=repo_config, port=get_free_port())
     sb_client.runner = Mock()
     yield sb_client