Switchyard lets LLM applications route traffic across models and providers while preserving native OpenAI and Anthropic API compatibility - enabling flexible model selection, benchmarking, and cost/performance optimization.
Rust
# SPDX-FileCopyrightText: Copyright (c) 2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
# SPDX-License-Identifier: Apache-2.0
"""CLI dispatcher for Harbor task converters.
Usage::
craft-taskgen-convert --adapter search-native \\
--candidates-dir tasks/candidates/search/ \\
--manifest repos/manifest.json \\
--output-dir harbor-tasks/craft-search/ \\
--limit 10
Register a new adapter by:
1. Creating `adapters/<name>/converter.py` with a `run_convert(...)` function.
2. Adding an entry to ``_ADAPTERS`` below that describes its CLI args and how
to call ``run_convert``.
"""
from __future__ import annotations
import argparse
from dataclasses import dataclass
from typing import Any, Callable
@dataclass(frozen=True)
class AdapterSpec:
"""Describes an adapter's CLI surface and entry point."""
name: str
help: str
add_args: Callable[[argparse.ArgumentParser], None]
run: Callable[[argparse.Namespace], Any]
def _search_native_add_args(parser: argparse.ArgumentParser) -> None:
parser.add_argument(
"--candidates-dir",
required=True,
help="Root dir with per-repo candidate subdirs ({repo}/{uuid}.json)",
)
parser.add_argument(
"--manifest",
required=True,
help="Path to repos/manifest.json (repo_name -> {url, commit})",
)
parser.add_argument(
"--output-dir",
required=True,
help="Output directory for Harbor task dirs",
)
parser.add_argument(
"--limit",
type=int,
default=0,
help="Stop after this many tasks (0 = all)",
)
def _search_native_run(args: argparse.Namespace) -> Any:
from craft_taskgen.adapters.search_native.converter import run_convert
return run_convert(
candidates_dir=args.candidates_dir,
manifest_path=args.manifest,
output_dir=args.output_dir,
limit=args.limit,
)
def _planning_add_args(parser: argparse.ArgumentParser) -> None:
parser.add_argument(
"--candidates-dir",
required=True,
help="Directory of bootstrapped planning candidate JSONs ({task_name}.json each)",
)
parser.add_argument(
"--output-dir",
required=True,
help="Output directory for Harbor planning task dirs",
)
parser.add_argument(
"--repo-cache",
default=None,
help="Local git clone cache (default: /tmp/craft-taskgen-repos)",
)
parser.add_argument(
"--limit",
type=int,
default=0,
help="Stop after this many tasks (0 = all)",
)
def _planning_run(args: argparse.Namespace) -> Any:
from craft_taskgen.adapters.planning.converter import run_convert
return run_convert(
candidates_dir=args.candidates_dir,
output_dir=args.output_dir,
repo_cache=args.repo_cache,
limit=args.limit,
)
_ADAPTERS: dict[str, AdapterSpec] = {
"search-native": AdapterSpec(
name="search-native",
help="Native Search tasks (builds fresh Dockerfile per task)",
add_args=_search_native_add_args,
run=_search_native_run,
),
"planning": AdapterSpec(
name="planning",
help="Iterative planning track tasks with binary F2P+P2P reward",
add_args=_planning_add_args,
run=_planning_run,
),
}
def _top_level_help(adapters: dict[str, AdapterSpec]) -> str:
lines = [
"usage: craft-taskgen-convert --adapter <name> [adapter args ...]",
"",
"Convert task candidates into Harbor task directories.",
"",
"Adapters:",
]
for name in sorted(adapters):
lines.append(f" {name:16s} {adapters[name].help}")
lines.append("")
lines.append("Use --adapter <name> --help to see adapter-specific arguments.")
return "\n".join(lines)
def main() -> None:
# Top-level parser: disable its own -h so --help can reach the adapter parser
# once an adapter is selected. If no --adapter is given, print our own help.
parser = argparse.ArgumentParser(add_help=False)
parser.add_argument(
"--adapter",
choices=sorted(_ADAPTERS.keys()),
help="Which adapter to run",
)
args, remaining = parser.parse_known_args()
if args.adapter is None:
print(_top_level_help(_ADAPTERS))
return
spec = _ADAPTERS[args.adapter]
sub_parser = argparse.ArgumentParser(
prog=f"craft-taskgen-convert --adapter {args.adapter}",
description=spec.help,
)
spec.add_args(sub_parser)
sub_args = sub_parser.parse_args(remaining)
spec.run(sub_args)
if __name__ == "__main__":
main()