model config wip

libs/inference/pyproject.toml

@@ -8,16 +8,12 @@ authors = [
 ]
 requires-python = ">=3.12"
 dependencies = [
+    "accelerate>=1.2.0",
     "huggingface-hub>=0.27.0",
     "llama-cpp-python>=0.3.0",
     "pyyaml>=6.0.2",
-]
-
-[project.optional-dependencies]
-transformers = [
-    "accelerate>=1.2.0",
     "torch>=2.5.0",
-    "transformers>=4.47.0",
+    "transformers>=5.7.0",
 ]
 
 [build-system]

libs/inference/src/inference/transformers.py

@@ -21,8 +21,8 @@ class TransformersBackend:
             from transformers import AutoModelForCausalLM, AutoTokenizer
         except ImportError as exc:
             raise ImportError(
-                "transformers backend requires optional deps. "
-                "Install with: uv sync --package inference --extra transformers"
+                "transformers backend requires torch and transformers. "
+                "Install with: uv sync --all-packages"
             ) from exc
 
         device = "cuda" if torch.cuda.is_available() else "cpu"

scripts/download_model.py

@@ -1,16 +1,76 @@
 #!/usr/bin/env python3
-"""Download a configured GGUF preset from Hugging Face Hub for offline dev."""
+"""Download a configured model preset from Hugging Face Hub for offline dev."""
 
 from __future__ import annotations
 
 import argparse
 from pathlib import Path
 
-from huggingface_hub import hf_hub_download
+from huggingface_hub import hf_hub_download, snapshot_download
 
 from inference.config import get_app_config, get_model_config
 
 
+def _is_local_path(value: str) -> bool:
+    return value.startswith(("./", "../")) or Path(value).is_absolute()
+
+
+def _download_llama_cpp(model, preset_key: str, output_dir: Path) -> Path:
+    if model.model_path:
+        path = Path(model.model_path)
+        if not path.exists():
+            raise SystemExit(f"Local MODEL_PATH does not exist: {model.model_path}")
+        print(f"Preset {preset_key!r} already points to local file: {path}")
+        print(f"Set MODEL_PATH={path} or update models.yaml model_path to use it directly.")
+        return path.resolve()
+
+    if not model.model_repo or not model.model_file:
+        raise SystemExit(f"Preset {preset_key!r} is missing model_repo/model_file.")
+
+    output_dir.mkdir(parents=True, exist_ok=True)
+    path = Path(
+        hf_hub_download(
+            repo_id=model.model_repo,
+            filename=model.model_file,
+            local_dir=output_dir,
+            local_dir_use_symlinks=False,
+        )
+    )
+    print(f"Preset: {preset_key} ({model.label})")
+    print(f"Model ready at: {path}")
+    print("Add to models.yaml under that preset:")
+    print(f"  model_path: {path.resolve()}")
+    return path.resolve()
+
+
+def _download_transformers(model, preset_key: str, output_dir: Path) -> Path:
+    if not model.model_id:
+        raise SystemExit(f"Preset {preset_key!r} is missing model_id.")
+
+    if _is_local_path(model.model_id):
+        path = Path(model.model_id).resolve()
+        if not path.exists():
+            raise SystemExit(f"Local model_id does not exist: {model.model_id}")
+        print(f"Preset {preset_key!r} already points to local path: {path}")
+        print(f"Set MODEL_ID={path} or update models.yaml model_id to use it directly.")
+        return path
+
+    output_dir.mkdir(parents=True, exist_ok=True)
+    dest = output_dir / model.model_id
+    path = Path(
+        snapshot_download(
+            repo_id=model.model_id,
+            local_dir=dest,
+            local_dir_use_symlinks=False,
+        )
+    )
+    print(f"Preset: {preset_key} ({model.label})")
+    print(f"Model ready at: {path}")
+    print("Add to models.yaml under that preset:")
+    print(f"  model_id: ./{path.relative_to(Path.cwd())}")
+    return path.resolve()
+
+
 def main() -> None:
     parser = argparse.ArgumentParser(description=__doc__)
     parser.add_argument(
@@ -22,7 +82,7 @@ def main() -> None:
         "--output-dir",
         type=Path,
         default=Path("models"),
-        help="Directory to copy/symlink the downloaded model into",
+        help="Directory to download the model into",
     )
     args = parser.parse_args()
 
@@ -30,35 +90,15 @@ def main() -> None:
     preset_key = args.preset or app_config.active_model
     model = get_model_config(preset_key)
 
-    if model.backend != "llama_cpp":
-        raise SystemExit(
-            f"Preset {preset_key!r} uses backend {model.backend!r}. "
-            "Only llama_cpp presets with model_repo/model_file can be downloaded."
-        )
-
-    if model.model_path:
-        path = Path(model.model_path)
-        if not path.exists():
-            raise SystemExit(f"Local MODEL_PATH does not exist: {model.model_path}")
-        print(f"Preset {preset_key!r} already points to local file: {path}")
-        print(f"Set MODEL_PATH={path} or update models.yaml model_path to use it directly.")
+    if model.backend == "llama_cpp":
+        _download_llama_cpp(model, preset_key, args.output_dir)
         return
 
-    if not model.model_repo or not model.model_file:
-        raise SystemExit(f"Preset {preset_key!r} is missing model_repo/model_file.")
-
-    args.output_dir.mkdir(parents=True, exist_ok=True)
+    if model.backend == "transformers":
+        _download_transformers(model, preset_key, args.output_dir)
+        return
 
-    path = hf_hub_download(
-        repo_id=model.model_repo,
-        filename=model.model_file,
-        local_dir=args.output_dir,
-        local_dir_use_symlinks=False,
-    )
-    print(f"Preset: {preset_key} ({model.label})")
-    print(f"Model ready at: {path}")
-    print("Add to models.yaml under that preset:")
-    print(f"  model_path: {Path(path).resolve()}")
+    raise SystemExit(f"Preset {preset_key!r} uses unsupported backend {model.backend!r}.")
 
 
 if __name__ == "__main__":

uv.lock

@@ -498,28 +498,23 @@ name = "inference"
 version = "0.1.0"
 source = { editable = "libs/inference" }
 dependencies = [
+    { name = "accelerate" },
     { name = "huggingface-hub" },
     { name = "llama-cpp-python" },
     { name = "pyyaml" },
-]
-
-[package.optional-dependencies]
-transformers = [
-    { name = "accelerate" },
     { name = "torch" },
     { name = "transformers" },
 ]
 
 [package.metadata]
 requires-dist = [
-    { name = "accelerate", marker = "extra == 'transformers'", specifier = ">=1.2.0" },
+    { name = "accelerate", specifier = ">=1.2.0" },
     { name = "huggingface-hub", specifier = ">=0.27.0" },
     { name = "llama-cpp-python", specifier = ">=0.3.0" },
     { name = "pyyaml", specifier = ">=6.0.2" },
-    { name = "torch", marker = "extra == 'transformers'", specifier = ">=2.5.0" },
-    { name = "transformers", marker = "extra == 'transformers'", specifier = ">=4.47.0" },
+    { name = "torch", specifier = ">=2.5.0" },
+    { name = "transformers", specifier = ">=5.7.0" },
 ]
-provides-extras = ["transformers"]
 
 [[package]]
 name = "iniconfig"