Ensure llama stack build --config <> --image-type <> works (#879)

Fix the issues brought up in
#870

Test all combinations of (conda, container) vs. (template, config)
combos.

llama_stack/cli/stack/_build.py

@@ -115,6 +115,8 @@ def run_stack_build_command(
                     f"Using conda environment {image_name}",
                     color="green",
                 )
+        else:
+            image_name = f"llamastack-{name}"
 
         cprint(
             textwrap.dedent(
@@ -171,11 +173,22 @@ def run_stack_build_command(
                 )
                 return
 
-    _run_stack_build_command_from_build_config(build_config, image_name=image_name)
+        if build_config.image_type == ImageType.container.value and not args.image_name:
+            cprint(
+                "Please specify --image-name when building a container from a config file",
+                color="red",
+            )
+            return
+
+    _run_stack_build_command_from_build_config(
+        build_config, image_name=image_name, config_path=args.config
+    )
 
 
 def _generate_run_config(
-    build_config: BuildConfig, build_dir: Path, image_name: str
+    build_config: BuildConfig,
+    build_dir: Path,
+    image_name: str,
 ) -> None:
     """
     Generate a run.yaml template file for user to edit from a build.yaml file
@@ -227,8 +240,9 @@ def _generate_run_config(
         to_write = json.loads(run_config.model_dump_json())
         f.write(yaml.dump(to_write, sort_keys=False))
 
+    # this path is only invoked when no template is provided
     cprint(
-        f"You can now edit {run_config_file} and run `llama stack run {image_name}`",
+        f"You can now run your stack with `llama stack run {run_config_file}`",
         color="green",
     )
 
@@ -237,6 +251,7 @@ def _run_stack_build_command_from_build_config(
     build_config: BuildConfig,
     image_name: Optional[str] = None,
     template_name: Optional[str] = None,
+    config_path: Optional[str] = None,
 ) -> None:
     if build_config.image_type == ImageType.container.value:
         if template_name:
@@ -263,7 +278,10 @@ def _run_stack_build_command_from_build_config(
         f.write(yaml.dump(to_write, sort_keys=False))
 
     return_code = build_image(
-        build_config, build_file_path, image_name, template_name=template_name
+        build_config,
+        build_file_path,
+        image_name,
+        template_or_config=template_name or config_path,
     )
     if return_code != 0:
         return
@@ -277,7 +295,7 @@ def _run_stack_build_command_from_build_config(
         with importlib.resources.as_file(template_path) as path:
             run_config_file = build_dir / f"{template_name}-run.yaml"
             shutil.copy(path, run_config_file)
-        # Find all ${env.VARIABLE} patterns
+
         cprint("Build Successful!", color="green")
     else:
         _generate_run_config(build_config, build_dir, image_name)

llama_stack/cli/stack/run.py

@@ -78,12 +78,15 @@ def _run_stack_run_cmd(self, args: argparse.Namespace) -> None:
 
         config_file = Path(args.config)
         has_yaml_suffix = args.config.endswith(".yaml")
+        template_name = None
 
         if not config_file.exists() and not has_yaml_suffix:
             # check if this is a template
             config_file = (
                 Path(REPO_ROOT) / "llama_stack" / "templates" / args.config / "run.yaml"
             )
+            if config_file.exists():
+                template_name = args.config
 
         if not config_file.exists() and not has_yaml_suffix:
             # check if it's a build config saved to conda dir
@@ -120,7 +123,12 @@ def _run_stack_run_cmd(self, args: argparse.Namespace) -> None:
                 importlib.resources.files("llama_stack")
                 / "distribution/start_container.sh"
             )
-            run_args = [script, config.container_image]
+            image_name = (
+                f"distribution-{template_name}"
+                if template_name
+                else config.container_image
+            )
+            run_args = [script, image_name]
         else:
             current_conda_env = os.environ.get("CONDA_DEFAULT_ENV")
             image_name = args.image_name or current_conda_env

llama_stack/distribution/build.py

@@ -10,7 +10,7 @@
 from enum import Enum
 
 from pathlib import Path
-from typing import Dict, List, Optional
+from typing import Dict, List
 
 from pydantic import BaseModel
 from termcolor import cprint
@@ -107,9 +107,9 @@ def build_image(
     build_config: BuildConfig,
     build_file_path: Path,
     image_name: str,
-    template_name: Optional[str] = None,
+    template_or_config: str,
 ):
-    container_image = (
+    container_base = (
         build_config.distribution_spec.container_image or "python:3.10-slim"
     )
 
@@ -119,16 +119,14 @@ def build_image(
     normal_deps += SERVER_DEPENDENCIES
 
     if build_config.image_type == ImageType.container.value:
-        if not template_name:
-            raise ValueError("template_name is required for container builds")
-
         script = str(
             importlib.resources.files("llama_stack") / "distribution/build_container.sh"
         )
         args = [
             script,
-            template_name,
-            container_image,
+            template_or_config,
+            image_name,
+            container_base,
             str(build_file_path),
             str(BUILDS_BASE_DIR / ImageType.container.value),
             " ".join(normal_deps),

llama_stack/distribution/build_container.sh

@@ -12,22 +12,22 @@ TEST_PYPI_VERSION=${TEST_PYPI_VERSION:-}
 PYPI_VERSION=${PYPI_VERSION:-}
 BUILD_PLATFORM=${BUILD_PLATFORM:-}
 
-if [ "$#" -lt 5 ]; then
+if [ "$#" -lt 6 ]; then
   # This only works for templates
-  echo "Usage: $0 <template_name> <container_base> <pip_dependencies> <host_build_dir> [<special_pip_deps>]" >&2
-  echo "Example: $0 fireworks python:3.9-slim 'fastapi uvicorn' /path/to/build/dir" >&2
+  echo "Usage: $0 <template_or_config> <image_name> <container_base> <build_file_path> <host_build_dir> <pip_dependencies> [<special_pip_deps>]" >&2
   exit 1
 fi
 
-special_pip_deps="$6"
-
 set -euo pipefail
 
-template_name="$1"
-container_base=$2
-build_file_path=$3
-host_build_dir=$4
-pip_dependencies=$5
+template_or_config="$1"
+image_name="$2"
+container_base="$3"
+build_file_path="$4"
+host_build_dir="$5"
+pip_dependencies="$6"
+special_pip_deps="$7"
+
 
 # Define color codes
 RED='\033[0;31m'
@@ -147,14 +147,16 @@ RUN pip install --no-cache $models_mount
 EOF
 fi
 
-add_to_container << EOF
-
-# This would be good in production but for debugging flexibility lets not add it right now
-# We need a more solid production ready entrypoint.sh anyway
-#
-ENTRYPOINT ["python", "-m", "llama_stack.distribution.server.server", "--template", "$template_name"]
-
+# if template_or_config ends with .yaml, it is not a template and we should not use the --template flag
+if [[ "$template_or_config" != *.yaml ]]; then
+  add_to_container << EOF
+ENTRYPOINT ["python", "-m", "llama_stack.distribution.server.server", "--template", "$template_or_config"]
 EOF
+else
+  add_to_container << EOF
+ENTRYPOINT ["python", "-m", "llama_stack.distribution.server.server"]
+EOF
+fi
 
 printf "Containerfile created successfully in $TEMP_DIR/Containerfile\n\n"
 cat $TEMP_DIR/Containerfile
@@ -174,7 +176,9 @@ if command -v selinuxenabled &>/dev/null && selinuxenabled; then
 fi
 
 # Set version tag based on PyPI version
-if [ -n "$TEST_PYPI_VERSION" ]; then
+if [ -n "$PYPI_VERSION" ]; then
+  version_tag="$PYPI_VERSION"
+elif [ -n "$TEST_PYPI_VERSION" ]; then
   version_tag="test-$TEST_PYPI_VERSION"
 elif [[ -n "$LLAMA_STACK_DIR" || -n "$LLAMA_MODELS_DIR" ]]; then
   version_tag="dev"
@@ -184,8 +188,7 @@ else
 fi
 
 # Add version tag to image name
-build_name="distribution-$template_name"
-image_tag="$build_name:$version_tag"
+image_tag="$image_name:$version_tag"
 
 # Detect platform architecture
 ARCH=$(uname -m)

llama_stack/distribution/start_container.sh

@@ -30,8 +30,8 @@ if [ $# -lt 3 ]; then
   exit 1
 fi
 
-build_name="$1"
-container_image="localhost/distribution-$build_name"
+image_name="$1"
+container_image="localhost/$image_name"
 shift
 
 yaml_config="$1"
@@ -76,13 +76,15 @@ if [ -n "$LLAMA_CHECKPOINT_DIR" ]; then
   CONTAINER_OPTS="$CONTAINER_OPTS --gpus=all"
 fi
 
-version_tag="latest"
 if [ -n "$PYPI_VERSION" ]; then
   version_tag="$PYPI_VERSION"
 elif [ -n "$LLAMA_STACK_DIR" ]; then
   version_tag="dev"
 elif [ -n "$TEST_PYPI_VERSION" ]; then
   version_tag="test-$TEST_PYPI_VERSION"
+else
+  URL="https://pypi.org/pypi/llama-stack/json"
+  version_tag=$(curl -s $URL | jq -r '.info.version')
 fi
 
 $CONTAINER_BINARY run $CONTAINER_OPTS -it \