feat: bump llama-stack to 0.3.0rc2+rhai0

https://github.com/opendatahub-io/llama-stack/releases/tag/v0.3.0rc2%2Brhai0

Relates to: RHAIENG-1685

Signed-off-by: Mustafa Elbehery <melbeher@redhat.com>

Author: Elbehery

.pre-commit-config.yaml

@@ -56,7 +56,7 @@ repos:
         always_run: true
         files: ^distribution/.*$
         additional_dependencies:
-          - llama-stack==0.2.23
+          - git+https://github.com/opendatahub-io/llama-stack.git@v0.3.0rc2+rhai0
 
       - id: doc-gen
         name: Distribution Documentation

distribution/Containerfile

@@ -20,12 +20,12 @@ RUN pip install \
     autoevals \
     boto3 \
     chardet \
+    einops \
     faiss-cpu \
     fastapi \
     fire \
     google-cloud-aiplatform \
     httpx \
-    ibm_watsonx_ai \
     litellm \
     matplotlib \
     nltk \
@@ -39,10 +39,12 @@ RUN pip install \
     pypdf \
     redis \
     requests \
+    safetensors \
     scikit-learn \
     scipy \
     sentencepiece \
     sqlalchemy[asyncio] \
+    tokenizers \
     tqdm \
     transformers \
     uvicorn
@@ -56,7 +58,7 @@ RUN pip install \
     llama_stack_provider_trustyai_fms==0.2.3
 RUN pip install 'torchao>=0.12.0' --extra-index-url https://download.pytorch.org/whl/cpu torch torchvision
 RUN pip install --no-deps sentence-transformers
-RUN pip install --no-cache llama-stack==0.2.23
+RUN pip install --no-cache --no-deps git+https://github.com/opendatahub-io/llama-stack.git@v0.3.0rc2+rhai0
 RUN mkdir -p ${HOME}/.llama ${HOME}/.cache
 COPY distribution/run.yaml ${APP_ROOT}/run.yaml
 ENTRYPOINT ["llama", "stack", "run", "/opt/app-root/run.yaml"]

distribution/Containerfile.in

@@ -3,7 +3,6 @@ WORKDIR /opt/app-root
 
 RUN pip install sqlalchemy # somehow sqlalchemy[asyncio] is not sufficient
 {dependencies}
-RUN pip install --no-cache llama-stack==0.2.23
 {llama_stack_install_source}
 RUN mkdir -p ${{HOME}}/.llama ${{HOME}}/.cache
 COPY distribution/run.yaml ${{APP_ROOT}}/run.yaml

distribution/README.md

@@ -4,7 +4,7 @@
 
 This image contains the official Open Data Hub Llama Stack distribution, with all the packages and configuration needed to run a Llama Stack server in a containerized environment.
 
-The image is currently shipping with upstream Llama Stack version [0.2.23](https://github.com/llamastack/llama-stack/releases/tag/v0.2.23)
+The image is currently shipping with upstream Llama Stack version [0.3.0rc2+rhai0](https://github.com/opendatahub-io/llama-stack/releases/tag/v0.3.0rc2+rhai0)
 
 You can see an overview of the APIs and Providers the image ships with in the table below.
 

distribution/build.py

@@ -13,7 +13,7 @@
 import os
 from pathlib import Path
 
-CURRENT_LLAMA_STACK_VERSION = "0.2.23"
+CURRENT_LLAMA_STACK_VERSION = "0.3.0rc2+rhai0"
 LLAMA_STACK_VERSION = os.getenv("LLAMA_STACK_VERSION", CURRENT_LLAMA_STACK_VERSION)
 BASE_REQUIREMENTS = [
     f"llama-stack=={LLAMA_STACK_VERSION}",
@@ -36,19 +36,30 @@
     git checkout {llama_stack_version} && \\
     pip install --no-cache -e ."""
 
+rhai_install_command = """RUN pip install --no-cache --no-deps git+https://github.com/opendatahub-io/llama-stack.git@v{llama_stack_version}"""
+
 
 def get_llama_stack_install(llama_stack_version):
     # If the version is a commit SHA or a short commit SHA, we need to install from source
     if is_install_from_source(llama_stack_version):
-        print(f"Installing llama-stack from source: {llama_stack_version}")
-        return source_install_command.format(
-            llama_stack_version=llama_stack_version
-        ).rstrip()
+        # Check if this is an RHAI version
+        if "+rhai" in llama_stack_version:
+            print(
+                f"Installing llama-stack from opendatahub-io repo: v{llama_stack_version}"
+            )
+            return rhai_install_command.format(
+                llama_stack_version=llama_stack_version
+            ).rstrip()
+        else:
+            print(f"Installing llama-stack from source: {llama_stack_version}")
+            return source_install_command.format(
+                llama_stack_version=llama_stack_version
+            ).rstrip()
 
 
 def is_install_from_source(llama_stack_version):
-    """Check if version string is a git commit SHA (no dots = SHA, has dots = version)."""
-    return "." not in llama_stack_version
+    """Check if version string is a git commit SHA (no dots = SHA, has dots = version) or contains +rhai suffix."""
+    return "." not in llama_stack_version or "+rhai" in llama_stack_version
 
 
 def check_llama_installed():

scripts/gen_distro_docs.py

@@ -21,15 +21,22 @@ def extract_llama_stack_version():
             content = file.read()
 
         # Look for llama-stack version in pip install commands
-        # Pattern matches: llama-stack==X.Y.Z
-        pattern = r"llama-stack==([0-9]+\.[0-9]+\.[0-9]+)"
+        # Pattern matches: llama-stack==X.Y.Z or llama-stack==X.Y.ZrcN+rhaiM
+        pattern = r"llama-stack==([0-9]+\.[0-9]+\.[0-9]+(?:rc[0-9]+)?(?:\+rhai[0-9]+)?)"
         match = re.search(pattern, content)
 
         if match:
             return match.group(1)
-        else:
-            print("Error: Could not find llama-stack version in Containerfile")
-            exit(1)
+
+        # Look for git URL format: git+https://github.com/*/llama-stack.git@vVERSION or @VERSION
+        git_pattern = r"git\+https://github\.com/[^/]+/llama-stack\.git@v?([0-9]+\.[0-9]+\.[0-9]+(?:rc[0-9]+)?(?:\+rhai[0-9]+)?)"
+        git_match = re.search(git_pattern, content)
+
+        if git_match:
+            return git_match.group(1)
+
+        print("Error: Could not find llama-stack version in Containerfile")
+        exit(1)
 
     except Exception as e:
         print(f"Error reading Containerfile: {e}")
@@ -163,14 +170,21 @@ def gen_distro_docs():
     # extract Llama Stack version from Containerfile
     version = extract_llama_stack_version()
 
+    # Determine repository URL based on version
+    # If version contains +rhai, use opendatahub-io, otherwise use llamastack
+    if "+rhai" in version:
+        repo_url = "https://github.com/opendatahub-io/llama-stack"
+    else:
+        repo_url = "https://github.com/llamastack/llama-stack"
+
     # header section
     header = f"""<!-- This file is automatically generated by scripts/gen_distro_doc.py - do not update manually -->
 
 # Open Data Hub Llama Stack Distribution Image
 
 This image contains the official Open Data Hub Llama Stack distribution, with all the packages and configuration needed to run a Llama Stack server in a containerized environment.
 
-The image is currently shipping with upstream Llama Stack version [{version}](https://github.com/llamastack/llama-stack/releases/tag/v{version})
+The image is currently shipping with upstream Llama Stack version [{version}]({repo_url}/releases/tag/v{version})
 
 You can see an overview of the APIs and Providers the image ships with in the table below.
 

tests/run_integration_tests.sh

@@ -3,20 +3,26 @@
 set -euo pipefail
 
 # Configuration
-LLAMA_STACK_REPO="https://github.com/meta-llama/llama-stack.git"
 WORK_DIR="/tmp/llama-stack-integration-tests"
 INFERENCE_MODEL="${INFERENCE_MODEL:-Qwen/Qwen3-0.6B}"
 
 SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
 
-# Get version dynamically from Containerfile.in (look in parent directory)
-CONTAINERFILE_IN="$SCRIPT_DIR/../distribution/Containerfile.in"
-LLAMA_STACK_VERSION=$(grep -o 'llama-stack==[0-9]\+\.[0-9]\+\.[0-9]\+' "$CONTAINERFILE_IN" | cut -d'=' -f3)
+# Get version dynamically from build.py (look in parent directory)
+BUILD_PY="$SCRIPT_DIR/../distribution/build.py"
+LLAMA_STACK_VERSION=$(grep '^CURRENT_LLAMA_STACK_VERSION' "$BUILD_PY" | head -1 | sed 's/.*= *"\(.*\)".*/\1/')
 if [ -z "$LLAMA_STACK_VERSION" ]; then
-    echo "Error: Could not extract llama-stack version from Containerfile.in"
+    echo "Error: Could not extract llama-stack version from build.py"
     exit 1
 fi
 
+# Determine which repo to use based on version suffix
+if [[ "$LLAMA_STACK_VERSION" == *"+rhai"* ]]; then
+    LLAMA_STACK_REPO="https://github.com/opendatahub-io/llama-stack.git"
+else
+    LLAMA_STACK_REPO="https://github.com/meta-llama/llama-stack.git"
+fi
+
 function clone_llama_stack() {
     # Clone the repository if it doesn't exist
     if [ ! -d "$WORK_DIR" ]; then