diff --git a/.github/actions/run-and-record-tests/action.yml b/.github/actions/run-and-record-tests/action.yml
index 7f028b104..a3eb31d9f 100644
--- a/.github/actions/run-and-record-tests/action.yml
+++ b/.github/actions/run-and-record-tests/action.yml
@@ -5,21 +5,22 @@ inputs:
   stack-config:
     description: 'Stack configuration to use'
     required: true
-  provider:
-    description: 'Provider to use for tests'
-    required: true
+  setup:
+    description: 'Setup to use for tests (e.g., ollama, gpt, vllm)'
+    required: false
+    default: ''
   inference-mode:
     description: 'Inference mode (record or replay)'
     required: true
-  test-suite:
+  suite:
     description: 'Test suite to use: base, responses, vision, etc.'
     required: false
     default: ''
-  test-subdirs:
-    description: 'Comma-separated list of test subdirectories to run; overrides test-suite'
+  subdirs:
+    description: 'Comma-separated list of test subdirectories to run; overrides suite'
     required: false
     default: ''
-  test-pattern:
+  pattern:
     description: 'Regex pattern to pass to pytest -k'
     required: false
     default: ''
@@ -37,14 +38,23 @@ runs:
     - name: Run Integration Tests
       shell: bash
       run: |
-        uv run --no-sync ./scripts/integration-tests.sh \
-          --stack-config '${{ inputs.stack-config }}' \
-          --provider '${{ inputs.provider }}' \
-          --test-subdirs '${{ inputs.test-subdirs }}' \
-          --test-pattern '${{ inputs.test-pattern }}' \
-          --inference-mode '${{ inputs.inference-mode }}' \
-          --test-suite '${{ inputs.test-suite }}' \
-          | tee pytest-${{ inputs.inference-mode }}.log
+        SCRIPT_ARGS="--stack-config ${{ inputs.stack-config }} --inference-mode ${{ inputs.inference-mode }}"
+
+        # Add optional arguments only if they are provided
+        if [ -n '${{ inputs.setup }}' ]; then
+          SCRIPT_ARGS="$SCRIPT_ARGS --setup ${{ inputs.setup }}"
+        fi
+        if [ -n '${{ inputs.suite }}' ]; then
+          SCRIPT_ARGS="$SCRIPT_ARGS --suite ${{ inputs.suite }}"
+        fi
+        if [ -n '${{ inputs.subdirs }}' ]; then
+          SCRIPT_ARGS="$SCRIPT_ARGS --subdirs ${{ inputs.subdirs }}"
+        fi
+        if [ -n '${{ inputs.pattern }}' ]; then
+          SCRIPT_ARGS="$SCRIPT_ARGS --pattern ${{ inputs.pattern }}"
+        fi
+
+        uv run --no-sync ./scripts/integration-tests.sh $SCRIPT_ARGS | tee pytest-${{ inputs.inference-mode }}.log
 
 
     - name: Commit and push recordings
@@ -58,7 +68,7 @@ runs:
           echo "New recordings detected, committing and pushing"
           git add tests/integration/recordings/
 
-          git commit -m "Recordings update from CI (test-suite: ${{ inputs.test-suite }})"
+          git commit -m "Recordings update from CI (suite: ${{ inputs.suite }})"
           git fetch origin ${{ github.ref_name }}
           git rebase origin/${{ github.ref_name }}
           echo "Rebased successfully"
diff --git a/.github/actions/setup-ollama/action.yml b/.github/actions/setup-ollama/action.yml
index dc2f87e8c..5c95d131d 100644
--- a/.github/actions/setup-ollama/action.yml
+++ b/.github/actions/setup-ollama/action.yml
@@ -1,7 +1,7 @@
 name: Setup Ollama
 description: Start Ollama
 inputs:
-  test-suite:
+  suite:
     description: 'Test suite to use: base, responses, vision, etc.'
     required: false
     default: ''
@@ -11,7 +11,7 @@ runs:
     - name: Start Ollama
       shell: bash
       run: |
-        if [ "${{ inputs.test-suite }}" == "vision" ]; then
+        if [ "${{ inputs.suite }}" == "vision" ]; then
           image="ollama-with-vision-model"
         else
           image="ollama-with-models"
diff --git a/.github/actions/setup-test-environment/action.yml b/.github/actions/setup-test-environment/action.yml
index 3be76f009..478e8f598 100644
--- a/.github/actions/setup-test-environment/action.yml
+++ b/.github/actions/setup-test-environment/action.yml
@@ -8,11 +8,11 @@ inputs:
   client-version:
     description: 'Client version (latest or published)'
     required: true
-  provider:
-    description: 'Provider to setup (ollama or vllm)'
-    required: true
+  setup:
+    description: 'Setup to configure (ollama, vllm, gpt, etc.)'
+    required: false
     default: 'ollama'
-  test-suite:
+  suite:
     description: 'Test suite to use: base, responses, vision, etc.'
     required: false
     default: ''
@@ -30,13 +30,13 @@ runs:
         client-version: ${{ inputs.client-version }}
 
     - name: Setup ollama
-      if: ${{ inputs.provider == 'ollama' && inputs.inference-mode == 'record' }}
+      if: ${{ (inputs.setup == 'ollama' || inputs.setup == 'ollama-vision') && inputs.inference-mode == 'record' }}
       uses: ./.github/actions/setup-ollama
       with:
-        test-suite: ${{ inputs.test-suite }}
+        suite: ${{ inputs.suite }}
 
     - name: Setup vllm
-      if: ${{ inputs.provider == 'vllm' && inputs.inference-mode == 'record' }}
+      if: ${{ inputs.setup == 'vllm' && inputs.inference-mode == 'record' }}
       uses: ./.github/actions/setup-vllm
 
     - name: Build Llama Stack
diff --git a/.github/workflows/integration-tests.yml b/.github/workflows/integration-tests.yml
index bb53eea2f..711eccd9e 100644
--- a/.github/workflows/integration-tests.yml
+++ b/.github/workflows/integration-tests.yml
@@ -28,8 +28,8 @@ on:
         description: 'Test against both the latest and published versions'
         type: boolean
         default: false
-      test-provider:
-        description: 'Test against a specific provider'
+      test-setup:
+        description: 'Test against a specific setup'
         type: string
         default: 'ollama'
 
@@ -42,18 +42,18 @@ jobs:
 
   run-replay-mode-tests:
     runs-on: ubuntu-latest
-    name: ${{ format('Integration Tests ({0}, {1}, {2}, client={3}, {4})', matrix.client-type, matrix.provider, matrix.python-version, matrix.client-version, matrix.test-suite) }}
+    name: ${{ format('Integration Tests ({0}, {1}, {2}, client={3}, {4})', matrix.client-type, matrix.setup, matrix.python-version, matrix.client-version, matrix.suite) }}
 
     strategy:
       fail-fast: false
       matrix:
         client-type: [library, server]
-        # Use vllm on weekly schedule, otherwise use test-provider input (defaults to ollama)
-        provider: ${{ (github.event.schedule == '1 0 * * 0') && fromJSON('["vllm"]') || fromJSON(format('["{0}"]', github.event.inputs.test-provider || 'ollama')) }}
+        # Use vllm on weekly schedule, otherwise use test-setup input (defaults to ollama)
+        setup: ${{ (github.event.schedule == '1 0 * * 0') && fromJSON('["vllm"]') || fromJSON(format('["{0}"]', github.event.inputs.test-setup || 'ollama')) }}
         # Use Python 3.13 only on nightly schedule (daily latest client test), otherwise use 3.12
         python-version: ${{ github.event.schedule == '0 0 * * *' && fromJSON('["3.12", "3.13"]') || fromJSON('["3.12"]') }}
         client-version: ${{ (github.event.schedule == '0 0 * * *' || github.event.inputs.test-all-client-versions == 'true') && fromJSON('["published", "latest"]') || fromJSON('["latest"]') }}
-        test-suite: [base, vision]
+        suite: [base, vision]
 
     steps:
       - name: Checkout repository
@@ -64,14 +64,14 @@ jobs:
         with:
           python-version: ${{ matrix.python-version }}
           client-version: ${{ matrix.client-version }}
-          provider: ${{ matrix.provider }}
-          test-suite: ${{ matrix.test-suite }}
+          setup: ${{ matrix.setup }}
+          suite: ${{ matrix.suite }}
           inference-mode: 'replay'
 
       - name: Run tests
         uses: ./.github/actions/run-and-record-tests
         with:
           stack-config: ${{ matrix.client-type == 'library' && 'ci-tests' || 'server:ci-tests' }}
-          provider: ${{ matrix.provider }}
+          setup: ${{ matrix.setup }}
           inference-mode: 'replay'
-          test-suite: ${{ matrix.test-suite }}
+          suite: ${{ matrix.suite }}
diff --git a/.github/workflows/record-integration-tests.yml b/.github/workflows/record-integration-tests.yml
index 01797a54b..65a04f125 100644
--- a/.github/workflows/record-integration-tests.yml
+++ b/.github/workflows/record-integration-tests.yml
@@ -10,19 +10,19 @@ run-name: Run the integration test suite from tests/integration
 on:
   workflow_dispatch:
     inputs:
-      test-provider:
-        description: 'Test against a specific provider'
+      test-setup:
+        description: 'Test against a specific setup'
         type: string
         default: 'ollama'
-      test-suite:
+      suite:
         description: 'Test suite to use: base, responses, vision, etc.'
         type: string
         default: ''
-      test-subdirs:
-        description: 'Comma-separated list of test subdirectories to run; overrides test-suite'
+      subdirs:
+        description: 'Comma-separated list of test subdirectories to run; overrides suite'
         type: string
         default: ''
-      test-pattern:
+      pattern:
         description: 'Regex pattern to pass to pytest -k'
         type: string
         default: ''
@@ -39,10 +39,10 @@ jobs:
         run: |
           echo "::group::Workflow Inputs"
           echo "branch: ${{ github.ref_name }}"
-          echo "test-provider: ${{ inputs.test-provider }}"
-          echo "test-suite: ${{ inputs.test-suite }}"
-          echo "test-subdirs: ${{ inputs.test-subdirs }}"
-          echo "test-pattern: ${{ inputs.test-pattern }}"
+          echo "test-setup: ${{ inputs.test-setup }}"
+          echo "suite: ${{ inputs.suite }}"
+          echo "subdirs: ${{ inputs.subdirs }}"
+          echo "pattern: ${{ inputs.pattern }}"
           echo "::endgroup::"
 
       - name: Checkout repository
@@ -55,16 +55,16 @@ jobs:
         with:
           python-version: "3.12"  # Use single Python version for recording
           client-version: "latest"
-          provider: ${{ inputs.test-provider || 'ollama' }}
-          test-suite: ${{ inputs.test-suite }}
+          setup: ${{ inputs.test-setup || 'ollama' }}
+          suite: ${{ inputs.suite }}
           inference-mode: 'record'
 
       - name: Run and record tests
         uses: ./.github/actions/run-and-record-tests
         with:
           stack-config: 'server:ci-tests'  # recording must be done with server since more tests are run
-          provider: ${{ inputs.test-provider || 'ollama' }}
+          setup: ${{ inputs.test-setup || 'ollama' }}
           inference-mode: 'record'
-          test-suite: ${{ inputs.test-suite }}
-          test-subdirs: ${{ inputs.test-subdirs }}
-          test-pattern: ${{ inputs.test-pattern }}
+          suite: ${{ inputs.suite }}
+          subdirs: ${{ inputs.subdirs }}
+          pattern: ${{ inputs.pattern }}
diff --git a/llama_stack/cli/stack/_build.py b/llama_stack/cli/stack/_build.py
index c6e204773..b14e6fe55 100644
--- a/llama_stack/cli/stack/_build.py
+++ b/llama_stack/cli/stack/_build.py
@@ -45,6 +45,7 @@ from llama_stack.core.utils.dynamic import instantiate_class_type
 from llama_stack.core.utils.exec import formulate_run_args, run_command
 from llama_stack.core.utils.image_types import LlamaStackImageType
 from llama_stack.providers.datatypes import Api
+from llama_stack.providers.utils.sqlstore.sqlstore import SqliteSqlStoreConfig
 
 DISTRIBS_PATH = Path(__file__).parent.parent.parent / "distributions"
 
@@ -294,6 +295,12 @@ def _generate_run_config(
         if build_config.external_providers_dir
         else EXTERNAL_PROVIDERS_DIR,
     )
+    if not run_config.inference_store:
+        run_config.inference_store = SqliteSqlStoreConfig(
+            **SqliteSqlStoreConfig.sample_run_config(
+                __distro_dir__=(DISTRIBS_BASE_DIR / image_name).as_posix(), db_name="inference_store.db"
+            )
+        )
     # build providers dict
     provider_registry = get_provider_registry(build_config)
     for api in apis:
diff --git a/llama_stack/core/library_client.py b/llama_stack/core/library_client.py
index 9e7a8006c..ea5a2ac8e 100644
--- a/llama_stack/core/library_client.py
+++ b/llama_stack/core/library_client.py
@@ -10,7 +10,6 @@ import json
 import logging  # allow-direct-logging
 import os
 import sys
-from concurrent.futures import ThreadPoolExecutor
 from enum import Enum
 from io import BytesIO
 from pathlib import Path
@@ -148,7 +147,6 @@ class LlamaStackAsLibraryClient(LlamaStackClient):
         self.async_client = AsyncLlamaStackAsLibraryClient(
             config_path_or_distro_name, custom_provider_registry, provider_data, skip_logger_removal
         )
-        self.pool_executor = ThreadPoolExecutor(max_workers=4)
         self.provider_data = provider_data
 
         self.loop = asyncio.new_event_loop()
diff --git a/llama_stack/providers/registry/datasetio.py b/llama_stack/providers/registry/datasetio.py
index 43cde83fb..f641b4ce3 100644
--- a/llama_stack/providers/registry/datasetio.py
+++ b/llama_stack/providers/registry/datasetio.py
@@ -30,7 +30,7 @@ def available_providers() -> list[ProviderSpec]:
             adapter=AdapterSpec(
                 adapter_type="huggingface",
                 pip_packages=[
-                    "datasets",
+                    "datasets>=4.0.0",
                 ],
                 module="llama_stack.providers.remote.datasetio.huggingface",
                 config_class="llama_stack.providers.remote.datasetio.huggingface.HuggingfaceDatasetIOConfig",
@@ -42,7 +42,7 @@ def available_providers() -> list[ProviderSpec]:
             adapter=AdapterSpec(
                 adapter_type="nvidia",
                 pip_packages=[
-                    "datasets",
+                    "datasets>=4.0.0",
                 ],
                 module="llama_stack.providers.remote.datasetio.nvidia",
                 config_class="llama_stack.providers.remote.datasetio.nvidia.NvidiaDatasetIOConfig",
diff --git a/llama_stack/providers/registry/post_training.py b/llama_stack/providers/registry/post_training.py
index 67238e3fc..47aeb401e 100644
--- a/llama_stack/providers/registry/post_training.py
+++ b/llama_stack/providers/registry/post_training.py
@@ -48,7 +48,7 @@ def available_providers() -> list[ProviderSpec]:
         InlineProviderSpec(
             api=Api.post_training,
             provider_type="inline::huggingface-gpu",
-            pip_packages=["trl", "transformers", "peft", "datasets", "torch"],
+            pip_packages=["trl", "transformers", "peft", "datasets>=4.0.0", "torch"],
             module="llama_stack.providers.inline.post_training.huggingface",
             config_class="llama_stack.providers.inline.post_training.huggingface.HuggingFacePostTrainingConfig",
             api_dependencies=[
diff --git a/llama_stack/ui/package-lock.json b/llama_stack/ui/package-lock.json
index cbe8e5557..1db1c61cd 100644
--- a/llama_stack/ui/package-lock.json
+++ b/llama_stack/ui/package-lock.json
@@ -18,7 +18,7 @@
         "class-variance-authority": "^0.7.1",
         "clsx": "^2.1.1",
         "framer-motion": "^12.23.12",
-        "llama-stack-client": "^0.2.20",
+        "llama-stack-client": "^0.2.21",
         "lucide-react": "^0.542.0",
         "next": "15.3.3",
         "next-auth": "^4.24.11",
@@ -10278,9 +10278,9 @@
       "license": "MIT"
     },
     "node_modules/llama-stack-client": {
-      "version": "0.2.20",
-      "resolved": "https://registry.npmjs.org/llama-stack-client/-/llama-stack-client-0.2.20.tgz",
-      "integrity": "sha512-1vD5nizTX5JEW8TADxKgy/P1W8YZoPSpdnmfxbdYbWgpQ3BWtbvLS6jmDk7VwVA5fRC4895VfHsRDfS1liHarw==",
+      "version": "0.2.21",
+      "resolved": "https://registry.npmjs.org/llama-stack-client/-/llama-stack-client-0.2.21.tgz",
+      "integrity": "sha512-rjU2Vx5xStxDYavU8K1An/SYXiQQjroLcK98B+p0Paz/a7OgRao2S0YwvThJjPUyChY4fO03UIXP9LpmHqlXWQ==",
       "license": "MIT",
       "dependencies": {
         "@types/node": "^18.11.18",
diff --git a/llama_stack/ui/package.json b/llama_stack/ui/package.json
index e817e9ae3..e50401fa6 100644
--- a/llama_stack/ui/package.json
+++ b/llama_stack/ui/package.json
@@ -23,7 +23,7 @@
     "class-variance-authority": "^0.7.1",
     "clsx": "^2.1.1",
     "framer-motion": "^12.23.12",
-    "llama-stack-client": "^0.2.20",
+    "llama-stack-client": "^0.2.21",
     "lucide-react": "^0.542.0",
     "next": "15.3.3",
     "next-auth": "^4.24.11",
diff --git a/pyproject.toml b/pyproject.toml
index fb6d3a330..0414aafb0 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -7,7 +7,7 @@ required-version = ">=0.7.0"
 
 [project]
 name = "llama_stack"
-version = "0.2.20"
+version = "0.2.21"
 authors = [{ name = "Meta Llama", email = "llama-oss@meta.com" }]
 description = "Llama Stack"
 readme = "README.md"
@@ -31,7 +31,7 @@ dependencies = [
     "huggingface-hub>=0.34.0,<1.0",
     "jinja2>=3.1.6",
     "jsonschema",
-    "llama-stack-client>=0.2.20",
+    "llama-stack-client>=0.2.21",
     "openai>=1.99.6",
     "prompt-toolkit",
     "python-dotenv",
@@ -55,7 +55,7 @@ dependencies = [
 ui = [
     "streamlit",
     "pandas",
-    "llama-stack-client>=0.2.20",
+    "llama-stack-client>=0.2.21",
     "streamlit-option-menu",
 ]
 
@@ -114,7 +114,7 @@ test = [
     "psycopg2-binary>=2.9.0",
     "pypdf",
     "mcp",
-    "datasets",
+    "datasets>=4.0.0",
     "autoevals",
     "transformers",
     "sqlalchemy",
diff --git a/scripts/get_setup_env.py b/scripts/get_setup_env.py
new file mode 100644
index 000000000..fad601e76
--- /dev/null
+++ b/scripts/get_setup_env.py
@@ -0,0 +1,71 @@
+#!/usr/bin/env python3
+# Copyright (c) Meta Platforms, Inc. and affiliates.
+# All rights reserved.
+#
+# This source code is licensed under the terms described in the LICENSE file in
+# the root directory of this source tree.
+
+"""
+Small helper script to extract environment variables from a test setup.
+Used by integration-tests.sh to set environment variables before starting the server.
+"""
+
+import argparse
+import sys
+
+from tests.integration.suites import SETUP_DEFINITIONS, SUITE_DEFINITIONS
+
+
+def get_setup_env_vars(setup_name, suite_name=None):
+    """
+    Get environment variables for a setup, with optional suite default fallback.
+
+    Args:
+        setup_name: Name of the setup (e.g., 'ollama', 'gpt')
+        suite_name: Optional suite name to get default setup if setup_name is None
+
+    Returns:
+        Dictionary of environment variables
+    """
+    # If no setup specified, try to get default from suite
+    if not setup_name and suite_name:
+        suite = SUITE_DEFINITIONS.get(suite_name)
+        if suite and suite.default_setup:
+            setup_name = suite.default_setup
+
+    if not setup_name:
+        return {}
+
+    setup = SETUP_DEFINITIONS.get(setup_name)
+    if not setup:
+        print(
+            f"Error: Unknown setup '{setup_name}'. Available: {', '.join(sorted(SETUP_DEFINITIONS.keys()))}",
+            file=sys.stderr,
+        )
+        sys.exit(1)
+
+    return setup.env
+
+
+def main():
+    parser = argparse.ArgumentParser(description="Extract environment variables from a test setup")
+    parser.add_argument("--setup", help="Setup name (e.g., ollama, gpt)")
+    parser.add_argument("--suite", help="Suite name to get default setup from if --setup not provided")
+    parser.add_argument("--format", choices=["bash", "json"], default="bash", help="Output format (default: bash)")
+
+    args = parser.parse_args()
+
+    env_vars = get_setup_env_vars(args.setup, args.suite)
+
+    if args.format == "bash":
+        # Output as bash export statements
+        for key, value in env_vars.items():
+            print(f"export {key}='{value}'")
+    elif args.format == "json":
+        import json
+
+        print(json.dumps(env_vars))
+
+
+if __name__ == "__main__":
+    main()
diff --git a/scripts/github/schedule-record-workflow.sh b/scripts/github/schedule-record-workflow.sh
index 09e055611..c292e53e6 100755
--- a/scripts/github/schedule-record-workflow.sh
+++ b/scripts/github/schedule-record-workflow.sh
@@ -14,7 +14,7 @@ set -euo pipefail
 # Default values
 BRANCH=""
 TEST_SUBDIRS=""
-TEST_PROVIDER="ollama"
+TEST_SETUP="ollama"
 TEST_SUITE="base"
 TEST_PATTERN=""
 
@@ -27,24 +27,24 @@ Trigger the integration test recording workflow remotely. This way you do not ne
 
 OPTIONS:
     -b, --branch BRANCH         Branch to run the workflow on (defaults to current branch)
-    -p, --test-provider PROVIDER Test provider to use: vllm or ollama (default: ollama)
-    -t, --test-suite SUITE      Test suite to use: base, responses, vision, etc. (default: base)
-    -s, --test-subdirs DIRS     Comma-separated list of test subdirectories to run (overrides suite)
-    -k, --test-pattern PATTERN  Regex pattern to pass to pytest -k
+    -t, --suite SUITE           Test suite to use: base, responses, vision, etc. (default: base)
+    -p, --setup SETUP           Test setup to use: vllm, ollama, gpt, etc. (default: ollama)
+    -s, --subdirs DIRS          Comma-separated list of test subdirectories to run (overrides suite)
+    -k, --pattern PATTERN       Regex pattern to pass to pytest -k
     -h, --help                  Show this help message
 
 EXAMPLES:
     # Record tests for current branch with agents subdirectory
-    $0 --test-subdirs "agents"
+    $0 --subdirs "agents"
 
     # Record tests for specific branch with vision tests
-    $0 -b my-feature-branch --test-suite vision
+    $0 -b my-feature-branch --suite vision
 
-    # Record multiple test subdirectories with specific provider
-    $0 --test-subdirs "agents,inference" --test-provider vllm
+    # Record multiple test subdirectories with specific setup
+    $0 --subdirs "agents,inference" --setup vllm
 
     # Record tests matching a specific pattern
-    $0 --test-subdirs "inference" --test-pattern "test_streaming"
+    $0 --subdirs "inference" --pattern "test_streaming"
 
 EOF
 }
@@ -63,19 +63,19 @@ while [[ $# -gt 0 ]]; do
             BRANCH="$2"
             shift 2
             ;;
-        -s|--test-subdirs)
+        -s|--subdirs)
             TEST_SUBDIRS="$2"
             shift 2
             ;;
-        -p|--test-provider)
-            TEST_PROVIDER="$2"
+        -p|--setup)
+            TEST_SETUP="$2"
             shift 2
             ;;
-        -t|--test-suite)
+        -t|--suite)
             TEST_SUITE="$2"
             shift 2
             ;;
-        -k|--test-pattern)
+        -k|--pattern)
             TEST_PATTERN="$2"
             shift 2
             ;;
@@ -93,21 +93,16 @@ done
 
 # Validate required parameters
 if [[ -z "$TEST_SUBDIRS" && -z "$TEST_SUITE" ]]; then
-    echo "Error: --test-subdirs or --test-suite is required"
+    echo "Error: --subdirs or --suite is required"
     echo "Please specify which test subdirectories to run or test suite to use, e.g.:"
-    echo "  $0 --test-subdirs \"agents,inference\""
-    echo "  $0 --test-suite vision"
+    echo "  $0 --subdirs \"agents,inference\""
+    echo "  $0 --suite vision"
     echo ""
     exit 1
 fi
 
-# Validate test provider
-if [[ "$TEST_PROVIDER" != "vllm" && "$TEST_PROVIDER" != "ollama" ]]; then
-    echo "❌ Error: Invalid test provider '$TEST_PROVIDER'"
-    echo "   Supported providers: vllm, ollama"
-    echo "   Example: $0 --test-subdirs \"agents\" --test-provider vllm"
-    exit 1
-fi
+# Validate test setup (optional - setups are validated by the workflow itself)
+# Common setups: ollama, vllm, gpt, etc.
 
 # Check if required tools are installed
 if ! command -v gh &> /dev/null; then
@@ -237,7 +232,7 @@ fi
 # Build the workflow dispatch command
 echo "Triggering integration test recording workflow..."
 echo "Branch: $BRANCH"
-echo "Test provider: $TEST_PROVIDER"
+echo "Test setup: $TEST_SETUP"
 echo "Test subdirs: $TEST_SUBDIRS"
 echo "Test suite: $TEST_SUITE"
 echo "Test pattern: ${TEST_PATTERN:-"(none)"}"
@@ -245,16 +240,16 @@ echo ""
 
 # Prepare inputs for gh workflow run
 if [[ -n "$TEST_SUBDIRS" ]]; then
-    INPUTS="-f test-subdirs='$TEST_SUBDIRS'"
+    INPUTS="-f subdirs='$TEST_SUBDIRS'"
 fi
-if [[ -n "$TEST_PROVIDER" ]]; then
-    INPUTS="$INPUTS -f test-provider='$TEST_PROVIDER'"
+if [[ -n "$TEST_SETUP" ]]; then
+    INPUTS="$INPUTS -f test-setup='$TEST_SETUP'"
 fi
 if [[ -n "$TEST_SUITE" ]]; then
-    INPUTS="$INPUTS -f test-suite='$TEST_SUITE'"
+    INPUTS="$INPUTS -f suite='$TEST_SUITE'"
 fi
 if [[ -n "$TEST_PATTERN" ]]; then
-    INPUTS="$INPUTS -f test-pattern='$TEST_PATTERN'"
+    INPUTS="$INPUTS -f pattern='$TEST_PATTERN'"
 fi
 
 # Run the workflow
diff --git a/scripts/integration-tests.sh b/scripts/integration-tests.sh
index ab7e37579..eee60951d 100755
--- a/scripts/integration-tests.sh
+++ b/scripts/integration-tests.sh
@@ -13,10 +13,10 @@ set -euo pipefail
 
 # Default values
 STACK_CONFIG=""
-PROVIDER=""
+TEST_SUITE="base"
+TEST_SETUP=""
 TEST_SUBDIRS=""
 TEST_PATTERN=""
-TEST_SUITE="base"
 INFERENCE_MODE="replay"
 EXTRA_PARAMS=""
 
@@ -27,29 +27,30 @@ Usage: $0 [OPTIONS]
 
 Options:
     --stack-config STRING    Stack configuration to use (required)
-    --provider STRING        Provider to use (ollama, vllm, etc.) (required)
-    --test-suite STRING      Comma-separated list of test suites to run (default: 'base')
+    --suite STRING           Test suite to run (default: 'base')
+    --setup STRING           Test setup (models, env) to use (e.g., 'ollama', 'ollama-vision', 'gpt', 'vllm')
     --inference-mode STRING  Inference mode: record or replay (default: replay)
-    --test-subdirs STRING    Comma-separated list of test subdirectories to run (overrides suite)
-    --test-pattern STRING    Regex pattern to pass to pytest -k
+    --subdirs STRING         Comma-separated list of test subdirectories to run (overrides suite)
+    --pattern STRING         Regex pattern to pass to pytest -k
     --help                   Show this help message
 
-Suites are defined in tests/integration/suites.py. They are used to narrow the collection of tests and provide default model options.
+Suites are defined in tests/integration/suites.py and define which tests to run.
+Setups are defined in tests/integration/setups.py and provide global configuration (models, env).
 
 You can also specify subdirectories (of tests/integration) to select tests from, which will override the suite.
 
 Examples:
     # Basic inference tests with ollama
-    $0 --stack-config server:ci-tests --provider ollama
+    $0 --stack-config server:ci-tests --suite base --setup ollama
 
     # Multiple test directories with vllm
-    $0 --stack-config server:ci-tests --provider vllm --test-subdirs 'inference,agents'
+    $0 --stack-config server:ci-tests --subdirs 'inference,agents' --setup vllm
 
     # Vision tests with ollama
-    $0 --stack-config server:ci-tests --provider ollama --test-suite vision
+    $0 --stack-config server:ci-tests --suite vision  # default setup for this suite is ollama-vision
 
     # Record mode for updating test recordings
-    $0 --stack-config server:ci-tests --provider ollama --inference-mode record
+    $0 --stack-config server:ci-tests --suite base --inference-mode record
 EOF
 }
 
@@ -60,15 +61,15 @@ while [[ $# -gt 0 ]]; do
             STACK_CONFIG="$2"
             shift 2
             ;;
-        --provider)
-            PROVIDER="$2"
+        --setup)
+            TEST_SETUP="$2"
             shift 2
             ;;
-        --test-subdirs)
+        --subdirs)
             TEST_SUBDIRS="$2"
             shift 2
             ;;
-        --test-suite)
+        --suite)
             TEST_SUITE="$2"
             shift 2
             ;;
@@ -76,7 +77,7 @@ while [[ $# -gt 0 ]]; do
             INFERENCE_MODE="$2"
             shift 2
             ;;
-        --test-pattern)
+        --pattern)
             TEST_PATTERN="$2"
             shift 2
             ;;
@@ -96,11 +97,13 @@ done
 # Validate required parameters
 if [[ -z "$STACK_CONFIG" ]]; then
     echo "Error: --stack-config is required"
+    usage
     exit 1
 fi
 
-if [[ -z "$PROVIDER" ]]; then
-    echo "Error: --provider is required"
+if [[ -z "$TEST_SETUP" && -n "$TEST_SUBDIRS" ]]; then
+    echo "Error: --test-setup is required when --test-subdirs is provided"
+    usage
     exit 1
 fi
 
@@ -111,7 +114,7 @@ fi
 
 echo "=== Llama Stack Integration Test Runner ==="
 echo "Stack Config: $STACK_CONFIG"
-echo "Provider: $PROVIDER"
+echo "Setup: $TEST_SETUP"
 echo "Inference Mode: $INFERENCE_MODE"
 echo "Test Suite: $TEST_SUITE"
 echo "Test Subdirs: $TEST_SUBDIRS"
@@ -129,21 +132,25 @@ echo ""
 
 # Set environment variables
 export LLAMA_STACK_CLIENT_TIMEOUT=300
-export LLAMA_STACK_TEST_INFERENCE_MODE="$INFERENCE_MODE"
-
-# Configure provider-specific settings
-if [[ "$PROVIDER" == "ollama" ]]; then
-    export OLLAMA_URL="http://0.0.0.0:11434"
-    export TEXT_MODEL="ollama/llama3.2:3b-instruct-fp16"
-    export SAFETY_MODEL="ollama/llama-guard3:1b"
-    EXTRA_PARAMS="--safety-shield=llama-guard"
-else
-    export VLLM_URL="http://localhost:8000/v1"
-    export TEXT_MODEL="vllm/meta-llama/Llama-3.2-1B-Instruct"
-    EXTRA_PARAMS=""
-fi
 
 THIS_DIR=$(dirname "$0")
+
+if [[ -n "$TEST_SETUP" ]]; then
+    EXTRA_PARAMS="--setup=$TEST_SETUP"
+fi
+
+# Apply setup-specific environment variables (needed for server startup and tests)
+echo "=== Applying Setup Environment Variables ==="
+
+# the server needs this
+export LLAMA_STACK_TEST_INFERENCE_MODE="$INFERENCE_MODE"
+
+SETUP_ENV=$(PYTHONPATH=$THIS_DIR/.. python "$THIS_DIR/get_setup_env.py" --suite "$TEST_SUITE" --setup "$TEST_SETUP" --format bash)
+echo "Setting up environment variables:"
+echo "$SETUP_ENV"
+eval "$SETUP_ENV"
+echo ""
+
 ROOT_DIR="$THIS_DIR/.."
 cd $ROOT_DIR
 
@@ -162,6 +169,18 @@ fi
 
 # Start Llama Stack Server if needed
 if [[ "$STACK_CONFIG" == *"server:"* ]]; then
+    stop_server() {
+        echo "Stopping Llama Stack Server..."
+        pids=$(lsof -i :8321 | awk 'NR>1 {print $2}')
+        if [[ -n "$pids" ]]; then
+            echo "Killing Llama Stack Server processes: $pids"
+            kill -9 $pids
+        else
+            echo "No Llama Stack Server processes found ?!"
+        fi
+        echo "Llama Stack Server stopped"
+    }
+
     # check if server is already running
     if curl -s http://localhost:8321/v1/health 2>/dev/null | grep -q "OK"; then
         echo "Llama Stack Server is already running, skipping start"
@@ -185,14 +204,16 @@ if [[ "$STACK_CONFIG" == *"server:"* ]]; then
         done
         echo ""
     fi
+
+    trap stop_server EXIT ERR INT TERM
 fi
 
 # Run tests
 echo "=== Running Integration Tests ==="
 EXCLUDE_TESTS="builtin_tool or safety_with_image or code_interpreter or test_rag"
 
-# Additional exclusions for vllm provider
-if [[ "$PROVIDER" == "vllm" ]]; then
+# Additional exclusions for vllm setup
+if [[ "$TEST_SETUP" == "vllm" ]]; then
     EXCLUDE_TESTS="${EXCLUDE_TESTS} or test_inference_store_tool_calls"
 fi
 
@@ -229,20 +250,22 @@ if [[ -n "$TEST_SUBDIRS" ]]; then
     echo "Total test files: $(echo $TEST_FILES | wc -w)"
 
     PYTEST_TARGET="$TEST_FILES"
-    EXTRA_PARAMS="$EXTRA_PARAMS --text-model=$TEXT_MODEL --embedding-model=sentence-transformers/all-MiniLM-L6-v2"
 else
     PYTEST_TARGET="tests/integration/"
     EXTRA_PARAMS="$EXTRA_PARAMS --suite=$TEST_SUITE"
 fi
 
 set +e
+set -x
 pytest -s -v $PYTEST_TARGET \
     --stack-config="$STACK_CONFIG" \
+    --inference-mode="$INFERENCE_MODE" \
     -k "$PYTEST_PATTERN" \
     $EXTRA_PARAMS \
     --color=yes \
     --capture=tee-sys
 exit_code=$?
+set +x
 set -e
 
 if [ $exit_code -eq 0 ]; then
@@ -260,18 +283,5 @@ echo "=== System Resources After Tests ==="
 free -h 2>/dev/null || echo "free command not available"
 df -h
 
-# stop server
-if [[ "$STACK_CONFIG" == *"server:"* ]]; then
-    echo "Stopping Llama Stack Server..."
-    pids=$(lsof -i :8321 | awk 'NR>1 {print $2}')
-    if [[ -n "$pids" ]]; then
-        echo "Killing Llama Stack Server processes: $pids"
-        kill -9 $pids
-    else
-        echo "No Llama Stack Server processes found ?!"
-    fi
-    echo "Llama Stack Server stopped"
-fi
-
 echo ""
 echo "=== Integration Tests Complete ==="
diff --git a/tests/integration/README.md b/tests/integration/README.md
index b05beeb98..467f97e02 100644
--- a/tests/integration/README.md
+++ b/tests/integration/README.md
@@ -6,9 +6,7 @@ Integration tests verify complete workflows across different providers using Lla
 
 ```bash
 # Run all integration tests with existing recordings
-LLAMA_STACK_TEST_INFERENCE_MODE=replay \
-  LLAMA_STACK_TEST_RECORDING_DIR=tests/integration/recordings \
-  uv run --group test \
+uv run --group test \
   pytest -sv tests/integration/ --stack-config=starter
 ```
 
@@ -42,25 +40,35 @@ Model parameters can be influenced by the following options:
 Each of these are comma-separated lists and can be used to generate multiple parameter combinations. Note that tests will be skipped
 if no model is specified.
 
-### Suites (fast selection + sane defaults)
+### Suites and Setups
 
-- `--suite`: comma-separated list of named suites that both narrow which tests are collected and prefill common model options (unless you pass them explicitly).
+- `--suite`: single named suite that narrows which tests are collected.
 - Available suites:
-  - `responses`: collects tests under `tests/integration/responses`; this is a separate suite because it needs a strong tool-calling model.
-  - `vision`: collects only `tests/integration/inference/test_vision_inference.py`; defaults `--vision-model=ollama/llama3.2-vision:11b`, `--embedding-model=sentence-transformers/all-MiniLM-L6-v2`.
-- Explicit flags always win. For example, `--suite=responses --text-model=<X>` overrides the suite’s text model.
+  - `base`: collects most tests (excludes responses and post_training)
+  - `responses`: collects tests under `tests/integration/responses` (needs strong tool-calling models)
+  - `vision`: collects only `tests/integration/inference/test_vision_inference.py`
+- `--setup`: global configuration that can be used with any suite. Setups prefill model/env defaults; explicit CLI flags always win.
+  - Available setups:
+    - `ollama`: Local Ollama provider with lightweight models (sets OLLAMA_URL, uses llama3.2:3b-instruct-fp16)
+    - `vllm`: VLLM provider for efficient local inference (sets VLLM_URL, uses Llama-3.2-1B-Instruct)
+    - `gpt`: OpenAI GPT models for high-quality responses (uses gpt-4o)
+    - `claude`: Anthropic Claude models for high-quality responses (uses claude-3-5-sonnet)
 
-Examples:
+Examples
 
 ```bash
-# Fast responses run with defaults
-pytest -s -v tests/integration --stack-config=server:starter --suite=responses
+# Fast responses run with a strong tool-calling model
+pytest -s -v tests/integration --stack-config=server:starter --suite=responses --setup=gpt
 
-# Fast single-file vision run with defaults
-pytest -s -v tests/integration --stack-config=server:starter --suite=vision
+# Fast single-file vision run with Ollama defaults
+pytest -s -v tests/integration --stack-config=server:starter --suite=vision --setup=ollama
 
-# Combine suites and override a default
-pytest -s -v tests/integration --stack-config=server:starter --suite=responses,vision --embedding-model=text-embedding-3-small
+# Base suite with VLLM for performance
+pytest -s -v tests/integration --stack-config=server:starter --suite=base --setup=vllm
+
+# Override a default from setup
+pytest -s -v tests/integration --stack-config=server:starter \
+  --suite=responses --setup=gpt --embedding-model=text-embedding-3-small
 ```
 
 ## Examples
@@ -127,14 +135,13 @@ pytest tests/integration/
 ### RECORD Mode
 Captures API interactions for later replay:
 ```bash
-LLAMA_STACK_TEST_INFERENCE_MODE=record \
-pytest tests/integration/inference/test_new_feature.py
+pytest tests/integration/inference/test_new_feature.py --inference-mode=record
 ```
 
 ### LIVE Mode
 Tests make real API calls (but not recorded):
 ```bash
-LLAMA_STACK_TEST_INFERENCE_MODE=live pytest tests/integration/
+pytest tests/integration/ --inference-mode=live
 ```
 
 By default, the recording directory is `tests/integration/recordings`. You can override this by setting the `LLAMA_STACK_TEST_RECORDING_DIR` environment variable.
@@ -155,15 +162,14 @@ cat recordings/responses/abc123.json | jq '.'
 #### Remote Re-recording (Recommended)
 Use the automated workflow script for easier re-recording:
 ```bash
-./scripts/github/schedule-record-workflow.sh --test-subdirs "inference,agents"
+./scripts/github/schedule-record-workflow.sh --subdirs "inference,agents"
 ```
 See the [main testing guide](../README.md#remote-re-recording-recommended) for full details.
 
 #### Local Re-recording
 ```bash
 # Re-record specific tests
-LLAMA_STACK_TEST_INFERENCE_MODE=record \
-pytest -s -v --stack-config=server:starter tests/integration/inference/test_modified.py
+pytest -s -v --stack-config=server:starter tests/integration/inference/test_modified.py --inference-mode=record
 ```
 
 Note that when re-recording tests, you must use a Stack pointing to a server (i.e., `server:starter`). This subtlety exists because the set of tests run in server are a superset of the set of tests run in the library client.
diff --git a/tests/integration/conftest.py b/tests/integration/conftest.py
index 96260fdb7..4735264c3 100644
--- a/tests/integration/conftest.py
+++ b/tests/integration/conftest.py
@@ -15,7 +15,7 @@ from dotenv import load_dotenv
 
 from llama_stack.log import get_logger
 
-from .suites import SUITE_DEFINITIONS
+from .suites import SETUP_DEFINITIONS, SUITE_DEFINITIONS
 
 logger = get_logger(__name__, category="tests")
 
@@ -63,19 +63,33 @@ def pytest_configure(config):
         key, value = env_var.split("=", 1)
         os.environ[key] = value
 
-    suites_raw = config.getoption("--suite")
-    suites: list[str] = []
-    if suites_raw:
-        suites = [p.strip() for p in str(suites_raw).split(",") if p.strip()]
-        unknown = [p for p in suites if p not in SUITE_DEFINITIONS]
-        if unknown:
+    inference_mode = config.getoption("--inference-mode")
+    os.environ["LLAMA_STACK_TEST_INFERENCE_MODE"] = inference_mode
+
+    suite = config.getoption("--suite")
+    if suite:
+        if suite not in SUITE_DEFINITIONS:
+            raise pytest.UsageError(f"Unknown suite: {suite}. Available: {', '.join(sorted(SUITE_DEFINITIONS.keys()))}")
+
+    # Apply setups (global parameterizations): env + defaults
+    setup = config.getoption("--setup")
+    if suite and not setup:
+        setup = SUITE_DEFINITIONS[suite].default_setup
+
+    if setup:
+        if setup not in SETUP_DEFINITIONS:
             raise pytest.UsageError(
-                f"Unknown suite(s): {', '.join(unknown)}. Available: {', '.join(sorted(SUITE_DEFINITIONS.keys()))}"
+                f"Unknown setup '{setup}'. Available: {', '.join(sorted(SETUP_DEFINITIONS.keys()))}"
             )
-    for suite in suites:
-        suite_def = SUITE_DEFINITIONS.get(suite, {})
-        defaults: dict = suite_def.get("defaults", {})
-        for dest, value in defaults.items():
+
+        setup_obj = SETUP_DEFINITIONS[setup]
+        logger.info(f"Applying setup '{setup}'{' for suite ' + suite if suite else ''}")
+        # Apply env first
+        for k, v in setup_obj.env.items():
+            if k not in os.environ:
+                os.environ[k] = str(v)
+        # Apply defaults if not provided explicitly
+        for dest, value in setup_obj.defaults.items():
             current = getattr(config.option, dest, None)
             if not current:
                 setattr(config.option, dest, value)
@@ -120,6 +134,13 @@ def pytest_addoption(parser):
         default=384,
         help="Output dimensionality of the embedding model to use for testing. Default: 384",
     )
+
+    parser.addoption(
+        "--inference-mode",
+        help="Inference mode: { record, replay, live } (default: replay)",
+        choices=["record", "replay", "live"],
+        default="replay",
+    )
     parser.addoption(
         "--report",
         help="Path where the test report should be written, e.g. --report=/path/to/report.md",
@@ -127,14 +148,18 @@ def pytest_addoption(parser):
 
     available_suites = ", ".join(sorted(SUITE_DEFINITIONS.keys()))
     suite_help = (
-        "Comma-separated integration test suites to narrow collection and prefill defaults. "
-        "Available: "
-        f"{available_suites}. "
-        "Explicit CLI flags (e.g., --text-model) override suite defaults. "
-        "Examples: --suite=responses or --suite=responses,vision."
+        f"Single test suite to run (narrows collection). Available: {available_suites}. Example: --suite=responses"
     )
     parser.addoption("--suite", help=suite_help)
 
+    # Global setups for any suite
+    available_setups = ", ".join(sorted(SETUP_DEFINITIONS.keys()))
+    setup_help = (
+        f"Global test setup configuration. Available: {available_setups}. "
+        "Can be used with any suite. Example: --setup=ollama"
+    )
+    parser.addoption("--setup", help=setup_help)
+
 
 MODEL_SHORT_IDS = {
     "meta-llama/Llama-3.2-3B-Instruct": "3B",
@@ -221,16 +246,12 @@ pytest_plugins = ["tests.integration.fixtures.common"]
 
 def pytest_ignore_collect(path: str, config: pytest.Config) -> bool:
     """Skip collecting paths outside the selected suite roots for speed."""
-    suites_raw = config.getoption("--suite")
-    if not suites_raw:
+    suite = config.getoption("--suite")
+    if not suite:
         return False
 
-    names = [p.strip() for p in str(suites_raw).split(",") if p.strip()]
-    roots: list[str] = []
-    for name in names:
-        suite_def = SUITE_DEFINITIONS.get(name)
-        if suite_def:
-            roots.extend(suite_def.get("roots", []))
+    sobj = SUITE_DEFINITIONS.get(suite)
+    roots: list[str] = sobj.get("roots", []) if isinstance(sobj, dict) else getattr(sobj, "roots", [])
     if not roots:
         return False
 
diff --git a/tests/integration/suites.py b/tests/integration/suites.py
index 602855055..bacd7ef52 100644
--- a/tests/integration/suites.py
+++ b/tests/integration/suites.py
@@ -8,46 +8,112 @@
 # For example:
 #
 # ```bash
-# pytest tests/integration/ --suite=vision
+# pytest tests/integration/ --suite=vision --setup=ollama
 # ```
 #
-# Each suite can:
-# - restrict collection to specific roots (dirs or files)
-# - provide default CLI option values (e.g. text_model, embedding_model, etc.)
+"""
+Each suite defines what to run (roots). Suites can be run with different global setups defined in setups.py.
+Setups provide environment variables and model defaults that can be reused across multiple suites.
+
+CLI examples:
+  pytest tests/integration --suite=responses --setup=gpt
+  pytest tests/integration --suite=vision --setup=ollama
+  pytest tests/integration --suite=base --setup=vllm
+"""
 
 from pathlib import Path
 
+from pydantic import BaseModel, Field
+
 this_dir = Path(__file__).parent
-default_roots = [
+
+
+class Suite(BaseModel):
+    name: str
+    roots: list[str]
+    default_setup: str | None = None
+
+
+class Setup(BaseModel):
+    """A reusable test configuration with environment and CLI defaults."""
+
+    name: str
+    description: str
+    defaults: dict[str, str] = Field(default_factory=dict)
+    env: dict[str, str] = Field(default_factory=dict)
+
+
+# Global setups - can be used with any suite "technically" but in reality, some setups might work
+# only for specific test suites.
+SETUP_DEFINITIONS: dict[str, Setup] = {
+    "ollama": Setup(
+        name="ollama",
+        description="Local Ollama provider with text + safety models",
+        env={
+            "OLLAMA_URL": "http://0.0.0.0:11434",
+            "SAFETY_MODEL": "ollama/llama-guard3:1b",
+        },
+        defaults={
+            "text_model": "ollama/llama3.2:3b-instruct-fp16",
+            "embedding_model": "sentence-transformers/all-MiniLM-L6-v2",
+            "safety_model": "ollama/llama-guard3:1b",
+            "safety_shield": "llama-guard",
+        },
+    ),
+    "ollama-vision": Setup(
+        name="ollama",
+        description="Local Ollama provider with a vision model",
+        env={
+            "OLLAMA_URL": "http://0.0.0.0:11434",
+        },
+        defaults={
+            "vision_model": "ollama/llama3.2-vision:11b",
+            "embedding_model": "sentence-transformers/all-MiniLM-L6-v2",
+        },
+    ),
+    "vllm": Setup(
+        name="vllm",
+        description="vLLM provider with a text model",
+        env={
+            "VLLM_URL": "http://localhost:8000/v1",
+        },
+        defaults={
+            "text_model": "vllm/meta-llama/Llama-3.2-1B-Instruct",
+            "embedding_model": "sentence-transformers/all-MiniLM-L6-v2",
+        },
+    ),
+    "gpt": Setup(
+        name="gpt",
+        description="OpenAI GPT models for high-quality responses and tool calling",
+        defaults={
+            "text_model": "openai/gpt-4o",
+            "embedding_model": "sentence-transformers/all-MiniLM-L6-v2",
+        },
+    ),
+}
+
+
+base_roots = [
     str(p)
     for p in this_dir.glob("*")
     if p.is_dir()
     and p.name not in ("__pycache__", "fixtures", "test_cases", "recordings", "responses", "post_training")
 ]
 
-SUITE_DEFINITIONS: dict[str, dict] = {
-    "base": {
-        "description": "Base suite that includes most tests but runs them with a text Ollama model",
-        "roots": default_roots,
-        "defaults": {
-            "text_model": "ollama/llama3.2:3b-instruct-fp16",
-            "embedding_model": "sentence-transformers/all-MiniLM-L6-v2",
-        },
-    },
-    "responses": {
-        "description": "Suite that includes only the OpenAI Responses tests; needs a strong tool-calling model",
-        "roots": ["tests/integration/responses"],
-        "defaults": {
-            "text_model": "openai/gpt-4o",
-            "embedding_model": "sentence-transformers/all-MiniLM-L6-v2",
-        },
-    },
-    "vision": {
-        "description": "Suite that includes only the vision tests",
-        "roots": ["tests/integration/inference/test_vision_inference.py"],
-        "defaults": {
-            "vision_model": "ollama/llama3.2-vision:11b",
-            "embedding_model": "sentence-transformers/all-MiniLM-L6-v2",
-        },
-    },
+SUITE_DEFINITIONS: dict[str, Suite] = {
+    "base": Suite(
+        name="base",
+        roots=base_roots,
+        default_setup="ollama",
+    ),
+    "responses": Suite(
+        name="responses",
+        roots=["tests/integration/responses"],
+        default_setup="gpt",
+    ),
+    "vision": Suite(
+        name="vision",
+        roots=["tests/integration/inference/test_vision_inference.py"],
+        default_setup="ollama-vision",
+    ),
 }
diff --git a/uv.lock b/uv.lock
index 43cc59c7a..2788c6fef 100644
--- a/uv.lock
+++ b/uv.lock
@@ -895,7 +895,6 @@ dependencies = [
     { name = "numpy" },
     { name = "packaging" },
 ]
-sdist = { url = "https://files.pythonhosted.org/packages/5c/f4/7c2136f4660ca504266cc08b38df2aa1db14fea93393b82e099ff34d7290/faiss_cpu-1.11.0.post1.tar.gz", hash = "sha256:06b1ea9ddec9e4d9a41c8ef7478d493b08d770e9a89475056e963081eed757d1", size = 70543, upload-time = "2025-07-15T09:15:02.127Z" }
 wheels = [
     { url = "https://files.pythonhosted.org/packages/30/1e/9980758efa55b4e7a5d6df1ae17c9ddbe5a636bfbf7d22d47c67f7a530f4/faiss_cpu-1.11.0.post1-cp312-cp312-macosx_13_0_x86_64.whl", hash = "sha256:68f6ce2d9c510a5765af2f5711bd76c2c37bd598af747f3300224bdccf45378c", size = 7913676, upload-time = "2025-07-15T09:14:06.077Z" },
     { url = "https://files.pythonhosted.org/packages/05/d1/bd785887085faa02916c52320527b8bb54288835b0a3138df89a0e323cc8/faiss_cpu-1.11.0.post1-cp312-cp312-macosx_14_0_arm64.whl", hash = "sha256:b940c530a8236cc0b9fd9d6e87b3d70b9c6c216bc2baf2649356c908902e52c9", size = 3313952, upload-time = "2025-07-15T09:14:07.584Z" },
@@ -1748,10 +1747,9 @@ wheels = [
     { url = "https://files.pythonhosted.org/packages/5f/e4/f1546746049c99c6b8b247e2f34485b9eae36faa9322b84e2a17262e6712/litellm-1.74.9-py3-none-any.whl", hash = "sha256:ab8f8a6e4d8689d3c7c4f9c3bbc7e46212cc3ebc74ddd0f3c0c921bb459c9874", size = 8740449, upload-time = "2025-07-28T16:42:36.8Z" },
 ]
 
-
 [[package]]
 name = "llama-stack"
-version = "0.2.20"
+version = "0.2.21"
 source = { editable = "." }
 dependencies = [
     { name = "aiohttp" },
@@ -1889,8 +1887,8 @@ requires-dist = [
     { name = "huggingface-hub", specifier = ">=0.34.0,<1.0" },
     { name = "jinja2", specifier = ">=3.1.6" },
     { name = "jsonschema" },
-    { name = "llama-stack-client", specifier = ">=0.2.20" },
-    { name = "llama-stack-client", marker = "extra == 'ui'", specifier = ">=0.2.20" },
+    { name = "llama-stack-client", specifier = ">=0.2.21" },
+    { name = "llama-stack-client", marker = "extra == 'ui'", specifier = ">=0.2.21" },
     { name = "openai", specifier = ">=1.99.6" },
     { name = "opentelemetry-exporter-otlp-proto-http", specifier = ">=1.30.0" },
     { name = "opentelemetry-sdk", specifier = ">=1.30.0" },
@@ -1958,7 +1956,7 @@ test = [
     { name = "aiosqlite" },
     { name = "autoevals" },
     { name = "chardet" },
-    { name = "datasets" },
+    { name = "datasets", specifier = ">=4.0.0" },
     { name = "mcp" },
     { name = "milvus-lite", specifier = ">=2.5.0" },
     { name = "openai", specifier = ">=1.100.0" },
@@ -1999,7 +1997,7 @@ unit = [
 
 [[package]]
 name = "llama-stack-client"
-version = "0.2.20"
+version = "0.2.21"
 source = { registry = "https://pypi.org/simple" }
 dependencies = [
     { name = "anyio" },
@@ -2018,9 +2016,9 @@ dependencies = [
     { name = "tqdm" },
     { name = "typing-extensions" },
 ]
-sdist = { url = "https://files.pythonhosted.org/packages/21/91/c5e32219a5192825dd601700e68205c815c5cfee60c64c22172e46a0c83e/llama_stack_client-0.2.20.tar.gz", hash = "sha256:356257f0a4bbb64205f89e113d715925853d5e34ec744e72466da72790ba415b", size = 318311, upload-time = "2025-08-29T21:10:12.854Z" }
+sdist = { url = "https://files.pythonhosted.org/packages/f1/d3/8c50561d167f1e9b601b8fffe852b44c1ff97aaa6db6cdedd611d9e02a65/llama_stack_client-0.2.21.tar.gz", hash = "sha256:bd931fdcadedec5ccdbaa3c54d0c17761af1c227711ad6150dc0dd33d7b66ce2", size = 318319, upload-time = "2025-09-08T22:26:57.668Z" }
 wheels = [
-    { url = "https://files.pythonhosted.org/packages/b0/ba/84914c4eead2fd9251c149fd6a7da28b78acd620793e3c4506116645cb60/llama_stack_client-0.2.20-py3-none-any.whl", hash = "sha256:6e178981d2ce971da2145c79d5b2b123fa50e063ed431494975c2ba01c5b8016", size = 369899, upload-time = "2025-08-29T21:10:11.113Z" },
+    { url = "https://files.pythonhosted.org/packages/02/77/dadc682046a2c7ad68be8d2d2afac7007bf4d22efb0d3929d85ab9706ffe/llama_stack_client-0.2.21-py3-none-any.whl", hash = "sha256:adba82fdf18ab3b8ac218cedba4927bd5d26c23c2318e75c8763a44bb6b40693", size = 369902, upload-time = "2025-09-08T22:26:56.308Z" },
 ]
 
 [[package]]