Merge branch 'main' into add-nvidia-inference-adapter

2025-08-01 16:24:44 +00:00 · 2024-11-19 10:25:50 -05:00 · 2024-11-19 10:25:50 -05:00 · 2a25ace2fa
commit 2a25ace2fa
parent c24f882f31 84d5f35a48
131 changed files with 3927 additions and 1286 deletions
--- a/.pre-commit-config.yaml
+++ b/.pre-commit-config.yaml
@ -57,3 +57,17 @@ repos:
 #   hooks:
 #     - id: markdown-link-check
 #       args: ['--quiet']
+
+# -   repo: local
+#     hooks:
+#       - id: distro-codegen
+#         name: Distribution Template Codegen
+#         additional_dependencies:
+#           - rich
+#           - pydantic
+#         entry: python -m llama_stack.scripts.distro_codegen
+#         language: python
+#         pass_filenames: false
+#         require_serial: true
+#         files: ^llama_stack/templates/.*$
+#         stages: [manual]
--- a/CONTRIBUTING.md
+++ b/CONTRIBUTING.md
@ -12,6 +12,11 @@ We actively welcome your pull requests.
 5. Make sure your code lints.
 6. If you haven't already, complete the Contributor License Agreement ("CLA").

+
+### Updating Provider Configurations
+
+If you have made changes to a provider's configuration in any form (introducing a new config key, or changing models, etc.), you should run `python llama_stack/scripts/distro_codegen.py` to re-generate various YAML files as well as the documentation. You should not change `docs/source/.../distributions/` files manually as they are auto-generated.
+
 ### Building the Documentation

 If you are making changes to the documentation at [https://llama-stack.readthedocs.io/en/latest/](https://llama-stack.readthedocs.io/en/latest/), you can use the following command to build the documentation and preview your changes. You will need [Sphinx](https://www.sphinx-doc.org/en/master/) and the readthedocs theme.
@ -26,6 +31,19 @@ make html
 sphinx-autobuild source build/html
 ```

+## Pre-commit Hooks
+
+We use [pre-commit](https://pre-commit.com/) to run linting and formatting checks on your code. You can install the pre-commit hooks by running:
+
+```bash
+$ cd llama-stack
+$ conda activate <your-environment>
+$ pip install pre-commit
+$ pre-commit install
+```
+
+After that, pre-commit hooks will run automatically before each commit.
+
 ## Contributor License Agreement ("CLA")
 In order to accept your pull request, we need you to submit a CLA. You only need
 to do this once to work on any of Meta's open source projects.
--- a/MANIFEST.in
+++ b/MANIFEST.in
@ -1,4 +1,4 @@
 include requirements.txt
 include llama_stack/distribution/*.sh
 include llama_stack/cli/scripts/*.sh
-include llama_stack/templates/*/build.yaml
+include llama_stack/templates/*/*.yaml
--- a/README.md
+++ b/README.md
@ -112,7 +112,7 @@ Please checkout our [Documentations](https://llama-stack.readthedocs.io/en/lates
 | Python |  [llama-stack-client-python](https://github.com/meta-llama/llama-stack-client-python) | [![PyPI version](https://img.shields.io/pypi/v/llama_stack_client.svg)](https://pypi.org/project/llama_stack_client/)
 | Swift  | [llama-stack-client-swift](https://github.com/meta-llama/llama-stack-client-swift) | [![Swift Package Index](https://img.shields.io/endpoint?url=https%3A%2F%2Fswiftpackageindex.com%2Fapi%2Fpackages%2Fmeta-llama%2Fllama-stack-client-swift%2Fbadge%3Ftype%3Dswift-versions)](https://swiftpackageindex.com/meta-llama/llama-stack-client-swift)
 | Node   | [llama-stack-client-node](https://github.com/meta-llama/llama-stack-client-node) | [![NPM version](https://img.shields.io/npm/v/llama-stack-client.svg)](https://npmjs.org/package/llama-stack-client)
-| Kotlin | [llama-stack-client-kotlin](https://github.com/meta-llama/llama-stack-client-kotlin) |
+| Kotlin | [llama-stack-client-kotlin](https://github.com/meta-llama/llama-stack-client-kotlin) | [![Maven version](https://img.shields.io/maven-central/v/com.llama.llamastack/llama-stack-client-kotlin)](https://central.sonatype.com/artifact/com.llama.llamastack/llama-stack-client-kotlin)

 Check out our client SDKs for connecting to Llama Stack server in your preferred language, you can choose from [python](https://github.com/meta-llama/llama-stack-client-python), [node](https://github.com/meta-llama/llama-stack-client-node), [swift](https://github.com/meta-llama/llama-stack-client-swift), and [kotlin](https://github.com/meta-llama/llama-stack-client-kotlin) programming languages to quickly build your applications.

--- a/distributions/bedrock/run.yaml
+++ b/distributions/bedrock/run.yaml
@ -1,5 +1,4 @@
 version: '2'
-built_at: '2024-11-01T17:40:45.325529'
 image_name: local
 name: bedrock
 docker_image: null
--- a/distributions/dell-tgi/run.yaml
+++ b/distributions/dell-tgi/run.yaml
@ -1,5 +1,4 @@
 version: '2'
-built_at: '2024-10-08T17:40:45.325529'
 image_name: local
 docker_image: null
 conda_env: local
--- a/distributions/fireworks/run.yaml
+++ b/distributions/fireworks/run.yaml
@ -1,51 +0,0 @@
-version: '2'
-built_at: '2024-10-08T17:40:45.325529'
-image_name: local
-docker_image: null
-conda_env: local
-apis:
- shields
- agents
- models
- memory
- memory_banks
- inference
- safety
-providers:
-  inference:
-  - provider_id: fireworks0
-    provider_type: remote::fireworks
-    config:
-      url: https://api.fireworks.ai/inference
-      # api_key: <ENTER_YOUR_API_KEY>
-  safety:
-  safety:
-  - provider_id: meta0
-    provider_type: inline::llama-guard
-    config:
-      model: Llama-Guard-3-1B
-      excluded_categories: []
-  - provider_id: meta1
-    provider_type: inline::prompt-guard
-    config:
-      model: Prompt-Guard-86M
-  memory:
-  - provider_id: meta0
-    provider_type: inline::meta-reference
-    config: {}
-  # Uncomment to use weaviate memory provider
-  # - provider_id: weaviate0
-  #   provider_type: remote::weaviate
-  #   config: {}
-  agents:
-  - provider_id: meta0
-    provider_type: inline::meta-reference
-    config:
-      persistence_store:
-        namespace: null
-        type: sqlite
-        db_path: ~/.llama/runtime/kvstore.db
-  telemetry:
-  - provider_id: meta0
-    provider_type: inline::meta-reference
-    config: {}
--- a/distributions/fireworks/run.yaml
+++ b/distributions/fireworks/run.yaml
@ -0,0 +1 @@
+../../llama_stack/templates/fireworks/run.yaml
--- a/distributions/inline-vllm/run.yaml
+++ b/distributions/inline-vllm/run.yaml
@ -1,5 +1,4 @@
 version: '2'
-built_at: '2024-10-08T17:40:45.325529'
 image_name: local
 docker_image: null
 conda_env: local
--- a/distributions/meta-reference-gpu/run-with-safety.yaml
+++ b/distributions/meta-reference-gpu/run-with-safety.yaml
@ -0,0 +1 @@
+../../llama_stack/templates/meta-reference-gpu/run-with-safety.yaml
--- a/distributions/meta-reference-gpu/run.yaml
+++ b/distributions/meta-reference-gpu/run.yaml
@ -1,69 +0,0 @@
-version: '2'
-built_at: '2024-10-08T17:40:45.325529'
-image_name: local
-docker_image: null
-conda_env: local
-apis:
- shields
- agents
- models
- memory
- memory_banks
- inference
- safety
-providers:
-  inference:
-  - provider_id: inference0
-    provider_type: inline::meta-reference
-    config:
-      model: Llama3.2-3B-Instruct
-      quantization: null
-      torch_seed: null
-      max_seq_len: 4096
-      max_batch_size: 1
-  - provider_id: inference1
-    provider_type: inline::meta-reference
-    config:
-      model: Llama-Guard-3-1B
-      quantization: null
-      torch_seed: null
-      max_seq_len: 2048
-      max_batch_size: 1
-  safety:
-  - provider_id: meta0
-    provider_type: inline::llama-guard
-    config:
-      model: Llama-Guard-3-1B
-      excluded_categories: []
-  - provider_id: meta1
-    provider_type: inline::prompt-guard
-    config:
-      model: Prompt-Guard-86M
-# Uncomment to use prompt guard
-#      prompt_guard_shield:
-#        model: Prompt-Guard-86M
-  memory:
-  - provider_id: meta0
-    provider_type: inline::meta-reference
-    config: {}
-  # Uncomment to use pgvector
-  # - provider_id: pgvector
-  #   provider_type: remote::pgvector
-  #   config:
-  #     host: 127.0.0.1
-  #     port: 5432
-  #     db: postgres
-  #     user: postgres
-  #     password: mysecretpassword
-  agents:
-  - provider_id: meta0
-    provider_type: inline::meta-reference
-    config:
-      persistence_store:
-        namespace: null
-        type: sqlite
-        db_path: ~/.llama/runtime/agents_store.db
-  telemetry:
-  - provider_id: meta0
-    provider_type: inline::meta-reference
-    config: {}
--- a/distributions/meta-reference-gpu/run.yaml
+++ b/distributions/meta-reference-gpu/run.yaml
@ -0,0 +1 @@
+../../llama_stack/templates/meta-reference-gpu/run.yaml
--- a/distributions/meta-reference-quantized-gpu/run.yaml
+++ b/distributions/meta-reference-quantized-gpu/run.yaml
@ -1,5 +1,4 @@
 version: '2'
-built_at: '2024-10-08T17:40:45.325529'
 image_name: local
 docker_image: null
 conda_env: local
--- a/distributions/ollama-gpu/run.yaml
+++ b/distributions/ollama-gpu/run.yaml
@ -1,5 +1,4 @@
 version: '2'
-built_at: '2024-10-08T17:40:45.325529'
 image_name: local
 docker_image: null
 conda_env: local
@ -13,20 +12,15 @@ apis:
 - safety
 providers:
  inference:
-  - provider_id: ollama0
+  - provider_id: ollama
    provider_type: remote::ollama
    config:
-      url: http://127.0.0.1:14343
+      url: ${env.OLLAMA_URL:http://127.0.0.1:11434}
  safety:
  - provider_id: meta0
    provider_type: inline::llama-guard
    config:
-      model: Llama-Guard-3-1B
      excluded_categories: []
-  - provider_id: meta1
-    provider_type: inline::prompt-guard
-    config:
-      model: Prompt-Guard-86M
  memory:
  - provider_id: meta0
    provider_type: inline::meta-reference
@ -43,3 +37,10 @@ providers:
  - provider_id: meta0
    provider_type: inline::meta-reference
    config: {}
+models:
+  - model_id: ${env.INFERENCE_MODEL:Llama3.2-3B-Instruct}
+    provider_id: ollama
+  - model_id: ${env.SAFETY_MODEL:Llama-Guard-3-1B}
+    provider_id: ollama
+shields:
+  - shield_id: ${env.SAFETY_MODEL:Llama-Guard-3-1B}
--- a/distributions/ollama/compose.yaml
+++ b/distributions/ollama/compose.yaml
@ -1,30 +1,71 @@
 services:
  ollama:
    image: ollama/ollama:latest
-    network_mode: "host"
+    network_mode: ${NETWORK_MODE:-bridge}
    volumes:
-      - ollama:/root/.ollama # this solution synchronizes with the docker volume and loads the model rocket fast
+      - ~/.ollama:/root/.ollama
    ports:
      - "11434:11434"
+    environment:
+      OLLAMA_DEBUG: 1
    command: []
+    deploy:
+      resources:
+        limits:
+          memory: 8G    # Set maximum memory
+        reservations:
+          memory: 8G    # Set minimum memory reservation
+    # healthcheck:
+    #   # ugh, no CURL in ollama image
+    #   test: ["CMD", "curl", "-f", "http://ollama:11434"]
+    #   interval: 10s
+    #   timeout: 5s
+    #   retries: 5
+
+  ollama-init:
+    image: ollama/ollama:latest
+    depends_on:
+      - ollama
+        # condition: service_healthy
+    network_mode: ${NETWORK_MODE:-bridge}
+    environment:
+      - OLLAMA_HOST=ollama
+      - INFERENCE_MODEL=${INFERENCE_MODEL}
+      - SAFETY_MODEL=${SAFETY_MODEL:-}
+    volumes:
+      - ~/.ollama:/root/.ollama
+      - ./pull-models.sh:/pull-models.sh
+    entrypoint: ["/pull-models.sh"]
+
  llamastack:
    depends_on:
-    - ollama
-    image: llamastack/distribution-ollama
-    network_mode: "host"
+      ollama:
+        condition: service_started
+      ollama-init:
+        condition: service_started
+    image: ${LLAMA_STACK_IMAGE:-llamastack/distribution-ollama}
+    network_mode: ${NETWORK_MODE:-bridge}
    volumes:
      - ~/.llama:/root/.llama
      # Link to ollama run.yaml file
-      - ./run.yaml:/root/my-run.yaml
+      - ~/local/llama-stack/:/app/llama-stack-source
+      - ./run${SAFETY_MODEL:+-with-safety}.yaml:/root/my-run.yaml
    ports:
-      - "5000:5000"
-    # Hack: wait for ollama server to start before starting docker
-    entrypoint: bash -c "sleep 60; python -m llama_stack.distribution.server.server --yaml_config /root/my-run.yaml"
+      - "${LLAMA_STACK_PORT:-5001}:${LLAMA_STACK_PORT:-5001}"
+    environment:
+      - INFERENCE_MODEL=${INFERENCE_MODEL}
+      - SAFETY_MODEL=${SAFETY_MODEL:-}
+      - OLLAMA_URL=http://ollama:11434
+    entrypoint: >
+        python -m llama_stack.distribution.server.server /root/my-run.yaml \
+        --port ${LLAMA_STACK_PORT:-5001}
    deploy:
      restart_policy:
        condition: on-failure
-        delay: 3s
-        max_attempts: 5
+        delay: 10s
+        max_attempts: 3
        window: 60s
 volumes:
  ollama:
+  ollama-init:
+  llamastack:
--- a/distributions/ollama/pull-models.sh
+++ b/distributions/ollama/pull-models.sh
@ -0,0 +1,18 @@
+#!/bin/sh
+
+# Copyright (c) Meta Platforms, Inc. and affiliates.
+# All rights reserved.
+#
+# This source code is licensed under the terms described in the LICENSE file in
+# the root directory of this source tree.
+
+echo "Preloading (${INFERENCE_MODEL}, ${SAFETY_MODEL})..."
+for model in ${INFERENCE_MODEL} ${SAFETY_MODEL}; do
+  echo "Preloading $model..."
+  if ! ollama run "$model"; then
+    echo "Failed to pull and run $model"
+    exit 1
+  fi
+done
+
+echo "All models pulled successfully"
--- a/distributions/ollama/run-with-safety.yaml
+++ b/distributions/ollama/run-with-safety.yaml
@ -0,0 +1 @@
+../../llama_stack/templates/ollama/run-with-safety.yaml
--- a/distributions/ollama/run.yaml
+++ b/distributions/ollama/run.yaml
@ -1,45 +0,0 @@
-version: '2'
-built_at: '2024-10-08T17:40:45.325529'
-image_name: local
-docker_image: null
-conda_env: local
-apis:
- shields
- agents
- models
- memory
- memory_banks
- inference
- safety
-providers:
-  inference:
-  - provider_id: ollama0
-    provider_type: remote::ollama
-    config:
-      url: http://127.0.0.1:14343
-  safety:
-  - provider_id: meta0
-    provider_type: inline::llama-guard
-    config:
-      model: Llama-Guard-3-1B
-      excluded_categories: []
-  - provider_id: meta1
-    provider_type: inline::prompt-guard
-    config:
-      model: Prompt-Guard-86M
-  memory:
-  - provider_id: meta0
-    provider_type: inline::meta-reference
-    config: {}
-  agents:
-  - provider_id: meta0
-    provider_type: inline::meta-reference
-    config:
-      persistence_store:
-        namespace: null
-        type: sqlite
-        db_path: ~/.llama/runtime/kvstore.db
-  telemetry:
-  - provider_id: meta0
-    provider_type: inline::meta-reference
-    config: {}
--- a/distributions/ollama/run.yaml
+++ b/distributions/ollama/run.yaml
@ -0,0 +1 @@
+../../llama_stack/templates/ollama/run.yaml
--- a/distributions/remote-vllm/compose.yaml
+++ b/distributions/remote-vllm/compose.yaml
@ -1,33 +1,28 @@
-# NOTES:
-#
-# This Docker Compose (and the associated run.yaml) assumes you will be
-# running in the default "bridged" network mode.
-#
-# If you need "host" network mode, please uncomment
-#  - network_mode: "host"
-#
-# Similarly change "host.docker.internal" to "localhost" in the run.yaml file
-#
 services:
-  vllm-0:
+  vllm-inference:
    image: vllm/vllm-openai:latest
    volumes:
      - $HOME/.cache/huggingface:/root/.cache/huggingface
-    # network_mode: "host"
+    network_mode: ${NETWORK_MODE:-bridged}
    ports:
-       - "5100:5100"
+       - "${VLLM_INFERENCE_PORT:-5100}:${VLLM_INFERENCE_PORT:-5100}"
    devices:
      - nvidia.com/gpu=all
    environment:
-      - CUDA_VISIBLE_DEVICES=0
+      - CUDA_VISIBLE_DEVICES=${VLLM_INFERENCE_GPU:-0}
      - HUGGING_FACE_HUB_TOKEN=$HF_TOKEN
    command: >
      --gpu-memory-utilization 0.75
-      --model meta-llama/Llama-3.1-8B-Instruct
+      --model ${VLLM_INFERENCE_MODEL:-meta-llama/Llama-3.2-3B-Instruct}
      --enforce-eager
      --max-model-len 8192
      --max-num-seqs 16
-      --port 5100
+      --port ${VLLM_INFERENCE_PORT:-5100}
+    healthcheck:
+      test: ["CMD", "curl", "-f", "http://localhost:${VLLM_INFERENCE_PORT:-5100}/v1/health"]
+      interval: 30s
+      timeout: 10s
+      retries: 5
    deploy:
      resources:
        reservations:
@ -35,25 +30,34 @@ services:
          - driver: nvidia
            capabilities: [gpu]
    runtime: nvidia
-  vllm-1:
+
+  # A little trick:
+  # if VLLM_SAFETY_MODEL is set, we will create a service for the safety model
+  # otherwise, the entry will end in a hyphen which gets ignored by docker compose
+  vllm-${VLLM_SAFETY_MODEL:+safety}:
    image: vllm/vllm-openai:latest
    volumes:
      - $HOME/.cache/huggingface:/root/.cache/huggingface
-    # network_mode: "host"
+    network_mode: ${NETWORK_MODE:-bridged}
    ports:
-      - "5101:5101"
+      - "${VLLM_SAFETY_PORT:-5101}:${VLLM_SAFETY_PORT:-5101}"
    devices:
      - nvidia.com/gpu=all
    environment:
-      - CUDA_VISIBLE_DEVICES=1
+      - CUDA_VISIBLE_DEVICES=${VLLM_SAFETY_GPU:-1}
      - HUGGING_FACE_HUB_TOKEN=$HF_TOKEN
    command: >
      --gpu-memory-utilization 0.75
-      --model meta-llama/Llama-Guard-3-1B
+      --model ${VLLM_SAFETY_MODEL}
      --enforce-eager
      --max-model-len 8192
      --max-num-seqs 16
-      --port 5101
+      --port ${VLLM_SAFETY_PORT:-5101}
+    healthcheck:
+      test: ["CMD", "curl", "-f", "http://localhost:${VLLM_SAFETY_PORT:-5101}/v1/health"]
+      interval: 30s
+      timeout: 10s
+      retries: 5
    deploy:
      resources:
        reservations:
@ -63,23 +67,25 @@ services:
    runtime: nvidia
  llamastack:
    depends_on:
-    - vllm-0
-    - vllm-1
-      # image: llamastack/distribution-remote-vllm
+      - vllm-inference:
+          condition: service_healthy
+      - vllm-${VLLM_SAFETY_MODEL:+safety}:
+          condition: service_healthy
+    # image: llamastack/distribution-remote-vllm
    image: llamastack/distribution-remote-vllm:test-0.0.52rc3
    volumes:
      - ~/.llama:/root/.llama
-      - ~/local/llama-stack/distributions/remote-vllm/run.yaml:/root/llamastack-run-remote-vllm.yaml
-    # network_mode: "host"
+      - ./run${VLLM_SAFETY_MODEL:+-with-safety}.yaml:/root/llamastack-run-remote-vllm.yaml
+    network_mode: ${NETWORK_MODE:-bridged}
    environment:
-      - LLAMA_INFERENCE_VLLM_URL=${LLAMA_INFERENCE_VLLM_URL:-http://host.docker.internal:5100/v1}
-      - LLAMA_INFERENCE_MODEL=${LLAMA_INFERENCE_MODEL:-Llama3.1-8B-Instruct}
+      - VLLM_URL=http://vllm-inference:${VLLM_INFERENCE_PORT:-5100}/v1
+      - VLLM_SAFETY_URL=http://vllm-safety:${VLLM_SAFETY_PORT:-5101}/v1
+      - INFERENCE_MODEL=${INFERENCE_MODEL:-meta-llama/Llama-3.2-3B-Instruct}
      - MAX_TOKENS=${MAX_TOKENS:-4096}
      - SQLITE_STORE_DIR=${SQLITE_STORE_DIR:-$HOME/.llama/distributions/remote-vllm}
-      - LLAMA_SAFETY_VLLM_URL=${LLAMA_SAFETY_VLLM_URL:-http://host.docker.internal:5101/v1}
-      - LLAMA_SAFETY_MODEL=${LLAMA_SAFETY_MODEL:-Llama-Guard-3-1B}
+      - SAFETY_MODEL=${SAFETY_MODEL:-meta-llama/Llama-Guard-3-1B}
    ports:
-      - "5001:5001"
+      - "${LLAMASTACK_PORT:-5001}:${LLAMASTACK_PORT:-5001}"
    # Hack: wait for vLLM server to start before starting docker
    entrypoint: bash -c "sleep 60; python -m llama_stack.distribution.server.server --yaml_config /root/llamastack-run-remote-vllm.yaml --port 5001"
    deploy:
@ -89,6 +95,6 @@ services:
        max_attempts: 5
        window: 60s
 volumes:
-  vllm-0:
-  vllm-1:
+  vllm-inference:
+  vllm-safety:
  llamastack:
--- a/distributions/remote-vllm/run-with-safety.yaml
+++ b/distributions/remote-vllm/run-with-safety.yaml
@ -0,0 +1 @@
+../../llama_stack/templates/remote-vllm/run-with-safety.yaml
--- a/distributions/remote-vllm/run.yaml
+++ b/distributions/remote-vllm/run.yaml
@ -1,68 +0,0 @@
-version: '2'
-built_at: '2024-11-11T20:09:45.988375'
-image_name: remote-vllm
-docker_image: remote-vllm
-conda_env: null
-apis:
- inference
- memory
- safety
- agents
- telemetry
-providers:
-  inference:
-  # serves main inference model
-  - provider_id: vllm-0
-    provider_type: remote::vllm
-    config:
-      # NOTE: replace with "localhost" if you are running in "host" network mode
-      url: ${env.LLAMA_INFERENCE_VLLM_URL:http://host.docker.internal:5100/v1}
-      max_tokens: ${env.MAX_TOKENS:4096}
-      api_token: fake
-  # serves safety llama_guard model
-  - provider_id: vllm-1
-    provider_type: remote::vllm
-    config:
-      # NOTE: replace with "localhost" if you are running in "host" network mode
-      url: ${env.LLAMA_SAFETY_VLLM_URL:http://host.docker.internal:5101/v1}
-      max_tokens: ${env.MAX_TOKENS:4096}
-      api_token: fake
-  memory:
-  - provider_id: faiss-0
-    provider_type: inline::faiss
-    config:
-      kvstore:
-        namespace: null
-        type: sqlite
-        db_path: "${env.SQLITE_STORE_DIR:/home/ashwin/.llama/distributions/remote-vllm}/faiss_store.db"
-  safety:
-  - provider_id: llama-guard
-    provider_type: inline::llama-guard
-    config: {}
-  memory:
-  - provider_id: meta0
-    provider_type: inline::faiss
-    config: {}
-  agents:
-  - provider_id: meta0
-    provider_type: inline::meta-reference
-    config:
-      persistence_store:
-        namespace: null
-        type: sqlite
-        db_path: "${env.SQLITE_STORE_DIR:/home/ashwin/.llama/distributions/remote-vllm}/agents_store.db"
-  telemetry:
-  - provider_id: meta0
-    provider_type: inline::meta-reference
-    config: {}
-metadata_store:
-  namespace: null
-  type: sqlite
-  db_path: "${env.SQLITE_STORE_DIR:/home/ashwin/.llama/distributions/remote-vllm}/registry.db"
-models:
-  - model_id: ${env.LLAMA_INFERENCE_MODEL:Llama3.1-8B-Instruct}
-    provider_id: vllm-0
-  - model_id: ${env.LLAMA_SAFETY_MODEL:Llama-Guard-3-1B}
-    provider_id: vllm-1
-shields:
-  - shield_id: ${env.LLAMA_SAFETY_MODEL:Llama-Guard-3-1B}
--- a/distributions/remote-vllm/run.yaml
+++ b/distributions/remote-vllm/run.yaml
@ -0,0 +1 @@
+../../llama_stack/templates/remote-vllm/run.yaml
--- a/distributions/tgi/compose.yaml
+++ b/distributions/tgi/compose.yaml
@ -1,51 +1,89 @@
 services:
-  text-generation-inference:
+  tgi-inference:
    image: ghcr.io/huggingface/text-generation-inference:latest
-    network_mode: "host"
    volumes:
      - $HOME/.cache/huggingface:/data
+    network_mode: ${NETWORK_MODE:-bridged}
    ports:
-      - "5009:5009"
+       - "${TGI_INFERENCE_PORT:-8080}:${TGI_INFERENCE_PORT:-8080}"
    devices:
      - nvidia.com/gpu=all
    environment:
-      - CUDA_VISIBLE_DEVICES=0
+      - CUDA_VISIBLE_DEVICES=${TGI_INFERENCE_GPU:-0}
+      - HF_TOKEN=$HF_TOKEN
      - HF_HOME=/data
      - HF_DATASETS_CACHE=/data
      - HF_MODULES_CACHE=/data
      - HF_HUB_CACHE=/data
-    command: ["--dtype", "bfloat16", "--usage-stats", "on", "--sharded", "false", "--model-id", "meta-llama/Llama-3.1-8B-Instruct", "--port", "5009", "--cuda-memory-fraction", "0.3"]
+    command: >
+      --dtype bfloat16
+      --usage-stats off
+      --sharded false
+      --model-id ${TGI_INFERENCE_MODEL:-meta-llama/Llama-3.2-3B-Instruct}
+      --port ${TGI_INFERENCE_PORT:-8080}
+      --cuda-memory-fraction 0.75
+    healthcheck:
+      test: ["CMD", "curl", "-f", "http://tgi-inference:${TGI_INFERENCE_PORT:-8080}/health"]
+      interval: 5s
+      timeout: 5s
+      retries: 30
    deploy:
      resources:
        reservations:
          devices:
          - driver: nvidia
-            # that's the closest analogue to --gpus; provide
-            # an integer amount of devices or 'all'
-            count: 1
-            # Devices are reserved using a list of capabilities, making
-            # capabilities the only required field. A device MUST
-            # satisfy all the requested capabilities for a successful
-            # reservation.
            capabilities: [gpu]
    runtime: nvidia
+
+  tgi-${TGI_SAFETY_MODEL:+safety}:
+    image: ghcr.io/huggingface/text-generation-inference:latest
+    volumes:
+      - $HOME/.cache/huggingface:/data
+    network_mode: ${NETWORK_MODE:-bridged}
+    ports:
+       - "${TGI_SAFETY_PORT:-8081}:${TGI_SAFETY_PORT:-8081}"
+    devices:
+      - nvidia.com/gpu=all
+    environment:
+      - CUDA_VISIBLE_DEVICES=${TGI_SAFETY_GPU:-1}
+      - HF_TOKEN=$HF_TOKEN
+      - HF_HOME=/data
+      - HF_DATASETS_CACHE=/data
+      - HF_MODULES_CACHE=/data
+      - HF_HUB_CACHE=/data
+    command: >
+      --dtype bfloat16
+      --usage-stats off
+      --sharded false
+      --model-id ${TGI_SAFETY_MODEL:-meta-llama/Llama-Guard-3-1B}
+      --port ${TGI_SAFETY_PORT:-8081}
+      --cuda-memory-fraction 0.75
    healthcheck:
-      test: ["CMD", "curl", "-f", "http://text-generation-inference:5009/health"]
+      test: ["CMD", "curl", "-f", "http://tgi-safety:${TGI_SAFETY_PORT:-8081}/health"]
      interval: 5s
      timeout: 5s
      retries: 30
+    deploy:
+      resources:
+        reservations:
+          devices:
+          - driver: nvidia
+            capabilities: [gpu]
+    runtime: nvidia
+
  llamastack:
    depends_on:
-      text-generation-inference:
+      tgi-inference:
        condition: service_healthy
-    image: llamastack/distribution-tgi
-    network_mode: "host"
+      tgi-${TGI_SAFETY_MODEL:+safety}:
+        condition: service_healthy
+    image: llamastack/distribution-tgi:test-0.0.52rc3
+    network_mode: ${NETWORK_MODE:-bridged}
    volumes:
      - ~/.llama:/root/.llama
-      # Link to TGI run.yaml file
-      - ./run.yaml:/root/my-run.yaml
+      - ./run${TGI_SAFETY_MODEL:+-with-safety}.yaml:/root/my-run.yaml
    ports:
-      - "5000:5000"
+      - "${LLAMA_STACK_PORT:-5001}:${LLAMA_STACK_PORT:-5001}"
    # Hack: wait for TGI server to start before starting docker
    entrypoint: bash -c "sleep 60; python -m llama_stack.distribution.server.server --yaml_config /root/my-run.yaml"
    restart_policy:
@ -53,3 +91,13 @@ services:
      delay: 3s
      max_attempts: 5
      window: 60s
+    environment:
+      - TGI_URL=http://tgi-inference:${TGI_INFERENCE_PORT:-8080}
+      - SAFETY_TGI_URL=http://tgi-safety:${TGI_SAFETY_PORT:-8081}
+      - INFERENCE_MODEL=${INFERENCE_MODEL:-meta-llama/Llama-3.2-3B-Instruct}
+      - SAFETY_MODEL=${SAFETY_MODEL:-meta-llama/Llama-Guard-3-1B}
+
+volumes:
+  tgi-inference:
+  tgi-safety:
+  llamastack:
--- a/distributions/tgi/run-with-safety.yaml
+++ b/distributions/tgi/run-with-safety.yaml
@ -0,0 +1 @@
+../../llama_stack/templates/tgi/run-with-safety.yaml
--- a/distributions/tgi/run.yaml
+++ b/distributions/tgi/run.yaml
@ -1,45 +0,0 @@
-version: '2'
-built_at: '2024-10-08T17:40:45.325529'
-image_name: local
-docker_image: null
-conda_env: local
-apis:
- shields
- agents
- models
- memory
- memory_banks
- inference
- safety
-providers:
-  inference:
-  - provider_id: tgi0
-    provider_type: remote::tgi
-    config:
-      url: http://127.0.0.1:5009
-  safety:
-  - provider_id: meta0
-    provider_type: inline::llama-guard
-    config:
-      model: Llama-Guard-3-1B
-      excluded_categories: []
-  - provider_id: meta1
-    provider_type: inline::prompt-guard
-    config:
-      model: Prompt-Guard-86M
-  memory:
-  - provider_id: meta0
-    provider_type: inline::meta-reference
-    config: {}
-  agents:
-  - provider_id: meta0
-    provider_type: inline::meta-reference
-    config:
-      persistence_store:
-        namespace: null
-        type: sqlite
-        db_path: ~/.llama/runtime/kvstore.db
-  telemetry:
-  - provider_id: meta0
-    provider_type: inline::meta-reference
-    config: {}
--- a/distributions/tgi/run.yaml
+++ b/distributions/tgi/run.yaml
@ -0,0 +1 @@
+../../llama_stack/templates/tgi/run.yaml
--- a/distributions/together/run.yaml
+++ b/distributions/together/run.yaml
@ -1,46 +0,0 @@
-version: '2'
-built_at: '2024-10-08T17:40:45.325529'
-image_name: local
-docker_image: null
-conda_env: local
-apis:
- shields
- agents
- models
- memory
- memory_banks
- inference
- safety
-providers:
-  inference:
-  - provider_id: together0
-    provider_type: remote::together
-    config:
-      url: https://api.together.xyz/v1
-      # api_key: <ENTER_YOUR_API_KEY>
-  safety:
-  - provider_id: meta0
-    provider_type: inline::llama-guard
-    config:
-      model: Llama-Guard-3-1B
-      excluded_categories: []
-  - provider_id: meta1
-    provider_type: inline::prompt-guard
-    config:
-      model: Prompt-Guard-86M
-  memory:
-  - provider_id: meta0
-    provider_type: remote::weaviate
-    config: {}
-  agents:
-  - provider_id: meta0
-    provider_type: inline::meta-reference
-    config:
-      persistence_store:
-        namespace: null
-        type: sqlite
-        db_path: ~/.llama/runtime/kvstore.db
-  telemetry:
-  - provider_id: meta0
-    provider_type: inline::meta-reference
-    config: {}
--- a/distributions/together/run.yaml
+++ b/distributions/together/run.yaml
@ -0,0 +1 @@
+../../llama_stack/templates/together/run.yaml
--- a/docs/openapi_generator/generate.py
+++ b/docs/openapi_generator/generate.py
@ -31,7 +31,10 @@ from .strong_typing.schema import json_schema_type

 schema_utils.json_schema_type = json_schema_type

-from llama_stack.distribution.stack import LlamaStack
+# this line needs to be here to ensure json_schema_type has been altered before
+# the imports use the annotation
+from llama_stack.apis.version import LLAMA_STACK_API_VERSION  # noqa: E402
+from llama_stack.distribution.stack import LlamaStack  # noqa: E402


 def main(output_dir: str):
@ -50,7 +53,7 @@ def main(output_dir: str):
            server=Server(url="http://any-hosted-llama-stack.com"),
            info=Info(
                title="[DRAFT] Llama Stack Specification",
-                version="0.0.1",
+                version=LLAMA_STACK_API_VERSION,
                description="""This is the specification of the llama stack that provides
                a set of endpoints and their corresponding interfaces that are tailored to
                best leverage Llama Models. The specification is still in draft and subject to change.
--- a/docs/openapi_generator/pyopenapi/generator.py
+++ b/docs/openapi_generator/pyopenapi/generator.py
@ -202,7 +202,9 @@ class ContentBuilder:
    ) -> MediaType:
        schema = self.schema_builder.classdef_to_ref(item_type)
        if self.schema_transformer:
-            schema_transformer: Callable[[SchemaOrRef], SchemaOrRef] = self.schema_transformer  # type: ignore
+            schema_transformer: Callable[[SchemaOrRef], SchemaOrRef] = (
+                self.schema_transformer
+            )
            schema = schema_transformer(schema)

        if not examples:
@ -630,6 +632,7 @@ class Generator:
                raise NotImplementedError(f"unknown HTTP method: {op.http_method}")

            route = op.get_route()
+            print(f"route: {route}")
            if route in paths:
                paths[route].update(pathItem)
            else:
--- a/docs/openapi_generator/pyopenapi/operations.py
+++ b/docs/openapi_generator/pyopenapi/operations.py
@ -12,6 +12,8 @@ import uuid
 from dataclasses import dataclass
 from typing import Any, Callable, Dict, Iterable, Iterator, List, Optional, Tuple, Union

+from llama_stack.apis.version import LLAMA_STACK_API_VERSION
+
 from termcolor import colored

 from ..strong_typing.inspection import (
@ -111,9 +113,12 @@ class EndpointOperation:

    def get_route(self) -> str:
        if self.route is not None:
-            return self.route
+            assert (
+                "_" not in self.route
+            ), f"route should not contain underscores: {self.route}"
+            return "/".join(["", LLAMA_STACK_API_VERSION, self.route.lstrip("/")])

-        route_parts = ["", self.name]
+        route_parts = ["", LLAMA_STACK_API_VERSION, self.name]
        for param_name, _ in self.path_params:
            route_parts.append("{" + param_name + "}")
        return "/".join(route_parts)
--- a/docs/resources/llama-stack-spec.html
+++ b/docs/resources/llama-stack-spec.html
@ -20,8 +20,8 @@
    "openapi": "3.1.0",
    "info": {
        "title": "[DRAFT] Llama Stack Specification",
-        "version": "0.0.1",
-        "description": "This is the specification of the llama stack that provides\n                a set of endpoints and their corresponding interfaces that are tailored to\n                best leverage Llama Models. The specification is still in draft and subject to change.\n                Generated at 2024-11-14 17:04:24.301559"
+        "version": "alpha",
+        "description": "This is the specification of the llama stack that provides\n                a set of endpoints and their corresponding interfaces that are tailored to\n                best leverage Llama Models. The specification is still in draft and subject to change.\n                Generated at 2024-11-18 23:37:24.867143"
    },
    "servers": [
        {
@ -29,7 +29,7 @@
        }
    ],
    "paths": {
-        "/batch_inference/chat_completion": {
+        "/alpha/batch-inference/chat-completion": {
            "post": {
                "responses": {
                    "200": {
@ -69,7 +69,7 @@
                }
            }
        },
-        "/batch_inference/completion": {
+        "/alpha/batch-inference/completion": {
            "post": {
                "responses": {
                    "200": {
@ -109,7 +109,7 @@
                }
            }
        },
-        "/post_training/job/cancel": {
+        "/alpha/post-training/job/cancel": {
            "post": {
                "responses": {
                    "200": {
@ -142,7 +142,7 @@
                }
            }
        },
-        "/inference/chat_completion": {
+        "/alpha/inference/chat-completion": {
            "post": {
                "responses": {
                    "200": {
@ -189,7 +189,7 @@
                }
            }
        },
-        "/inference/completion": {
+        "/alpha/inference/completion": {
            "post": {
                "responses": {
                    "200": {
@ -236,7 +236,7 @@
                }
            }
        },
-        "/agents/create": {
+        "/alpha/agents/create": {
            "post": {
                "responses": {
                    "200": {
@ -276,7 +276,7 @@
                }
            }
        },
-        "/agents/session/create": {
+        "/alpha/agents/session/create": {
            "post": {
                "responses": {
                    "200": {
@ -316,7 +316,7 @@
                }
            }
        },
-        "/agents/turn/create": {
+        "/alpha/agents/turn/create": {
            "post": {
                "responses": {
                    "200": {
@ -363,7 +363,7 @@
                }
            }
        },
-        "/agents/delete": {
+        "/alpha/agents/delete": {
            "post": {
                "responses": {
                    "200": {
@ -396,7 +396,7 @@
                }
            }
        },
-        "/agents/session/delete": {
+        "/alpha/agents/session/delete": {
            "post": {
                "responses": {
                    "200": {
@ -429,7 +429,7 @@
                }
            }
        },
-        "/inference/embeddings": {
+        "/alpha/inference/embeddings": {
            "post": {
                "responses": {
                    "200": {
@ -469,7 +469,7 @@
                }
            }
        },
-        "/eval/evaluate_rows": {
+        "/alpha/eval/evaluate-rows": {
            "post": {
                "responses": {
                    "200": {
@ -509,7 +509,7 @@
                }
            }
        },
-        "/agents/session/get": {
+        "/alpha/agents/session/get": {
            "post": {
                "responses": {
                    "200": {
@ -565,7 +565,7 @@
                }
            }
        },
-        "/agents/step/get": {
+        "/alpha/agents/step/get": {
            "get": {
                "responses": {
                    "200": {
@ -627,7 +627,7 @@
                ]
            }
        },
-        "/agents/turn/get": {
+        "/alpha/agents/turn/get": {
            "get": {
                "responses": {
                    "200": {
@ -681,7 +681,7 @@
                ]
            }
        },
-        "/datasets/get": {
+        "/alpha/datasets/get": {
            "get": {
                "responses": {
                    "200": {
@ -726,7 +726,7 @@
                ]
            }
        },
-        "/eval_tasks/get": {
+        "/alpha/eval-tasks/get": {
            "get": {
                "responses": {
                    "200": {
@ -771,7 +771,7 @@
                ]
            }
        },
-        "/memory_banks/get": {
+        "/alpha/memory-banks/get": {
            "get": {
                "responses": {
                    "200": {
@ -829,7 +829,7 @@
                ]
            }
        },
-        "/models/get": {
+        "/alpha/models/get": {
            "get": {
                "responses": {
                    "200": {
@ -874,7 +874,7 @@
                ]
            }
        },
-        "/datasetio/get_rows_paginated": {
+        "/alpha/datasetio/get-rows-paginated": {
            "get": {
                "responses": {
                    "200": {
@ -936,7 +936,7 @@
                ]
            }
        },
-        "/scoring_functions/get": {
+        "/alpha/scoring-functions/get": {
            "get": {
                "responses": {
                    "200": {
@ -981,7 +981,7 @@
                ]
            }
        },
-        "/shields/get": {
+        "/alpha/shields/get": {
            "get": {
                "responses": {
                    "200": {
@ -1026,7 +1026,7 @@
                ]
            }
        },
-        "/telemetry/get_trace": {
+        "/alpha/telemetry/get-trace": {
            "get": {
                "responses": {
                    "200": {
@ -1064,7 +1064,7 @@
                ]
            }
        },
-        "/post_training/job/artifacts": {
+        "/alpha/post-training/job/artifacts": {
            "get": {
                "responses": {
                    "200": {
@ -1102,7 +1102,7 @@
                ]
            }
        },
-        "/post_training/job/logs": {
+        "/alpha/post-training/job/logs": {
            "get": {
                "responses": {
                    "200": {
@ -1140,7 +1140,7 @@
                ]
            }
        },
-        "/post_training/job/status": {
+        "/alpha/post-training/job/status": {
            "get": {
                "responses": {
                    "200": {
@ -1178,7 +1178,7 @@
                ]
            }
        },
-        "/post_training/jobs": {
+        "/alpha/post-training/jobs": {
            "get": {
                "responses": {
                    "200": {
@ -1208,7 +1208,7 @@
                ]
            }
        },
-        "/health": {
+        "/alpha/health": {
            "get": {
                "responses": {
                    "200": {
@ -1238,7 +1238,7 @@
                ]
            }
        },
-        "/memory/insert": {
+        "/alpha/memory/insert": {
            "post": {
                "responses": {
                    "200": {
@ -1271,7 +1271,7 @@
                }
            }
        },
-        "/eval/job/cancel": {
+        "/alpha/eval/job/cancel": {
            "post": {
                "responses": {
                    "200": {
@ -1304,7 +1304,7 @@
                }
            }
        },
-        "/eval/job/result": {
+        "/alpha/eval/job/result": {
            "get": {
                "responses": {
                    "200": {
@ -1350,7 +1350,7 @@
                ]
            }
        },
-        "/eval/job/status": {
+        "/alpha/eval/job/status": {
            "get": {
                "responses": {
                    "200": {
@ -1403,7 +1403,7 @@
                ]
            }
        },
-        "/datasets/list": {
+        "/alpha/datasets/list": {
            "get": {
                "responses": {
                    "200": {
@ -1433,7 +1433,7 @@
                ]
            }
        },
-        "/eval_tasks/list": {
+        "/alpha/eval-tasks/list": {
            "get": {
                "responses": {
                    "200": {
@ -1463,7 +1463,7 @@
                ]
            }
        },
-        "/memory_banks/list": {
+        "/alpha/memory-banks/list": {
            "get": {
                "responses": {
                    "200": {
@ -1506,7 +1506,7 @@
                ]
            }
        },
-        "/models/list": {
+        "/alpha/models/list": {
            "get": {
                "responses": {
                    "200": {
@ -1536,7 +1536,7 @@
                ]
            }
        },
-        "/providers/list": {
+        "/alpha/providers/list": {
            "get": {
                "responses": {
                    "200": {
@ -1569,7 +1569,7 @@
                ]
            }
        },
-        "/routes/list": {
+        "/alpha/routes/list": {
            "get": {
                "responses": {
                    "200": {
@ -1605,7 +1605,7 @@
                ]
            }
        },
-        "/scoring_functions/list": {
+        "/alpha/scoring-functions/list": {
            "get": {
                "responses": {
                    "200": {
@ -1635,7 +1635,7 @@
                ]
            }
        },
-        "/shields/list": {
+        "/alpha/shields/list": {
            "get": {
                "responses": {
                    "200": {
@ -1665,7 +1665,7 @@
                ]
            }
        },
-        "/telemetry/log_event": {
+        "/alpha/telemetry/log-event": {
            "post": {
                "responses": {
                    "200": {
@ -1698,7 +1698,7 @@
                }
            }
        },
-        "/post_training/preference_optimize": {
+        "/alpha/post-training/preference-optimize": {
            "post": {
                "responses": {
                    "200": {
@ -1738,7 +1738,7 @@
                }
            }
        },
-        "/memory/query": {
+        "/alpha/memory/query": {
            "post": {
                "responses": {
                    "200": {
@ -1778,7 +1778,7 @@
                }
            }
        },
-        "/datasets/register": {
+        "/alpha/datasets/register": {
            "post": {
                "responses": {
                    "200": {
@ -1811,7 +1811,7 @@
                }
            }
        },
-        "/eval_tasks/register": {
+        "/alpha/eval-tasks/register": {
            "post": {
                "responses": {
                    "200": {
@ -1844,7 +1844,7 @@
                }
            }
        },
-        "/memory_banks/register": {
+        "/alpha/memory-banks/register": {
            "post": {
                "responses": {},
                "tags": [
@ -1873,7 +1873,7 @@
                }
            }
        },
-        "/models/register": {
+        "/alpha/models/register": {
            "post": {
                "responses": {
                    "200": {
@ -1913,7 +1913,7 @@
                }
            }
        },
-        "/scoring_functions/register": {
+        "/alpha/scoring-functions/register": {
            "post": {
                "responses": {
                    "200": {
@ -1946,7 +1946,7 @@
                }
            }
        },
-        "/shields/register": {
+        "/alpha/shields/register": {
            "post": {
                "responses": {
                    "200": {
@ -1986,7 +1986,7 @@
                }
            }
        },
-        "/eval/run_eval": {
+        "/alpha/eval/run-eval": {
            "post": {
                "responses": {
                    "200": {
@ -2026,7 +2026,7 @@
                }
            }
        },
-        "/safety/run_shield": {
+        "/alpha/safety/run-shield": {
            "post": {
                "responses": {
                    "200": {
@ -2066,7 +2066,7 @@
                }
            }
        },
-        "/scoring/score": {
+        "/alpha/scoring/score": {
            "post": {
                "responses": {
                    "200": {
@ -2106,7 +2106,7 @@
                }
            }
        },
-        "/scoring/score_batch": {
+        "/alpha/scoring/score-batch": {
            "post": {
                "responses": {
                    "200": {
@ -2146,7 +2146,7 @@
                }
            }
        },
-        "/post_training/supervised_fine_tune": {
+        "/alpha/post-training/supervised-fine-tune": {
            "post": {
                "responses": {
                    "200": {
@ -2186,7 +2186,7 @@
                }
            }
        },
-        "/synthetic_data_generation/generate": {
+        "/alpha/synthetic-data-generation/generate": {
            "post": {
                "responses": {
                    "200": {
@ -2226,7 +2226,7 @@
                }
            }
        },
-        "/memory_banks/unregister": {
+        "/alpha/memory-banks/unregister": {
            "post": {
                "responses": {
                    "200": {
@ -2259,7 +2259,7 @@
                }
            }
        },
-        "/models/unregister": {
+        "/alpha/models/unregister": {
            "post": {
                "responses": {
                    "200": {
--- a/docs/resources/llama-stack-spec.yaml
+++ b/docs/resources/llama-stack-spec.yaml
@ -3400,13 +3400,13 @@ info:
  description: "This is the specification of the llama stack that provides\n     \
    \           a set of endpoints and their corresponding interfaces that are tailored\
    \ to\n                best leverage Llama Models. The specification is still in\
-    \ draft and subject to change.\n                Generated at 2024-11-14 17:04:24.301559"
+    \ draft and subject to change.\n                Generated at 2024-11-18 23:37:24.867143"
  title: '[DRAFT] Llama Stack Specification'
-  version: 0.0.1
+  version: alpha
 jsonSchemaDialect: https://json-schema.org/draft/2020-12/schema
 openapi: 3.1.0
 paths:
-  /agents/create:
+  /alpha/agents/create:
    post:
      parameters:
      - description: JSON-encoded provider data which will be made available to the
@ -3431,7 +3431,7 @@ paths:
          description: OK
      tags:
      - Agents
-  /agents/delete:
+  /alpha/agents/delete:
    post:
      parameters:
      - description: JSON-encoded provider data which will be made available to the
@ -3452,7 +3452,7 @@ paths:
          description: OK
      tags:
      - Agents
-  /agents/session/create:
+  /alpha/agents/session/create:
    post:
      parameters:
      - description: JSON-encoded provider data which will be made available to the
@ -3477,7 +3477,7 @@ paths:
          description: OK
      tags:
      - Agents
-  /agents/session/delete:
+  /alpha/agents/session/delete:
    post:
      parameters:
      - description: JSON-encoded provider data which will be made available to the
@ -3498,7 +3498,7 @@ paths:
          description: OK
      tags:
      - Agents
-  /agents/session/get:
+  /alpha/agents/session/get:
    post:
      parameters:
      - in: query
@ -3533,7 +3533,7 @@ paths:
          description: OK
      tags:
      - Agents
-  /agents/step/get:
+  /alpha/agents/step/get:
    get:
      parameters:
      - in: query
@ -3572,7 +3572,7 @@ paths:
          description: OK
      tags:
      - Agents
-  /agents/turn/create:
+  /alpha/agents/turn/create:
    post:
      parameters:
      - description: JSON-encoded provider data which will be made available to the
@ -3600,7 +3600,7 @@ paths:
            streamed agent turn completion response.
      tags:
      - Agents
-  /agents/turn/get:
+  /alpha/agents/turn/get:
    get:
      parameters:
      - in: query
@ -3634,7 +3634,7 @@ paths:
          description: OK
      tags:
      - Agents
-  /batch_inference/chat_completion:
+  /alpha/batch-inference/chat-completion:
    post:
      parameters:
      - description: JSON-encoded provider data which will be made available to the
@ -3659,7 +3659,7 @@ paths:
          description: OK
      tags:
      - BatchInference
-  /batch_inference/completion:
+  /alpha/batch-inference/completion:
    post:
      parameters:
      - description: JSON-encoded provider data which will be made available to the
@ -3684,7 +3684,7 @@ paths:
          description: OK
      tags:
      - BatchInference
-  /datasetio/get_rows_paginated:
+  /alpha/datasetio/get-rows-paginated:
    get:
      parameters:
      - in: query
@ -3723,7 +3723,7 @@ paths:
          description: OK
      tags:
      - DatasetIO
-  /datasets/get:
+  /alpha/datasets/get:
    get:
      parameters:
      - in: query
@ -3749,7 +3749,7 @@ paths:
          description: OK
      tags:
      - Datasets
-  /datasets/list:
+  /alpha/datasets/list:
    get:
      parameters:
      - description: JSON-encoded provider data which will be made available to the
@ -3768,7 +3768,7 @@ paths:
          description: OK
      tags:
      - Datasets
-  /datasets/register:
+  /alpha/datasets/register:
    post:
      parameters:
      - description: JSON-encoded provider data which will be made available to the
@ -3789,7 +3789,73 @@ paths:
          description: OK
      tags:
      - Datasets
-  /eval/evaluate_rows:
+  /alpha/eval-tasks/get:
+    get:
+      parameters:
+      - in: query
+        name: name
+        required: true
+        schema:
+          type: string
+      - description: JSON-encoded provider data which will be made available to the
+          adapter servicing the API
+        in: header
+        name: X-LlamaStack-ProviderData
+        required: false
+        schema:
+          type: string
+      responses:
+        '200':
+          content:
+            application/json:
+              schema:
+                oneOf:
+                - $ref: '#/components/schemas/EvalTask'
+                - type: 'null'
+          description: OK
+      tags:
+      - EvalTasks
+  /alpha/eval-tasks/list:
+    get:
+      parameters:
+      - description: JSON-encoded provider data which will be made available to the
+          adapter servicing the API
+        in: header
+        name: X-LlamaStack-ProviderData
+        required: false
+        schema:
+          type: string
+      responses:
+        '200':
+          content:
+            application/jsonl:
+              schema:
+                $ref: '#/components/schemas/EvalTask'
+          description: OK
+      tags:
+      - EvalTasks
+  /alpha/eval-tasks/register:
+    post:
+      parameters:
+      - description: JSON-encoded provider data which will be made available to the
+          adapter servicing the API
+        in: header
+        name: X-LlamaStack-ProviderData
+        required: false
+        schema:
+          type: string
+      requestBody:
+        content:
+          application/json:
+            schema:
+              $ref: '#/components/schemas/RegisterEvalTaskRequest'
+        required: true
+      responses:
+        '200':
+          description: OK
+      tags:
+      - EvalTasks
+  /alpha/eval/evaluate-rows:
    post:
      parameters:
      - description: JSON-encoded provider data which will be made available to the
@ -3814,7 +3880,7 @@ paths:
          description: OK
      tags:
      - Eval
-  /eval/job/cancel:
+  /alpha/eval/job/cancel:
    post:
      parameters:
      - description: JSON-encoded provider data which will be made available to the
@ -3835,7 +3901,7 @@ paths:
          description: OK
      tags:
      - Eval
-  /eval/job/result:
+  /alpha/eval/job/result:
    get:
      parameters:
      - in: query
@ -3864,7 +3930,7 @@ paths:
          description: OK
      tags:
      - Eval
-  /eval/job/status:
+  /alpha/eval/job/status:
    get:
      parameters:
      - in: query
@ -3895,7 +3961,7 @@ paths:
          description: OK
      tags:
      - Eval
-  /eval/run_eval:
+  /alpha/eval/run-eval:
    post:
      parameters:
      - description: JSON-encoded provider data which will be made available to the
@ -3920,73 +3986,7 @@ paths:
          description: OK
      tags:
      - Eval
-  /eval_tasks/get:
-    get:
-      parameters:
-      - in: query
-        name: name
-        required: true
-        schema:
-          type: string
-      - description: JSON-encoded provider data which will be made available to the
-          adapter servicing the API
-        in: header
-        name: X-LlamaStack-ProviderData
-        required: false
-        schema:
-          type: string
-      responses:
-        '200':
-          content:
-            application/json:
-              schema:
-                oneOf:
-                - $ref: '#/components/schemas/EvalTask'
-                - type: 'null'
-          description: OK
-      tags:
-      - EvalTasks
-  /eval_tasks/list:
-    get:
-      parameters:
-      - description: JSON-encoded provider data which will be made available to the
-          adapter servicing the API
-        in: header
-        name: X-LlamaStack-ProviderData
-        required: false
-        schema:
-          type: string
-      responses:
-        '200':
-          content:
-            application/jsonl:
-              schema:
-                $ref: '#/components/schemas/EvalTask'
-          description: OK
-      tags:
-      - EvalTasks
-  /eval_tasks/register:
-    post:
-      parameters:
-      - description: JSON-encoded provider data which will be made available to the
-          adapter servicing the API
-        in: header
-        name: X-LlamaStack-ProviderData
-        required: false
-        schema:
-          type: string
-      requestBody:
-        content:
-          application/json:
-            schema:
-              $ref: '#/components/schemas/RegisterEvalTaskRequest'
-        required: true
-      responses:
-        '200':
-          description: OK
-      tags:
-      - EvalTasks
-  /health:
+  /alpha/health:
    get:
      parameters:
      - description: JSON-encoded provider data which will be made available to the
@ -4005,7 +4005,7 @@ paths:
          description: OK
      tags:
      - Inspect
-  /inference/chat_completion:
+  /alpha/inference/chat-completion:
    post:
      parameters:
      - description: JSON-encoded provider data which will be made available to the
@ -4032,7 +4032,7 @@ paths:
          description: Chat completion response. **OR** SSE-stream of these events.
      tags:
      - Inference
-  /inference/completion:
+  /alpha/inference/completion:
    post:
      parameters:
      - description: JSON-encoded provider data which will be made available to the
@ -4059,7 +4059,7 @@ paths:
          description: Completion response. **OR** streamed completion response.
      tags:
      - Inference
-  /inference/embeddings:
+  /alpha/inference/embeddings:
    post:
      parameters:
      - description: JSON-encoded provider data which will be made available to the
@ -4084,53 +4084,7 @@ paths:
          description: OK
      tags:
      - Inference
-  /memory/insert:
-    post:
-      parameters:
-      - description: JSON-encoded provider data which will be made available to the
-          adapter servicing the API
-        in: header
-        name: X-LlamaStack-ProviderData
-        required: false
-        schema:
-          type: string
-      requestBody:
-        content:
-          application/json:
-            schema:
-              $ref: '#/components/schemas/InsertDocumentsRequest'
-        required: true
-      responses:
-        '200':
-          description: OK
-      tags:
-      - Memory
-  /memory/query:
-    post:
-      parameters:
-      - description: JSON-encoded provider data which will be made available to the
-          adapter servicing the API
-        in: header
-        name: X-LlamaStack-ProviderData
-        required: false
-        schema:
-          type: string
-      requestBody:
-        content:
-          application/json:
-            schema:
-              $ref: '#/components/schemas/QueryDocumentsRequest'
-        required: true
-      responses:
-        '200':
-          content:
-            application/json:
-              schema:
-                $ref: '#/components/schemas/QueryDocumentsResponse'
-          description: OK
-      tags:
-      - Memory
-  /memory_banks/get:
+  /alpha/memory-banks/get:
    get:
      parameters:
      - in: query
@ -4160,7 +4114,7 @@ paths:
          description: OK
      tags:
      - MemoryBanks
-  /memory_banks/list:
+  /alpha/memory-banks/list:
    get:
      parameters:
      - description: JSON-encoded provider data which will be made available to the
@ -4183,7 +4137,7 @@ paths:
          description: OK
      tags:
      - MemoryBanks
-  /memory_banks/register:
+  /alpha/memory-banks/register:
    post:
      parameters:
      - description: JSON-encoded provider data which will be made available to the
@ -4202,7 +4156,7 @@ paths:
      responses: {}
      tags:
      - MemoryBanks
-  /memory_banks/unregister:
+  /alpha/memory-banks/unregister:
    post:
      parameters:
      - description: JSON-encoded provider data which will be made available to the
@ -4223,7 +4177,53 @@ paths:
          description: OK
      tags:
      - MemoryBanks
-  /models/get:
+  /alpha/memory/insert:
+    post:
+      parameters:
+      - description: JSON-encoded provider data which will be made available to the
+          adapter servicing the API
+        in: header
+        name: X-LlamaStack-ProviderData
+        required: false
+        schema:
+          type: string
+      requestBody:
+        content:
+          application/json:
+            schema:
+              $ref: '#/components/schemas/InsertDocumentsRequest'
+        required: true
+      responses:
+        '200':
+          description: OK
+      tags:
+      - Memory
+  /alpha/memory/query:
+    post:
+      parameters:
+      - description: JSON-encoded provider data which will be made available to the
+          adapter servicing the API
+        in: header
+        name: X-LlamaStack-ProviderData
+        required: false
+        schema:
+          type: string
+      requestBody:
+        content:
+          application/json:
+            schema:
+              $ref: '#/components/schemas/QueryDocumentsRequest'
+        required: true
+      responses:
+        '200':
+          content:
+            application/json:
+              schema:
+                $ref: '#/components/schemas/QueryDocumentsResponse'
+          description: OK
+      tags:
+      - Memory
+  /alpha/models/get:
    get:
      parameters:
      - in: query
@ -4249,7 +4249,7 @@ paths:
          description: OK
      tags:
      - Models
-  /models/list:
+  /alpha/models/list:
    get:
      parameters:
      - description: JSON-encoded provider data which will be made available to the
@ -4268,7 +4268,7 @@ paths:
          description: OK
      tags:
      - Models
-  /models/register:
+  /alpha/models/register:
    post:
      parameters:
      - description: JSON-encoded provider data which will be made available to the
@ -4293,7 +4293,7 @@ paths:
          description: OK
      tags:
      - Models
-  /models/unregister:
+  /alpha/models/unregister:
    post:
      parameters:
      - description: JSON-encoded provider data which will be made available to the
@ -4314,7 +4314,7 @@ paths:
          description: OK
      tags:
      - Models
-  /post_training/job/artifacts:
+  /alpha/post-training/job/artifacts:
    get:
      parameters:
      - in: query
@ -4338,7 +4338,7 @@ paths:
          description: OK
      tags:
      - PostTraining
-  /post_training/job/cancel:
+  /alpha/post-training/job/cancel:
    post:
      parameters:
      - description: JSON-encoded provider data which will be made available to the
@ -4359,7 +4359,7 @@ paths:
          description: OK
      tags:
      - PostTraining
-  /post_training/job/logs:
+  /alpha/post-training/job/logs:
    get:
      parameters:
      - in: query
@ -4383,7 +4383,7 @@ paths:
          description: OK
      tags:
      - PostTraining
-  /post_training/job/status:
+  /alpha/post-training/job/status:
    get:
      parameters:
      - in: query
@ -4407,7 +4407,7 @@ paths:
          description: OK
      tags:
      - PostTraining
-  /post_training/jobs:
+  /alpha/post-training/jobs:
    get:
      parameters:
      - description: JSON-encoded provider data which will be made available to the
@ -4426,7 +4426,7 @@ paths:
          description: OK
      tags:
      - PostTraining
-  /post_training/preference_optimize:
+  /alpha/post-training/preference-optimize:
    post:
      parameters:
      - description: JSON-encoded provider data which will be made available to the
@ -4451,7 +4451,7 @@ paths:
          description: OK
      tags:
      - PostTraining
-  /post_training/supervised_fine_tune:
+  /alpha/post-training/supervised-fine-tune:
    post:
      parameters:
      - description: JSON-encoded provider data which will be made available to the
@ -4476,7 +4476,7 @@ paths:
          description: OK
      tags:
      - PostTraining
-  /providers/list:
+  /alpha/providers/list:
    get:
      parameters:
      - description: JSON-encoded provider data which will be made available to the
@ -4497,7 +4497,7 @@ paths:
          description: OK
      tags:
      - Inspect
-  /routes/list:
+  /alpha/routes/list:
    get:
      parameters:
      - description: JSON-encoded provider data which will be made available to the
@ -4520,7 +4520,7 @@ paths:
          description: OK
      tags:
      - Inspect
-  /safety/run_shield:
+  /alpha/safety/run-shield:
    post:
      parameters:
      - description: JSON-encoded provider data which will be made available to the
@ -4545,57 +4545,7 @@ paths:
          description: OK
      tags:
      - Safety
-  /scoring/score:
-    post:
-      parameters:
-      - description: JSON-encoded provider data which will be made available to the
-          adapter servicing the API
-        in: header
-        name: X-LlamaStack-ProviderData
-        required: false
-        schema:
-          type: string
-      requestBody:
-        content:
-          application/json:
-            schema:
-              $ref: '#/components/schemas/ScoreRequest'
-        required: true
-      responses:
-        '200':
-          content:
-            application/json:
-              schema:
-                $ref: '#/components/schemas/ScoreResponse'
-          description: OK
-      tags:
-      - Scoring
-  /scoring/score_batch:
-    post:
-      parameters:
-      - description: JSON-encoded provider data which will be made available to the
-          adapter servicing the API
-        in: header
-        name: X-LlamaStack-ProviderData
-        required: false
-        schema:
-          type: string
-      requestBody:
-        content:
-          application/json:
-            schema:
-              $ref: '#/components/schemas/ScoreBatchRequest'
-        required: true
-      responses:
-        '200':
-          content:
-            application/json:
-              schema:
-                $ref: '#/components/schemas/ScoreBatchResponse'
-          description: OK
-      tags:
-      - Scoring
-  /scoring_functions/get:
+  /alpha/scoring-functions/get:
    get:
      parameters:
      - in: query
@ -4621,7 +4571,7 @@ paths:
          description: OK
      tags:
      - ScoringFunctions
-  /scoring_functions/list:
+  /alpha/scoring-functions/list:
    get:
      parameters:
      - description: JSON-encoded provider data which will be made available to the
@ -4640,7 +4590,7 @@ paths:
          description: OK
      tags:
      - ScoringFunctions
-  /scoring_functions/register:
+  /alpha/scoring-functions/register:
    post:
      parameters:
      - description: JSON-encoded provider data which will be made available to the
@ -4661,7 +4611,57 @@ paths:
          description: OK
      tags:
      - ScoringFunctions
-  /shields/get:
+  /alpha/scoring/score:
+    post:
+      parameters:
+      - description: JSON-encoded provider data which will be made available to the
+          adapter servicing the API
+        in: header
+        name: X-LlamaStack-ProviderData
+        required: false
+        schema:
+          type: string
+      requestBody:
+        content:
+          application/json:
+            schema:
+              $ref: '#/components/schemas/ScoreRequest'
+        required: true
+      responses:
+        '200':
+          content:
+            application/json:
+              schema:
+                $ref: '#/components/schemas/ScoreResponse'
+          description: OK
+      tags:
+      - Scoring
+  /alpha/scoring/score-batch:
+    post:
+      parameters:
+      - description: JSON-encoded provider data which will be made available to the
+          adapter servicing the API
+        in: header
+        name: X-LlamaStack-ProviderData
+        required: false
+        schema:
+          type: string
+      requestBody:
+        content:
+          application/json:
+            schema:
+              $ref: '#/components/schemas/ScoreBatchRequest'
+        required: true
+      responses:
+        '200':
+          content:
+            application/json:
+              schema:
+                $ref: '#/components/schemas/ScoreBatchResponse'
+          description: OK
+      tags:
+      - Scoring
+  /alpha/shields/get:
    get:
      parameters:
      - in: query
@ -4687,7 +4687,7 @@ paths:
          description: OK
      tags:
      - Shields
-  /shields/list:
+  /alpha/shields/list:
    get:
      parameters:
      - description: JSON-encoded provider data which will be made available to the
@ -4706,7 +4706,7 @@ paths:
          description: OK
      tags:
      - Shields
-  /shields/register:
+  /alpha/shields/register:
    post:
      parameters:
      - description: JSON-encoded provider data which will be made available to the
@ -4731,7 +4731,7 @@ paths:
          description: OK
      tags:
      - Shields
-  /synthetic_data_generation/generate:
+  /alpha/synthetic-data-generation/generate:
    post:
      parameters:
      - description: JSON-encoded provider data which will be made available to the
@ -4756,7 +4756,7 @@ paths:
          description: OK
      tags:
      - SyntheticDataGeneration
-  /telemetry/get_trace:
+  /alpha/telemetry/get-trace:
    get:
      parameters:
      - in: query
@ -4780,7 +4780,7 @@ paths:
          description: OK
      tags:
      - Telemetry
-  /telemetry/log_event:
+  /alpha/telemetry/log-event:
    post:
      parameters:
      - description: JSON-encoded provider data which will be made available to the
--- a/docs/source/getting_started/distributions/self_hosted_distro/fireworks.md
+++ b/docs/source/getting_started/distributions/self_hosted_distro/fireworks.md
@ -2,63 +2,67 @@

 The `llamastack/distribution-fireworks` distribution consists of the following provider configurations.

+| API | Provider(s) |
+|-----|-------------|
+| agents | `inline::meta-reference` |
+| inference | `remote::fireworks` |
+| memory | `inline::faiss`, `remote::chromadb`, `remote::pgvector` |
+| safety | `inline::llama-guard` |
+| telemetry | `inline::meta-reference` |

-| **API**         	| **Inference** 	| **Agents**     	| **Memory**                                       	| **Safety**     	| **Telemetry**  	|
-|-----------------	|---------------	|----------------	|--------------------------------------------------	|----------------	|----------------	|
-| **Provider(s)** 	| remote::fireworks   	| meta-reference 	| meta-reference 	| meta-reference 	| meta-reference 	|

-### Step 0. Prerequisite
- Make sure you have access to a fireworks API Key. You can get one by visiting [fireworks.ai](https://fireworks.ai/)
+### Environment Variables

-### Step 1. Start the Distribution (Single Node CPU)
+The following environment variables can be configured:

-#### (Option 1) Start Distribution Via Docker
-> [!NOTE]
-> This assumes you have an hosted endpoint at Fireworks with API Key.
+- `LLAMASTACK_PORT`: Port for the Llama Stack distribution server (default: `5001`)
+- `FIREWORKS_API_KEY`: Fireworks.AI API Key (default: ``)

-```
-$ cd distributions/fireworks && docker compose up
+### Models
+
+The following models are available by default:
+
+- `meta-llama/Llama-3.1-8B-Instruct (fireworks/llama-v3p1-8b-instruct)`
+- `meta-llama/Llama-3.1-70B-Instruct (fireworks/llama-v3p1-70b-instruct)`
+- `meta-llama/Llama-3.1-405B-Instruct-FP8 (fireworks/llama-v3p1-405b-instruct)`
+- `meta-llama/Llama-3.2-3B-Instruct (fireworks/llama-v3p2-1b-instruct)`
+- `meta-llama/Llama-3.2-11B-Vision-Instruct (fireworks/llama-v3p2-3b-instruct)`
+- `meta-llama/Llama-3.2-11B-Vision-Instruct (fireworks/llama-v3p2-11b-vision-instruct)`
+- `meta-llama/Llama-3.2-90B-Vision-Instruct (fireworks/llama-v3p2-90b-vision-instruct)`
+- `meta-llama/Llama-Guard-3-8B (fireworks/llama-guard-3-8b)`
+- `meta-llama/Llama-Guard-3-11B-Vision (fireworks/llama-guard-3-11b-vision)`
+
+
+### Prerequisite: API Keys
+
+Make sure you have access to a Fireworks API Key. You can get one by visiting [fireworks.ai](https://fireworks.ai/).
+
+
+## Running Llama Stack with Fireworks
+
+You can do this via Conda (build code) or Docker which has a pre-built image.
+
+### Via Docker
+
+This method allows you to get started quickly without having to build the distribution code.
+
+```bash
+LLAMA_STACK_PORT=5001
+docker run \
+  -it \
+  -p $LLAMA_STACK_PORT:$LLAMA_STACK_PORT \
+  -v ./run.yaml:/root/my-run.yaml \
+  llamastack/distribution-fireworks \
+  --yaml-config /root/my-run.yaml \
+  --port $LLAMA_STACK_PORT \
+  --env FIREWORKS_API_KEY=$FIREWORKS_API_KEY
 ```

-Make sure in you `run.yaml` file, you inference provider is pointing to the correct Fireworks URL server endpoint. E.g.
-```
-inference:
-  - provider_id: fireworks
-    provider_type: remote::fireworks
-    config:
-      url: https://api.fireworks.ai/inference
-      api_key: <optional api key>
-```
-
-#### (Option 2) Start Distribution Via Conda
+### Via Conda

 ```bash
 llama stack build --template fireworks --image-type conda
-# -- modify run.yaml to a valid Fireworks server endpoint
-llama stack run ./run.yaml
-```
-
-
-### (Optional) Model Serving
-
-Use `llama-stack-client models list` to check the available models served by Fireworks.
-```
-$ llama-stack-client models list
-+------------------------------+------------------------------+---------------+------------+
-| identifier                   | llama_model                  | provider_id   | metadata   |
-+==============================+==============================+===============+============+
-| Llama3.1-8B-Instruct         | Llama3.1-8B-Instruct         | fireworks0    | {}         |
-+------------------------------+------------------------------+---------------+------------+
-| Llama3.1-70B-Instruct        | Llama3.1-70B-Instruct        | fireworks0    | {}         |
-+------------------------------+------------------------------+---------------+------------+
-| Llama3.1-405B-Instruct       | Llama3.1-405B-Instruct       | fireworks0    | {}         |
-+------------------------------+------------------------------+---------------+------------+
-| Llama3.2-1B-Instruct         | Llama3.2-1B-Instruct         | fireworks0    | {}         |
-+------------------------------+------------------------------+---------------+------------+
-| Llama3.2-3B-Instruct         | Llama3.2-3B-Instruct         | fireworks0    | {}         |
-+------------------------------+------------------------------+---------------+------------+
-| Llama3.2-11B-Vision-Instruct | Llama3.2-11B-Vision-Instruct | fireworks0    | {}         |
-+------------------------------+------------------------------+---------------+------------+
-| Llama3.2-90B-Vision-Instruct | Llama3.2-90B-Vision-Instruct | fireworks0    | {}         |
-+------------------------------+------------------------------+---------------+------------+
+llama stack run ./run.yaml \
+  --port 5001 \
+  --env FIREWORKS_API_KEY=$FIREWORKS_API_KEY
 ```
--- a/docs/source/getting_started/distributions/self_hosted_distro/meta-reference-gpu.md
+++ b/docs/source/getting_started/distributions/self_hosted_distro/meta-reference-gpu.md
@ -1,15 +1,32 @@
 # Meta Reference Distribution

-The `llamastack/distribution-meta-reference-gpu` distribution consists of the following provider configurations.
+The `llamastack/distribution-meta-reference-gpu` distribution consists of the following provider configurations:
+
+| API | Provider(s) |
+|-----|-------------|
+| agents | `inline::meta-reference` |
+| inference | `inline::meta-reference` |
+| memory | `inline::faiss`, `remote::chromadb`, `remote::pgvector` |
+| safety | `inline::llama-guard` |
+| telemetry | `inline::meta-reference` |


-| **API**         	| **Inference** 	| **Agents**     	| **Memory**                                       	| **Safety**     	| **Telemetry**  	|
-|-----------------	|---------------	|----------------	|--------------------------------------------------	|----------------	|----------------	|
-| **Provider(s)** 	| meta-reference  	| meta-reference 	| meta-reference, remote::pgvector, remote::chroma 	| meta-reference 	| meta-reference 	|
+Note that you need access to nvidia GPUs to run this distribution. This distribution is not compatible with CPU-only machines or machines with AMD GPUs.
+
+### Environment Variables
+
+The following environment variables can be configured:
+
+- `LLAMASTACK_PORT`: Port for the Llama Stack distribution server (default: `5001`)
+- `INFERENCE_MODEL`: Inference model loaded into the Meta Reference server (default: `meta-llama/Llama-3.2-3B-Instruct`)
+- `INFERENCE_CHECKPOINT_DIR`: Directory containing the Meta Reference model checkpoint (default: `null`)
+- `SAFETY_MODEL`: Name of the safety (Llama-Guard) model to use (default: `meta-llama/Llama-Guard-3-1B`)
+- `SAFETY_CHECKPOINT_DIR`: Directory containing the Llama-Guard model checkpoint (default: `null`)


-### Step 0. Prerequisite - Downloading Models
-Please make sure you have llama model checkpoints downloaded in `~/.llama` before proceeding. See [installation guide](https://llama-stack.readthedocs.io/en/latest/cli_reference/download_models.html) here to download the models.
+## Prerequisite: Downloading Models
+
+Please make sure you have llama model checkpoints downloaded in `~/.llama` before proceeding. See [installation guide](https://llama-stack.readthedocs.io/en/latest/cli_reference/download_models.html) here to download the models. Run `llama model list` to see the available models to download, and `llama model download` to download the checkpoints.

 ```
 $ ls ~/.llama/checkpoints
@ -17,55 +34,56 @@ Llama3.1-8B           Llama3.2-11B-Vision-Instruct  Llama3.2-1B-Instruct  Llama3
 Llama3.1-8B-Instruct  Llama3.2-1B                   Llama3.2-3B-Instruct  Llama-Guard-3-1B              Prompt-Guard-86M
 ```

-### Step 1. Start the Distribution
+## Running the Distribution

-#### (Option 1) Start with Docker
-```
-$ cd distributions/meta-reference-gpu && docker compose up
+You can do this via Conda (build code) or Docker which has a pre-built image.
+
+### Via Docker
+
+This method allows you to get started quickly without having to build the distribution code.
+
+```bash
+LLAMA_STACK_PORT=5001
+docker run \
+  -it \
+  -p $LLAMA_STACK_PORT:$LLAMA_STACK_PORT \
+  -v ./run.yaml:/root/my-run.yaml \
+  llamastack/distribution-meta-reference-gpu \
+  /root/my-run.yaml \
+  --port $LLAMA_STACK_PORT \
+  --env INFERENCE_MODEL=meta-llama/Llama-3.2-3B-Instruct
 ```

-> [!NOTE]
-> This assumes you have access to GPU to start a local server with access to your GPU.
+If you are using Llama Stack Safety / Shield APIs, use:

-
-> [!NOTE]
-> `~/.llama` should be the path containing downloaded weights of Llama models.
-
-
-This will download and start running a pre-built docker container. Alternatively, you may use the following commands:
-
-```
-docker run -it -p 5000:5000 -v ~/.llama:/root/.llama -v ./run.yaml:/root/my-run.yaml --gpus=all distribution-meta-reference-gpu --yaml_config /root/my-run.yaml
+```bash
+docker run \
+  -it \
+  -p $LLAMA_STACK_PORT:$LLAMA_STACK_PORT \
+  -v ./run-with-safety.yaml:/root/my-run.yaml \
+  llamastack/distribution-meta-reference-gpu \
+  /root/my-run.yaml \
+  --port $LLAMA_STACK_PORT \
+  --env INFERENCE_MODEL=meta-llama/Llama-3.2-3B-Instruct \
+  --env SAFETY_MODEL=meta-llama/Llama-Guard-3-1B
 ```

-#### (Option 2) Start with Conda
+### Via Conda

-1. Install the `llama` CLI. See [CLI Reference](https://llama-stack.readthedocs.io/en/latest/cli_reference/index.html)
+Make sure you have done `pip install llama-stack` and have the Llama Stack CLI available.

-2. Build the `meta-reference-gpu` distribution
-
-```
-$ llama stack build --template meta-reference-gpu --image-type conda
+```bash
+llama stack build --template meta-reference-gpu --image-type conda
+llama stack run ./run.yaml \
+  --port 5001 \
+  --env INFERENCE_MODEL=meta-llama/Llama-3.2-3B-Instruct
 ```

-3. Start running distribution
-```
-$ cd distributions/meta-reference-gpu
-$ llama stack run ./run.yaml
-```
+If you are using Llama Stack Safety / Shield APIs, use:

-### (Optional) Serving a new model
-You may change the `config.model` in `run.yaml` to update the model currently being served by the distribution. Make sure you have the model checkpoint downloaded in your `~/.llama`.
+```bash
+llama stack run ./run-with-safety.yaml \
+  --port 5001 \
+  --env INFERENCE_MODEL=meta-llama/Llama-3.2-3B-Instruct \
+  --env SAFETY_MODEL=meta-llama/Llama-Guard-3-1B
 ```
-inference:
-  - provider_id: meta0
-    provider_type: inline::meta-reference
-    config:
-      model: Llama3.2-11B-Vision-Instruct
-      quantization: null
-      torch_seed: null
-      max_seq_len: 4096
-      max_batch_size: 1
-```
-
-Run `llama model list` to see the available models to download, and `llama model download` to download the checkpoints.
--- a/docs/source/getting_started/distributions/self_hosted_distro/ollama.md
+++ b/docs/source/getting_started/distributions/self_hosted_distro/ollama.md
@ -2,103 +2,114 @@

 The `llamastack/distribution-ollama` distribution consists of the following provider configurations.

-| **API**         	| **Inference**  	| **Agents**     	| **Memory**                       	  | **Safety**     	| **Telemetry**  	|
-|-----------------	|----------------	|----------------	|------------------------------------	|----------------	|----------------	|
-| **Provider(s)** 	| remote::ollama 	| meta-reference 	| remote::pgvector, remote::chromadb 	| meta-reference 	| meta-reference 	|
+| API | Provider(s) |
+|-----|-------------|
+| agents | `inline::meta-reference` |
+| inference | `remote::ollama` |
+| memory | `inline::faiss`, `remote::chromadb`, `remote::pgvector` |
+| safety | `inline::llama-guard` |
+| telemetry | `inline::meta-reference` |


-## Using Docker Compose
+You should use this distribution if you have a regular desktop machine without very powerful GPUs. Of course, if you have powerful GPUs, you can still continue using this distribution since Ollama supports GPU acceleration.### Environment Variables

-You can use `docker compose` to start a Ollama server and connect with Llama Stack server in a single command.
+The following environment variables can be configured:

-### Docker: Start the Distribution (Single Node regular Desktop machine)
+- `LLAMASTACK_PORT`: Port for the Llama Stack distribution server (default: `5001`)
+- `OLLAMA_URL`: URL of the Ollama server (default: `http://127.0.0.1:11434`)
+- `INFERENCE_MODEL`: Inference model loaded into the Ollama server (default: `meta-llama/Llama-3.2-3B-Instruct`)
+- `SAFETY_MODEL`: Safety model loaded into the Ollama server (default: `meta-llama/Llama-Guard-3-1B`)

-> [!NOTE]
-> This will start an ollama server with CPU only, please see [Ollama Documentations](https://github.com/ollama/ollama) for serving models on CPU only.
+
+## Setting up Ollama server
+
+Please check the [Ollama Documentation](https://github.com/ollama/ollama) on how to install and run Ollama. After installing Ollama, you need to run `ollama serve` to start the server.
+
+In order to load models, you can run:

 ```bash
-$ cd distributions/ollama; docker compose up
+export INFERENCE_MODEL="meta-llama/Llama-3.2-3B-Instruct"
+
+# ollama names this model differently, and we must use the ollama name when loading the model
+export OLLAMA_INFERENCE_MODEL="llama3.2:3b-instruct-fp16"
+ollama run $OLLAMA_INFERENCE_MODEL --keepalive 60m
 ```

-### Docker: Start a Distribution (Single Node with nvidia GPUs)
-
-> [!NOTE]
-> This assumes you have access to GPU to start a Ollama server with access to your GPU.
+If you are using Llama Stack Safety / Shield APIs, you will also need to pull and run the safety model.

 ```bash
-$ cd distributions/ollama-gpu; docker compose up
+export SAFETY_MODEL="meta-llama/Llama-Guard-3-1B"
+
+# ollama names this model differently, and we must use the ollama name when loading the model
+export OLLAMA_SAFETY_MODEL="llama-guard3:1b"
+ollama run $OLLAMA_SAFETY_MODEL --keepalive 60m
 ```

-You will see outputs similar to following ---
+## Running Llama Stack
+
+Now you are ready to run Llama Stack with Ollama as the inference provider. You can do this via Conda (build code) or Docker which has a pre-built image.
+
+### Via Docker
+
+This method allows you to get started quickly without having to build the distribution code.
+
 ```bash
-[ollama]               | [GIN] 2024/10/18 - 21:19:41 | 200 |     226.841µs |             ::1 | GET      "/api/ps"
-[ollama]               | [GIN] 2024/10/18 - 21:19:42 | 200 |      60.908µs |             ::1 | GET      "/api/ps"
-INFO:     Started server process [1]
-INFO:     Waiting for application startup.
-INFO:     Application startup complete.
-INFO:     Uvicorn running on http://[::]:5000 (Press CTRL+C to quit)
-[llamastack] | Resolved 12 providers
-[llamastack] |  inner-inference => ollama0
-[llamastack] |  models => __routing_table__
-[llamastack] |  inference => __autorouted__
+export LLAMA_STACK_PORT=5001
+docker run \
+  -it \
+  -p $LLAMA_STACK_PORT:$LLAMA_STACK_PORT \
+  -v ~/.llama:/root/.llama \
+  -v ./run.yaml:/root/my-run.yaml \
+  llamastack/distribution-ollama \
+  --yaml-config /root/my-run.yaml \
+  --port $LLAMA_STACK_PORT \
+  --env INFERENCE_MODEL=$INFERENCE_MODEL \
+  --env OLLAMA_URL=http://host.docker.internal:11434
 ```

-To kill the server
+If you are using Llama Stack Safety / Shield APIs, use:
+
 ```bash
-docker compose down
+docker run \
+  -it \
+  -p $LLAMA_STACK_PORT:$LLAMA_STACK_PORT \
+  -v ~/.llama:/root/.llama \
+  -v ./run-with-safety.yaml:/root/my-run.yaml \
+  llamastack/distribution-ollama \
+  --yaml-config /root/my-run.yaml \
+  --port $LLAMA_STACK_PORT \
+  --env INFERENCE_MODEL=$INFERENCE_MODEL \
+  --env SAFETY_MODEL=$SAFETY_MODEL \
+  --env OLLAMA_URL=http://host.docker.internal:11434
 ```

-## Starting Ollama and Llama Stack separately
+### Via Conda

-If you wish to separately spin up a Ollama server, and connect with Llama Stack, you should use the following commands.
-
-#### Start Ollama server
- Please check the [Ollama Documentation](https://github.com/ollama/ollama) for more details.
-
-**Via Docker**
-```bash
-docker run -d -v ollama:/root/.ollama -p 11434:11434 --name ollama ollama/ollama
-```
-
-**Via CLI**
-```bash
-ollama run <model_id>
-```
-
-#### Start Llama Stack server pointing to Ollama server
-
-**Via Conda**
+Make sure you have done `pip install llama-stack` and have the Llama Stack CLI available.

 ```bash
+export LLAMA_STACK_PORT=5001
+
 llama stack build --template ollama --image-type conda
-llama stack run ./gpu/run.yaml
+llama stack run ./run.yaml \
+  --port $LLAMA_STACK_PORT \
+  --env INFERENCE_MODEL=$INFERENCE_MODEL \
+  --env OLLAMA_URL=http://localhost:11434
 ```

-**Via Docker**
-```
-docker run --network host -it -p 5000:5000 -v ~/.llama:/root/.llama -v ./gpu/run.yaml:/root/llamastack-run-ollama.yaml --gpus=all llamastack/distribution-ollama --yaml_config /root/llamastack-run-ollama.yaml
+If you are using Llama Stack Safety / Shield APIs, use:
+
+```bash
+llama stack run ./run-with-safety.yaml \
+  --port $LLAMA_STACK_PORT \
+  --env INFERENCE_MODEL=$INFERENCE_MODEL \
+  --env SAFETY_MODEL=$SAFETY_MODEL \
+  --env OLLAMA_URL=http://localhost:11434
 ```

-Make sure in your `run.yaml` file, your inference provider is pointing to the correct Ollama endpoint. E.g.
-```yaml
-inference:
-  - provider_id: ollama0
-    provider_type: remote::ollama
-    config:
-      url: http://127.0.0.1:14343
-```

 ### (Optional) Update Model Serving Configuration

-#### Downloading model via Ollama
-
-You can use ollama for managing model downloads.
-
-```bash
-ollama pull llama3.1:8b-instruct-fp16
-ollama pull llama3.1:70b-instruct-fp16
-```
-
 > [!NOTE]
 > Please check the [OLLAMA_SUPPORTED_MODELS](https://github.com/meta-llama/llama-stack/blob/main/llama_stack/providers.remote/inference/ollama/ollama.py) for the supported Ollama models.

--- a/docs/source/getting_started/distributions/self_hosted_distro/remote-vllm.md
+++ b/docs/source/getting_started/distributions/self_hosted_distro/remote-vllm.md
@ -0,0 +1,144 @@
+# Remote vLLM Distribution
+
+The `llamastack/distribution-remote-vllm` distribution consists of the following provider configurations:
+
+| API | Provider(s) |
+|-----|-------------|
+| agents | `inline::meta-reference` |
+| inference | `remote::vllm` |
+| memory | `inline::faiss`, `remote::chromadb`, `remote::pgvector` |
+| safety | `inline::llama-guard` |
+| telemetry | `inline::meta-reference` |
+
+
+You can use this distribution if you have GPUs and want to run an independent vLLM server container for running inference.
+
+### Environment Variables
+
+The following environment variables can be configured:
+
+- `LLAMASTACK_PORT`: Port for the Llama Stack distribution server (default: `5001`)
+- `INFERENCE_MODEL`: Inference model loaded into the vLLM server (default: `meta-llama/Llama-3.2-3B-Instruct`)
+- `VLLM_URL`: URL of the vLLM server with the main inference model (default: `http://host.docker.internal:5100}/v1`)
+- `MAX_TOKENS`: Maximum number of tokens for generation (default: `4096`)
+- `SAFETY_VLLM_URL`: URL of the vLLM server with the safety model (default: `http://host.docker.internal:5101/v1`)
+- `SAFETY_MODEL`: Name of the safety (Llama-Guard) model to use (default: `meta-llama/Llama-Guard-3-1B`)
+
+
+## Setting up vLLM server
+
+Please check the [vLLM Documentation](https://docs.vllm.ai/en/v0.5.5/serving/deploying_with_docker.html) to get a vLLM endpoint. Here is a sample script to start a vLLM server locally via Docker:
+
+```bash
+export INFERENCE_PORT=8000
+export INFERENCE_MODEL=meta-llama/Llama-3.2-3B-Instruct
+export CUDA_VISIBLE_DEVICES=0
+
+docker run \
+    --runtime nvidia \
+    --gpus $CUDA_VISIBLE_DEVICES \
+    -v ~/.cache/huggingface:/root/.cache/huggingface \
+    --env "HUGGING_FACE_HUB_TOKEN=$HF_TOKEN" \
+    -p $INFERENCE_PORT:$INFERENCE_PORT \
+    --ipc=host \
+    vllm/vllm-openai:latest \
+    --gpu-memory-utilization 0.7 \
+    --model $INFERENCE_MODEL \
+    --port $INFERENCE_PORT
+```
+
+If you are using Llama Stack Safety / Shield APIs, then you will need to also run another instance of a vLLM with a corresponding safety model like `meta-llama/Llama-Guard-3-1B` using a script like:
+
+```bash
+export SAFETY_PORT=8081
+export SAFETY_MODEL=meta-llama/Llama-Guard-3-1B
+export CUDA_VISIBLE_DEVICES=1
+
+docker run \
+    --runtime nvidia \
+    --gpus $CUDA_VISIBLE_DEVICES \
+    -v ~/.cache/huggingface:/root/.cache/huggingface \
+    --env "HUGGING_FACE_HUB_TOKEN=$HF_TOKEN" \
+    -p $SAFETY_PORT:$SAFETY_PORT \
+    --ipc=host \
+    vllm/vllm-openai:latest \
+    --gpu-memory-utilization 0.7 \
+    --model $SAFETY_MODEL \
+    --port $SAFETY_PORT
+```
+
+## Running Llama Stack
+
+Now you are ready to run Llama Stack with vLLM as the inference provider. You can do this via Conda (build code) or Docker which has a pre-built image.
+
+### Via Docker
+
+This method allows you to get started quickly without having to build the distribution code.
+
+```bash
+export INFERENCE_PORT=8000
+export INFERENCE_MODEL=meta-llama/Llama-3.2-3B-Instruct
+export LLAMA_STACK_PORT=5001
+
+docker run \
+  -it \
+  -p $LLAMA_STACK_PORT:$LLAMA_STACK_PORT \
+  -v ./run.yaml:/root/my-run.yaml \
+  llamastack/distribution-remote-vllm \
+  --yaml-config /root/my-run.yaml \
+  --port $LLAMA_STACK_PORT \
+  --env INFERENCE_MODEL=$INFERENCE_MODEL \
+  --env VLLM_URL=http://host.docker.internal:$INFERENCE_PORT/v1
+```
+
+If you are using Llama Stack Safety / Shield APIs, use:
+
+```bash
+export SAFETY_PORT=8081
+export SAFETY_MODEL=meta-llama/Llama-Guard-3-1B
+
+docker run \
+  -it \
+  -p $LLAMA_STACK_PORT:$LLAMA_STACK_PORT \
+  -v ./run-with-safety.yaml:/root/my-run.yaml \
+  llamastack/distribution-remote-vllm \
+  --yaml-config /root/my-run.yaml \
+  --port $LLAMA_STACK_PORT \
+  --env INFERENCE_MODEL=$INFERENCE_MODEL \
+  --env VLLM_URL=http://host.docker.internal:$INFERENCE_PORT/v1 \
+  --env SAFETY_MODEL=$SAFETY_MODEL \
+  --env SAFETY_VLLM_URL=http://host.docker.internal:$SAFETY_PORT/v1
+```
+
+
+### Via Conda
+
+Make sure you have done `pip install llama-stack` and have the Llama Stack CLI available.
+
+```bash
+export INFERENCE_PORT=8000
+export INFERENCE_MODEL=meta-llama/Llama-3.2-3B-Instruct
+export LLAMA_STACK_PORT=5001
+
+cd distributions/remote-vllm
+llama stack build --template remote-vllm --image-type conda
+
+llama stack run ./run.yaml \
+  --port $LLAMA_STACK_PORT \
+  --env INFERENCE_MODEL=$INFERENCE_MODEL \
+  --env VLLM_URL=http://localhost:$INFERENCE_PORT/v1
+```
+
+If you are using Llama Stack Safety / Shield APIs, use:
+
+```bash
+export SAFETY_PORT=8081
+export SAFETY_MODEL=meta-llama/Llama-Guard-3-1B
+
+llama stack run ./run-with-safety.yaml \
+  --port $LLAMA_STACK_PORT \
+  --env INFERENCE_MODEL=$INFERENCE_MODEL \
+  --env VLLM_URL=http://localhost:$INFERENCE_PORT/v1 \
+  --env SAFETY_MODEL=$SAFETY_MODEL \
+  --env SAFETY_VLLM_URL=http://localhost:$SAFETY_PORT/v1
+```
--- a/docs/source/getting_started/distributions/self_hosted_distro/remote_vllm.md
+++ b/docs/source/getting_started/distributions/self_hosted_distro/remote_vllm.md
@ -1,83 +0,0 @@
-# Remote vLLM Distribution
-
-The `llamastack/distribution-remote-vllm` distribution consists of the following provider configurations.
-
-| **API**           | **Inference**   | **Agents**      | **Memory**                          | **Safety**     	| **Telemetry**  	|
-|-----------------  |---------------- |---------------- |------------------------------------	|----------------	|----------------	|
-| **Provider(s)**   | remote::vllm  	| meta-reference 	| remote::pgvector, remote::chromadb 	| meta-reference 	| meta-reference 	|
-
-You can use this distribution if you have GPUs and want to run an independent vLLM server container for running inference.
-
-## Using Docker Compose
-
-You can use `docker compose` to start a vLLM container and Llama Stack server container together.
-
-> [!NOTE]
-> This assumes you have access to GPU to start a vLLM server with access to your GPU.
-
-```bash
-$ cd distributions/remote-vllm; docker compose up
-```
-
-You will see outputs similar to following ---
-```
-<TO BE FILLED>
-```
-
-To kill the server
-```bash
-docker compose down
-```
-
-## Starting vLLM and Llama Stack separately
-
-You may want to start a vLLM server and connect with Llama Stack manually. There are two ways to start a vLLM server and connect with Llama Stack.
-
-
-#### Start vLLM server.
-
-```bash
-docker run --runtime nvidia --gpus all \
-    -v ~/.cache/huggingface:/root/.cache/huggingface \
-    --env "HUGGING_FACE_HUB_TOKEN=<secret>" \
-    -p 8000:8000 \
-    --ipc=host \
-    vllm/vllm-openai:latest \
-    --model meta-llama/Llama-3.1-8B-Instruct
-```
-
-Please check the [vLLM Documentation](https://docs.vllm.ai/en/v0.5.5/serving/deploying_with_docker.html) for more details.
-
-
-#### Start Llama Stack server pointing to your vLLM server
-
-
-We have provided a template `run.yaml` file in the `distributions/remote-vllm` directory. Please make sure to modify the `inference.provider_id` to point to your vLLM server endpoint. As an example, if your vLLM server is running on `http://127.0.0.1:8000`, your `run.yaml` file should look like the following:
-```yaml
-inference:
-  - provider_id: vllm0
-    provider_type: remote::vllm
-    config:
-      url: http://127.0.0.1:8000
-```
-
-**Via Conda**
-
-If you are using Conda, you can build and run the Llama Stack server with the following commands:
-```bash
-cd distributions/remote-vllm
-llama stack build --template remote_vllm --image-type conda
-llama stack run run.yaml
-```
-
-**Via Docker**
-
-You can use the Llama Stack Docker image to start the server with the following command:
-```bash
-docker run --network host -it -p 5000:5000 \
-  -v ~/.llama:/root/.llama \
-  -v ./gpu/run.yaml:/root/llamastack-run-remote-vllm.yaml \
-  --gpus=all \
-  llamastack/distribution-remote-vllm \
-  --yaml_config /root/llamastack-run-remote-vllm.yaml
-```
--- a/docs/source/getting_started/distributions/self_hosted_distro/tgi.md
+++ b/docs/source/getting_started/distributions/self_hosted_distro/tgi.md
@ -2,94 +2,125 @@

 The `llamastack/distribution-tgi` distribution consists of the following provider configurations.

-
-| **API**         	| **Inference** 	| **Agents**     	| **Memory**                                       	| **Safety**     	| **Telemetry**  	|
-|-----------------	|---------------	|----------------	|--------------------------------------------------	|----------------	|----------------	|
-| **Provider(s)** 	| remote::tgi   	| meta-reference 	| meta-reference, remote::pgvector, remote::chroma 	| meta-reference 	| meta-reference 	|
+| API | Provider(s) |
+|-----|-------------|
+| agents | `inline::meta-reference` |
+| inference | `remote::tgi` |
+| memory | `inline::faiss`, `remote::chromadb`, `remote::pgvector` |
+| safety | `inline::llama-guard` |
+| telemetry | `inline::meta-reference` |


-### Docker: Start the Distribution (Single Node GPU)
+You can use this distribution if you have GPUs and want to run an independent TGI server container for running inference.

-> [!NOTE]
-> This assumes you have access to GPU to start a TGI server with access to your GPU.
+### Environment Variables
+
+The following environment variables can be configured:
+
+- `LLAMASTACK_PORT`: Port for the Llama Stack distribution server (default: `5001`)
+- `INFERENCE_MODEL`: Inference model loaded into the TGI server (default: `meta-llama/Llama-3.2-3B-Instruct`)
+- `TGI_URL`: URL of the TGI server with the main inference model (default: `http://127.0.0.1:8080}/v1`)
+- `TGI_SAFETY_URL`: URL of the TGI server with the safety model (default: `http://127.0.0.1:8081/v1`)
+- `SAFETY_MODEL`: Name of the safety (Llama-Guard) model to use (default: `meta-llama/Llama-Guard-3-1B`)


-```
-$ cd distributions/tgi && docker compose up
+## Setting up TGI server
+
+Please check the [TGI Getting Started Guide](https://github.com/huggingface/text-generation-inference?tab=readme-ov-file#get-started) to get a TGI endpoint. Here is a sample script to start a TGI server locally via Docker:
+
+```bash
+export INFERENCE_PORT=8080
+export INFERENCE_MODEL=meta-llama/Llama-3.2-3B-Instruct
+export CUDA_VISIBLE_DEVICES=0
+
+docker run --rm -it \
+  -v $HOME/.cache/huggingface:/data \
+  -p $INFERENCE_PORT:$INFERENCE_PORT \
+  --gpus $CUDA_VISIBLE_DEVICES \
+  ghcr.io/huggingface/text-generation-inference:2.3.1 \
+  --dtype bfloat16 \
+  --usage-stats off \
+  --sharded false \
+  --cuda-memory-fraction 0.7 \
+  --model-id $INFERENCE_MODEL \
+  --port $INFERENCE_PORT
 ```

-The script will first start up TGI server, then start up Llama Stack distribution server hooking up to the remote TGI provider for inference. You should be able to see the following outputs --
-```
-[text-generation-inference] | 2024-10-15T18:56:33.810397Z  INFO text_generation_router::server: router/src/server.rs:1813: Using config Some(Llama)
-[text-generation-inference] | 2024-10-15T18:56:33.810448Z  WARN text_generation_router::server: router/src/server.rs:1960: Invalid hostname, defaulting to 0.0.0.0
-[text-generation-inference] | 2024-10-15T18:56:33.864143Z  INFO text_generation_router::server: router/src/server.rs:2353: Connected
-INFO:     Started server process [1]
-INFO:     Waiting for application startup.
-INFO:     Application startup complete.
-INFO:     Uvicorn running on http://[::]:5000 (Press CTRL+C to quit)
+If you are using Llama Stack Safety / Shield APIs, then you will need to also run another instance of a TGI with a corresponding safety model like `meta-llama/Llama-Guard-3-1B` using a script like:
+
+```bash
+export SAFETY_PORT=8081
+export SAFETY_MODEL=meta-llama/Llama-Guard-3-1B
+export CUDA_VISIBLE_DEVICES=1
+
+docker run --rm -it \
+  -v $HOME/.cache/huggingface:/data \
+  -p $SAFETY_PORT:$SAFETY_PORT \
+  --gpus $CUDA_VISIBLE_DEVICES \
+  ghcr.io/huggingface/text-generation-inference:2.3.1 \
+  --dtype bfloat16 \
+  --usage-stats off \
+  --sharded false \
+  --model-id $SAFETY_MODEL \
+  --port $SAFETY_PORT
 ```

-To kill the server
-```
-docker compose down
+## Running Llama Stack
+
+Now you are ready to run Llama Stack with TGI as the inference provider. You can do this via Conda (build code) or Docker which has a pre-built image.
+
+### Via Docker
+
+This method allows you to get started quickly without having to build the distribution code.
+
+```bash
+LLAMA_STACK_PORT=5001
+docker run \
+  -it \
+  -p $LLAMA_STACK_PORT:$LLAMA_STACK_PORT \
+  -v ./run.yaml:/root/my-run.yaml \
+  llamastack/distribution-tgi \
+  --yaml-config /root/my-run.yaml \
+  --port $LLAMA_STACK_PORT \
+  --env INFERENCE_MODEL=$INFERENCE_MODEL \
+  --env TGI_URL=http://host.docker.internal:$INFERENCE_PORT
 ```

+If you are using Llama Stack Safety / Shield APIs, use:

-### Conda: TGI server + llama stack run
-
-If you wish to separately spin up a TGI server, and connect with Llama Stack, you may use the following commands.
-
-#### Start TGI server locally
- Please check the [TGI Getting Started Guide](https://github.com/huggingface/text-generation-inference?tab=readme-ov-file#get-started) to get a TGI endpoint.
-
-```
-docker run --rm -it -v $HOME/.cache/huggingface:/data -p 5009:5009 --gpus all ghcr.io/huggingface/text-generation-inference:latest --dtype bfloat16 --usage-stats on --sharded false --model-id meta-llama/Llama-3.1-8B-Instruct --port 5009
+```bash
+docker run \
+  -it \
+  -p $LLAMA_STACK_PORT:$LLAMA_STACK_PORT \
+  -v ./run-with-safety.yaml:/root/my-run.yaml \
+  llamastack/distribution-tgi \
+  --yaml-config /root/my-run.yaml \
+  --port $LLAMA_STACK_PORT \
+  --env INFERENCE_MODEL=$INFERENCE_MODEL \
+  --env TGI_URL=http://host.docker.internal:$INFERENCE_PORT \
+  --env SAFETY_MODEL=$SAFETY_MODEL \
+  --env TGI_SAFETY_URL=http://host.docker.internal:$SAFETY_PORT
 ```

-#### Start Llama Stack server pointing to TGI server
+### Via Conda

-**Via Conda**
+Make sure you have done `pip install llama-stack` and have the Llama Stack CLI available.

 ```bash
 llama stack build --template tgi --image-type conda
-# -- start a TGI server endpoint
-llama stack run ./gpu/run.yaml
+llama stack run ./run.yaml
+  --port 5001
+  --env INFERENCE_MODEL=$INFERENCE_MODEL
+  --env TGI_URL=http://127.0.0.1:$INFERENCE_PORT
 ```

-**Via Docker**
-```
-docker run --network host -it -p 5000:5000 -v ./run.yaml:/root/my-run.yaml --gpus=all llamastack/distribution-tgi --yaml_config /root/my-run.yaml
-```
+If you are using Llama Stack Safety / Shield APIs, use:

-Make sure in you `run.yaml` file, you inference provider is pointing to the correct TGI server endpoint. E.g.
-```
-inference:
-  - provider_id: tgi0
-    provider_type: remote::tgi
-    config:
-      url: http://127.0.0.1:5009
-```
-
-
-### (Optional) Update Model Serving Configuration
-To serve a new model with `tgi`, change the docker command flag `--model-id <model-to-serve>`.
-
-This can be done by edit the `command` args in `compose.yaml`. E.g. Replace "Llama-3.2-1B-Instruct" with the model you want to serve.
-
-```
-command: ["--dtype", "bfloat16", "--usage-stats", "on", "--sharded", "false", "--model-id", "meta-llama/Llama-3.2-1B-Instruct", "--port", "5009", "--cuda-memory-fraction", "0.3"]
-```
-
-or by changing the docker run command's `--model-id` flag
-```
-docker run --rm -it -v $HOME/.cache/huggingface:/data -p 5009:5009 --gpus all ghcr.io/huggingface/text-generation-inference:latest --dtype bfloat16 --usage-stats on --sharded false --model-id meta-llama/Llama-3.2-1B-Instruct --port 5009
-```
-
-In `run.yaml`, make sure you point the correct server endpoint to the TGI server endpoint serving your model.
-```
-inference:
-  - provider_id: tgi0
-    provider_type: remote::tgi
-    config:
-      url: http://127.0.0.1:5009
+```bash
+llama stack run ./run-with-safety.yaml
+  --port 5001
+  --env INFERENCE_MODEL=$INFERENCE_MODEL
+  --env TGI_URL=http://127.0.0.1:$INFERENCE_PORT
+  --env SAFETY_MODEL=$SAFETY_MODEL
+  --env TGI_SAFETY_URL=http://127.0.0.1:$SAFETY_PORT
 ```
--- a/docs/source/getting_started/distributions/self_hosted_distro/together.md
+++ b/docs/source/getting_started/distributions/self_hosted_distro/together.md
@ -1,62 +1,67 @@
-# Together Distribution
-
-### Connect to a Llama Stack Together Endpoint
- You may connect to a hosted endpoint `https://llama-stack.together.ai`, serving a Llama Stack distribution
+# Fireworks Distribution

 The `llamastack/distribution-together` distribution consists of the following provider configurations.

-
-| **API**         	| **Inference** 	| **Agents**     	| **Memory**                                       	| **Safety**     	| **Telemetry**  	|
-|-----------------	|---------------	|----------------	|--------------------------------------------------	|----------------	|----------------	|
-| **Provider(s)** 	| remote::together   	| meta-reference 	| meta-reference, remote::weaviate 	| meta-reference 	| meta-reference 	|
+| API | Provider(s) |
+|-----|-------------|
+| agents | `inline::meta-reference` |
+| inference | `remote::together` |
+| memory | `inline::faiss`, `remote::chromadb`, `remote::pgvector` |
+| safety | `inline::llama-guard` |
+| telemetry | `inline::meta-reference` |


-### Docker: Start the Distribution (Single Node CPU)
+### Environment Variables

-> [!NOTE]
-> This assumes you have an hosted endpoint at Together with API Key.
+The following environment variables can be configured:

-```
-$ cd distributions/together && docker compose up
+- `LLAMASTACK_PORT`: Port for the Llama Stack distribution server (default: `5001`)
+- `TOGETHER_API_KEY`: Together.AI API Key (default: ``)
+
+### Models
+
+The following models are available by default:
+
+- `meta-llama/Llama-3.1-8B-Instruct`
+- `meta-llama/Llama-3.1-70B-Instruct`
+- `meta-llama/Llama-3.1-405B-Instruct-FP8`
+- `meta-llama/Llama-3.2-3B-Instruct`
+- `meta-llama/Llama-3.2-11B-Vision-Instruct`
+- `meta-llama/Llama-3.2-90B-Vision-Instruct`
+- `meta-llama/Llama-Guard-3-8B`
+- `meta-llama/Llama-Guard-3-11B-Vision`
+
+
+### Prerequisite: API Keys
+
+Make sure you have access to a Together API Key. You can get one by visiting [together.xyz](https://together.xyz/).
+
+
+## Running Llama Stack with Together
+
+You can do this via Conda (build code) or Docker which has a pre-built image.
+
+### Via Docker
+
+This method allows you to get started quickly without having to build the distribution code.
+
+```bash
+LLAMA_STACK_PORT=5001
+docker run \
+  -it \
+  -p $LLAMA_STACK_PORT:$LLAMA_STACK_PORT \
+  -v ./run.yaml:/root/my-run.yaml \
+  llamastack/distribution-together \
+  --yaml-config /root/my-run.yaml \
+  --port $LLAMA_STACK_PORT \
+  --env TOGETHER_API_KEY=$TOGETHER_API_KEY
 ```

-Make sure in your `run.yaml` file, your inference provider is pointing to the correct Together URL server endpoint. E.g.
-```
-inference:
-  - provider_id: together
-    provider_type: remote::together
-    config:
-      url: https://api.together.xyz/v1
-      api_key: <optional api key>
-```
-
-### Conda llama stack run (Single Node CPU)
+### Via Conda

 ```bash
 llama stack build --template together --image-type conda
-# -- modify run.yaml to a valid Together server endpoint
-llama stack run ./run.yaml
-```
-
-### (Optional) Update Model Serving Configuration
-
-Use `llama-stack-client models list` to check the available models served by together.
-
-```
-$ llama-stack-client models list
-+------------------------------+------------------------------+---------------+------------+
-| identifier                   | llama_model                  | provider_id   | metadata   |
-+==============================+==============================+===============+============+
-| Llama3.1-8B-Instruct         | Llama3.1-8B-Instruct         | together0     | {}         |
-+------------------------------+------------------------------+---------------+------------+
-| Llama3.1-70B-Instruct        | Llama3.1-70B-Instruct        | together0     | {}         |
-+------------------------------+------------------------------+---------------+------------+
-| Llama3.1-405B-Instruct       | Llama3.1-405B-Instruct       | together0     | {}         |
-+------------------------------+------------------------------+---------------+------------+
-| Llama3.2-3B-Instruct         | Llama3.2-3B-Instruct         | together0     | {}         |
-+------------------------------+------------------------------+---------------+------------+
-| Llama3.2-11B-Vision-Instruct | Llama3.2-11B-Vision-Instruct | together0     | {}         |
-+------------------------------+------------------------------+---------------+------------+
-| Llama3.2-90B-Vision-Instruct | Llama3.2-90B-Vision-Instruct | together0     | {}         |
-+------------------------------+------------------------------+---------------+------------+
+llama stack run ./run.yaml \
+  --port 5001 \
+  --env TOGETHER_API_KEY=$TOGETHER_API_KEY
 ```
--- a/docs/source/index.md
+++ b/docs/source/index.md
@ -74,7 +74,7 @@ A Distribution is where APIs and Providers are assembled together to provide a c
 | Python |  [llama-stack-client-python](https://github.com/meta-llama/llama-stack-client-python) | [![PyPI version](https://img.shields.io/pypi/v/llama_stack_client.svg)](https://pypi.org/project/llama_stack_client/)
 | Swift  | [llama-stack-client-swift](https://github.com/meta-llama/llama-stack-client-swift) | [![Swift Package Index](https://img.shields.io/endpoint?url=https%3A%2F%2Fswiftpackageindex.com%2Fapi%2Fpackages%2Fmeta-llama%2Fllama-stack-client-swift%2Fbadge%3Ftype%3Dswift-versions)](https://swiftpackageindex.com/meta-llama/llama-stack-client-swift)
 | Node   | [llama-stack-client-node](https://github.com/meta-llama/llama-stack-client-node) | [![NPM version](https://img.shields.io/npm/v/llama-stack-client.svg)](https://npmjs.org/package/llama-stack-client)
-| Kotlin | [llama-stack-client-kotlin](https://github.com/meta-llama/llama-stack-client-kotlin) |
+| Kotlin | [llama-stack-client-kotlin](https://github.com/meta-llama/llama-stack-client-kotlin) | [![Maven version](https://img.shields.io/maven-central/v/com.llama.llamastack/llama-stack-client-kotlin)](https://central.sonatype.com/artifact/com.llama.llamastack/llama-stack-client-kotlin)

 Check out our client SDKs for connecting to Llama Stack server in your preferred language, you can choose from [python](https://github.com/meta-llama/llama-stack-client-python), [node](https://github.com/meta-llama/llama-stack-client-node), [swift](https://github.com/meta-llama/llama-stack-client-swift), and [kotlin](https://github.com/meta-llama/llama-stack-client-kotlin) programming languages to quickly build your applications.

--- a/llama_stack/apis/batch_inference/batch_inference.py
+++ b/llama_stack/apis/batch_inference/batch_inference.py
@ -49,7 +49,7 @@ class BatchChatCompletionResponse(BaseModel):

@runtime_checkable
 class BatchInference(Protocol):
-    @webmethod(route="/batch_inference/completion")
+    @webmethod(route="/batch-inference/completion")
    async def batch_completion(
        self,
        model: str,
@ -58,7 +58,7 @@ class BatchInference(Protocol):
        logprobs: Optional[LogProbConfig] = None,
    ) -> BatchCompletionResponse: ...

-    @webmethod(route="/batch_inference/chat_completion")
+    @webmethod(route="/batch-inference/chat-completion")
    async def batch_chat_completion(
        self,
        model: str,
--- a/llama_stack/apis/datasetio/datasetio.py
+++ b/llama_stack/apis/datasetio/datasetio.py
@ -29,7 +29,7 @@ class DatasetIO(Protocol):
    # keeping for aligning with inference/safety, but this is not used
    dataset_store: DatasetStore

-    @webmethod(route="/datasetio/get_rows_paginated", method="GET")
+    @webmethod(route="/datasetio/get-rows-paginated", method="GET")
    async def get_rows_paginated(
        self,
        dataset_id: str,
--- a/llama_stack/apis/eval/eval.py
+++ b/llama_stack/apis/eval/eval.py
@ -74,14 +74,14 @@ class EvaluateResponse(BaseModel):


 class Eval(Protocol):
-    @webmethod(route="/eval/run_eval", method="POST")
+    @webmethod(route="/eval/run-eval", method="POST")
    async def run_eval(
        self,
        task_id: str,
        task_config: EvalTaskConfig,
    ) -> Job: ...

-    @webmethod(route="/eval/evaluate_rows", method="POST")
+    @webmethod(route="/eval/evaluate-rows", method="POST")
    async def evaluate_rows(
        self,
        task_id: str,
--- a/llama_stack/apis/eval_tasks/eval_tasks.py
+++ b/llama_stack/apis/eval_tasks/eval_tasks.py
@ -42,13 +42,13 @@ class EvalTaskInput(CommonEvalTaskFields, BaseModel):

@runtime_checkable
 class EvalTasks(Protocol):
-    @webmethod(route="/eval_tasks/list", method="GET")
+    @webmethod(route="/eval-tasks/list", method="GET")
    async def list_eval_tasks(self) -> List[EvalTask]: ...

-    @webmethod(route="/eval_tasks/get", method="GET")
+    @webmethod(route="/eval-tasks/get", method="GET")
    async def get_eval_task(self, name: str) -> Optional[EvalTask]: ...

-    @webmethod(route="/eval_tasks/register", method="POST")
+    @webmethod(route="/eval-tasks/register", method="POST")
    async def register_eval_task(
        self,
        eval_task_id: str,
--- a/llama_stack/apis/inference/inference.py
+++ b/llama_stack/apis/inference/inference.py
@ -234,7 +234,7 @@ class Inference(Protocol):
        logprobs: Optional[LogProbConfig] = None,
    ) -> Union[CompletionResponse, AsyncIterator[CompletionResponseStreamChunk]]: ...

-    @webmethod(route="/inference/chat_completion")
+    @webmethod(route="/inference/chat-completion")
    async def chat_completion(
        self,
        model_id: str,
--- a/llama_stack/apis/memory_banks/memory_banks.py
+++ b/llama_stack/apis/memory_banks/memory_banks.py
@ -130,13 +130,13 @@ class MemoryBankInput(BaseModel):

@runtime_checkable
 class MemoryBanks(Protocol):
-    @webmethod(route="/memory_banks/list", method="GET")
+    @webmethod(route="/memory-banks/list", method="GET")
    async def list_memory_banks(self) -> List[MemoryBank]: ...

-    @webmethod(route="/memory_banks/get", method="GET")
+    @webmethod(route="/memory-banks/get", method="GET")
    async def get_memory_bank(self, memory_bank_id: str) -> Optional[MemoryBank]: ...

-    @webmethod(route="/memory_banks/register", method="POST")
+    @webmethod(route="/memory-banks/register", method="POST")
    async def register_memory_bank(
        self,
        memory_bank_id: str,
@ -145,5 +145,5 @@ class MemoryBanks(Protocol):
        provider_memory_bank_id: Optional[str] = None,
    ) -> MemoryBank: ...

-    @webmethod(route="/memory_banks/unregister", method="POST")
+    @webmethod(route="/memory-banks/unregister", method="POST")
    async def unregister_memory_bank(self, memory_bank_id: str) -> None: ...
--- a/llama_stack/apis/models/models.py
+++ b/llama_stack/apis/models/models.py
@ -31,6 +31,8 @@ class Model(CommonModelFields, Resource):
    def provider_model_id(self) -> str:
        return self.provider_resource_id

+    model_config = ConfigDict(protected_namespaces=())
+

 class ModelInput(CommonModelFields):
    model_id: str
--- a/llama_stack/apis/post_training/post_training.py
+++ b/llama_stack/apis/post_training/post_training.py
@ -176,7 +176,7 @@ class PostTrainingJobArtifactsResponse(BaseModel):


 class PostTraining(Protocol):
-    @webmethod(route="/post_training/supervised_fine_tune")
+    @webmethod(route="/post-training/supervised-fine-tune")
    def supervised_fine_tune(
        self,
        job_uuid: str,
@ -193,7 +193,7 @@ class PostTraining(Protocol):
        logger_config: Dict[str, Any],
    ) -> PostTrainingJob: ...

-    @webmethod(route="/post_training/preference_optimize")
+    @webmethod(route="/post-training/preference-optimize")
    def preference_optimize(
        self,
        job_uuid: str,
@ -208,22 +208,22 @@ class PostTraining(Protocol):
        logger_config: Dict[str, Any],
    ) -> PostTrainingJob: ...

-    @webmethod(route="/post_training/jobs")
+    @webmethod(route="/post-training/jobs")
    def get_training_jobs(self) -> List[PostTrainingJob]: ...

    # sends SSE stream of logs
-    @webmethod(route="/post_training/job/logs")
+    @webmethod(route="/post-training/job/logs")
    def get_training_job_logstream(self, job_uuid: str) -> PostTrainingJobLogStream: ...

-    @webmethod(route="/post_training/job/status")
+    @webmethod(route="/post-training/job/status")
    def get_training_job_status(
        self, job_uuid: str
    ) -> PostTrainingJobStatusResponse: ...

-    @webmethod(route="/post_training/job/cancel")
+    @webmethod(route="/post-training/job/cancel")
    def cancel_training_job(self, job_uuid: str) -> None: ...

-    @webmethod(route="/post_training/job/artifacts")
+    @webmethod(route="/post-training/job/artifacts")
    def get_training_job_artifacts(
        self, job_uuid: str
    ) -> PostTrainingJobArtifactsResponse: ...
--- a/llama_stack/apis/safety/safety.py
+++ b/llama_stack/apis/safety/safety.py
@ -46,7 +46,7 @@ class ShieldStore(Protocol):
 class Safety(Protocol):
    shield_store: ShieldStore

-    @webmethod(route="/safety/run_shield")
+    @webmethod(route="/safety/run-shield")
    async def run_shield(
        self,
        shield_id: str,
--- a/llama_stack/apis/scoring/scoring.py
+++ b/llama_stack/apis/scoring/scoring.py
@ -44,7 +44,7 @@ class ScoringFunctionStore(Protocol):
 class Scoring(Protocol):
    scoring_function_store: ScoringFunctionStore

-    @webmethod(route="/scoring/score_batch")
+    @webmethod(route="/scoring/score-batch")
    async def score_batch(
        self,
        dataset_id: str,
--- a/llama_stack/apis/scoring_functions/scoring_functions.py
+++ b/llama_stack/apis/scoring_functions/scoring_functions.py
@ -104,13 +104,13 @@ class ScoringFnInput(CommonScoringFnFields, BaseModel):

@runtime_checkable
 class ScoringFunctions(Protocol):
-    @webmethod(route="/scoring_functions/list", method="GET")
+    @webmethod(route="/scoring-functions/list", method="GET")
    async def list_scoring_functions(self) -> List[ScoringFn]: ...

-    @webmethod(route="/scoring_functions/get", method="GET")
+    @webmethod(route="/scoring-functions/get", method="GET")
    async def get_scoring_function(self, scoring_fn_id: str) -> Optional[ScoringFn]: ...

-    @webmethod(route="/scoring_functions/register", method="POST")
+    @webmethod(route="/scoring-functions/register", method="POST")
    async def register_scoring_function(
        self,
        scoring_fn_id: str,
--- a/llama_stack/apis/synthetic_data_generation/synthetic_data_generation.py
+++ b/llama_stack/apis/synthetic_data_generation/synthetic_data_generation.py
@ -44,7 +44,7 @@ class SyntheticDataGenerationResponse(BaseModel):


 class SyntheticDataGeneration(Protocol):
-    @webmethod(route="/synthetic_data_generation/generate")
+    @webmethod(route="/synthetic-data-generation/generate")
    def synthetic_data_generate(
        self,
        dialogs: List[Message],
--- a/llama_stack/apis/telemetry/telemetry.py
+++ b/llama_stack/apis/telemetry/telemetry.py
@ -125,8 +125,8 @@ Event = Annotated[

@runtime_checkable
 class Telemetry(Protocol):
-    @webmethod(route="/telemetry/log_event")
+    @webmethod(route="/telemetry/log-event")
    async def log_event(self, event: Event) -> None: ...

-    @webmethod(route="/telemetry/get_trace", method="GET")
+    @webmethod(route="/telemetry/get-trace", method="GET")
    async def get_trace(self, trace_id: str) -> Trace: ...
--- a/llama_stack/apis/version.py
+++ b/llama_stack/apis/version.py
@ -0,0 +1,7 @@
+# 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.
+
+LLAMA_STACK_API_VERSION = "alpha"
--- a/llama_stack/cli/download.py
+++ b/llama_stack/cli/download.py
@ -19,7 +19,7 @@ import httpx

 from llama_models.datatypes import Model
 from llama_models.sku_list import LlamaDownloadInfo
-from pydantic import BaseModel
+from pydantic import BaseModel, ConfigDict

 from rich.console import Console
 from rich.progress import (
@ -293,8 +293,8 @@ class ParallelDownloader:

            if free_space < required_space:
                self.console.print(
-                    f"[red]Not enough disk space. Required: {required_space // (1024*1024)} MB, "
-                    f"Available: {free_space // (1024*1024)} MB[/red]"
+                    f"[red]Not enough disk space. Required: {required_space // (1024 * 1024)} MB, "
+                    f"Available: {free_space // (1024 * 1024)} MB[/red]"
                )
                return False
            return True
@ -413,8 +413,7 @@ class ModelEntry(BaseModel):
    model_id: str
    files: Dict[str, str]

-    class Config:
-        protected_namespaces = ()
+    model_config = ConfigDict(protected_namespaces=())


 class Manifest(BaseModel):
--- a/llama_stack/cli/stack/build.py
+++ b/llama_stack/cli/stack/build.py
@ -8,10 +8,14 @@ import argparse

 from llama_stack.cli.subcommand import Subcommand
 from llama_stack.distribution.datatypes import *  # noqa: F403
+import importlib
 import os
+import shutil
 from functools import lru_cache
 from pathlib import Path

+import pkg_resources
+
 from llama_stack.distribution.distribution import get_provider_registry
 from llama_stack.distribution.utils.dynamic import instantiate_class_type

@ -99,7 +103,9 @@ class StackBuild(Subcommand):
                        self.parser.error(
                            f"Please specify a image-type (docker | conda) for {args.template}"
                        )
-                    self._run_stack_build_command_from_build_config(build_config)
+                    self._run_stack_build_command_from_build_config(
+                        build_config, template_name=args.template
+                    )
                    return

            self.parser.error(
@ -193,7 +199,6 @@ class StackBuild(Subcommand):

        apis = list(build_config.distribution_spec.providers.keys())
        run_config = StackRunConfig(
-            built_at=datetime.now(),
            docker_image=(
                build_config.name
                if build_config.image_type == ImageType.docker.value
@ -217,15 +222,23 @@ class StackBuild(Subcommand):
                provider_types = [provider_types]

            for i, provider_type in enumerate(provider_types):
-                p_spec = Provider(
-                    provider_id=f"{provider_type}-{i}",
-                    provider_type=provider_type,
-                    config={},
-                )
+                pid = provider_type.split("::")[-1]
+
                config_type = instantiate_class_type(
                    provider_registry[Api(api)][provider_type].config_class
                )
-                p_spec.config = config_type()
+                if hasattr(config_type, "sample_run_config"):
+                    config = config_type.sample_run_config(
+                        __distro_dir__=f"distributions/{build_config.name}"
+                    )
+                else:
+                    config = {}
+
+                p_spec = Provider(
+                    provider_id=f"{pid}-{i}" if len(provider_types) > 1 else pid,
+                    provider_type=provider_type,
+                    config=config,
+                )
                run_config.providers[api].append(p_spec)

        os.makedirs(build_dir, exist_ok=True)
@ -241,12 +254,13 @@ class StackBuild(Subcommand):
        )

    def _run_stack_build_command_from_build_config(
-        self, build_config: BuildConfig
+        self, build_config: BuildConfig, template_name: Optional[str] = None
    ) -> None:
        import json
        import os

        import yaml
+        from termcolor import cprint

        from llama_stack.distribution.build import build_image
        from llama_stack.distribution.utils.config_dirs import DISTRIBS_BASE_DIR
@ -264,7 +278,29 @@ class StackBuild(Subcommand):
        if return_code != 0:
            return

-        self._generate_run_config(build_config, build_dir)
+        if template_name:
+            # copy run.yaml from template to build_dir instead of generating it again
+            template_path = pkg_resources.resource_filename(
+                "llama_stack", f"templates/{template_name}/run.yaml"
+            )
+            os.makedirs(build_dir, exist_ok=True)
+            run_config_file = build_dir / f"{build_config.name}-run.yaml"
+            shutil.copy(template_path, run_config_file)
+            module_name = f"llama_stack.templates.{template_name}"
+            module = importlib.import_module(module_name)
+            distribution_template = module.get_distribution_template()
+            cprint("Build Successful! Next steps: ", color="green")
+            env_vars = ", ".join(distribution_template.run_config_env_vars.keys())
+            cprint(
+                f"   1. Set the environment variables: {env_vars}",
+                color="green",
+            )
+            cprint(
+                f"   2. `llama stack run {run_config_file}`",
+                color="green",
+            )
+        else:
+            self._generate_run_config(build_config, build_dir)

    def _run_template_list_cmd(self, args: argparse.Namespace) -> None:
        import json
--- a/llama_stack/cli/stack/run.py
+++ b/llama_stack/cli/stack/run.py
@ -39,6 +39,13 @@ class StackRun(Subcommand):
            help="Disable IPv6 support",
            default=False,
        )
+        self.parser.add_argument(
+            "--env",
+            action="append",
+            help="Environment variables to pass to the server in KEY=VALUE format. Can be specified multiple times.",
+            default=[],
+            metavar="KEY=VALUE",
+        )

    def _run_stack_run_cmd(self, args: argparse.Namespace) -> None:
        from pathlib import Path
@ -108,4 +115,16 @@ class StackRun(Subcommand):
        if args.disable_ipv6:
            run_args.append("--disable-ipv6")

+        for env_var in args.env:
+            if "=" not in env_var:
+                self.parser.error(
+                    f"Environment variable '{env_var}' must be in KEY=VALUE format"
+                )
+                return
+            key, value = env_var.split("=", 1)  # split on first = only
+            if not key:
+                self.parser.error(f"Environment variable '{env_var}' has empty key")
+                return
+            run_args.extend(["--env", f"{key}={value}"])
+
        run_with_pty(run_args)
--- a/llama_stack/distribution/build_container.sh
+++ b/llama_stack/distribution/build_container.sh
@ -146,6 +146,8 @@ fi
 # Set version tag based on PyPI version
 if [ -n "$TEST_PYPI_VERSION" ]; then
  version_tag="test-$TEST_PYPI_VERSION"
+elif [[ -n "$LLAMA_STACK_DIR" || -n "$LLAMA_MODELS_DIR" ]]; then
+  version_tag="dev"
 else
  URL="https://pypi.org/pypi/llama-stack/json"
  version_tag=$(curl -s $URL | jq -r '.info.version')
--- a/llama_stack/distribution/configure.py
+++ b/llama_stack/distribution/configure.py
@ -186,6 +186,5 @@ def parse_and_maybe_upgrade_config(config_dict: Dict[str, Any]) -> StackRunConfi
        config_dict = upgrade_from_routing_table(config_dict)

    config_dict["version"] = LLAMA_STACK_RUN_CONFIG_VERSION
-    config_dict["built_at"] = datetime.now().isoformat()

    return StackRunConfig(**config_dict)
--- a/llama_stack/distribution/datatypes.py
+++ b/llama_stack/distribution/datatypes.py
@ -4,8 +4,6 @@
 # This source code is licensed under the terms described in the LICENSE file in
 # the root directory of this source tree.

-from datetime import datetime
-
 from typing import Dict, List, Optional, Union

 from pydantic import BaseModel, Field
@ -115,7 +113,6 @@ class Provider(BaseModel):

 class StackRunConfig(BaseModel):
    version: str = LLAMA_STACK_RUN_CONFIG_VERSION
-    built_at: datetime

    image_name: str = Field(
        ...,
--- a/llama_stack/distribution/server/server.py
+++ b/llama_stack/distribution/server/server.py
@ -4,12 +4,12 @@
 # This source code is licensed under the terms described in the LICENSE file in
 # the root directory of this source tree.

+import argparse
 import asyncio
 import functools
 import inspect
 import json
 import os
-import re
 import signal
 import sys
 import traceback
@ -19,7 +19,6 @@ from contextlib import asynccontextmanager
 from ssl import SSLError
 from typing import Any, Dict, Optional

-import fire
 import httpx
 import yaml

@ -41,7 +40,11 @@ from llama_stack.providers.utils.telemetry.tracing import (
 from llama_stack.distribution.datatypes import *  # noqa: F403
 from llama_stack.distribution.request_headers import set_request_provider_data
 from llama_stack.distribution.resolver import InvalidProviderError
-from llama_stack.distribution.stack import construct_stack
+from llama_stack.distribution.stack import (
+    construct_stack,
+    replace_env_vars,
+    validate_env_pair,
+)

 from .endpoints import get_all_api_endpoints

@ -271,64 +274,36 @@ def create_dynamic_typed_route(func: Any, method: str):
    return endpoint


-class EnvVarError(Exception):
-    def __init__(self, var_name: str, path: str = ""):
-        self.var_name = var_name
-        self.path = path
-        super().__init__(
-            f"Environment variable '{var_name}' not set or empty{f' at {path}' if path else ''}"
-        )
+def main():
+    """Start the LlamaStack server."""
+    parser = argparse.ArgumentParser(description="Start the LlamaStack server.")
+    parser.add_argument(
+        "--yaml-config",
+        default="llamastack-run.yaml",
+        help="Path to YAML configuration file",
+    )
+    parser.add_argument("--port", type=int, default=5000, help="Port to listen on")
+    parser.add_argument(
+        "--disable-ipv6", action="store_true", help="Whether to disable IPv6 support"
+    )
+    parser.add_argument(
+        "--env",
+        action="append",
+        help="Environment variables in KEY=value format. Can be specified multiple times.",
+    )

-
-def replace_env_vars(config: Any, path: str = "") -> Any:
-    if isinstance(config, dict):
-        result = {}
-        for k, v in config.items():
+    args = parser.parse_args()
+    if args.env:
+        for env_pair in args.env:
            try:
-                result[k] = replace_env_vars(v, f"{path}.{k}" if path else k)
-            except EnvVarError as e:
-                raise EnvVarError(e.var_name, e.path) from None
-        return result
+                key, value = validate_env_pair(env_pair)
+                print(f"Setting CLI environment variable {key} => {value}")
+                os.environ[key] = value
+            except ValueError as e:
+                print(f"Error: {str(e)}")
+                sys.exit(1)

-    elif isinstance(config, list):
-        result = []
-        for i, v in enumerate(config):
-            try:
-                result.append(replace_env_vars(v, f"{path}[{i}]"))
-            except EnvVarError as e:
-                raise EnvVarError(e.var_name, e.path) from None
-        return result
-
-    elif isinstance(config, str):
-        pattern = r"\${env\.([A-Z0-9_]+)(?::([^}]*))?}"
-
-        def get_env_var(match):
-            env_var = match.group(1)
-            default_val = match.group(2)
-
-            value = os.environ.get(env_var)
-            if not value:
-                if default_val is None:
-                    raise EnvVarError(env_var, path)
-                else:
-                    value = default_val
-
-            return value
-
-        try:
-            return re.sub(pattern, get_env_var, config)
-        except EnvVarError as e:
-            raise EnvVarError(e.var_name, e.path) from None
-
-    return config
-
-
-def main(
-    yaml_config: str = "llamastack-run.yaml",
-    port: int = 5000,
-    disable_ipv6: bool = False,
-):
-    with open(yaml_config, "r") as fp:
+    with open(args.yaml_config, "r") as fp:
        config = replace_env_vars(yaml.safe_load(fp))
        config = StackRunConfig(**config)

@ -395,10 +370,10 @@ def main(

    # FYI this does not do hot-reloads

-    listen_host = ["::", "0.0.0.0"] if not disable_ipv6 else "0.0.0.0"
-    print(f"Listening on {listen_host}:{port}")
-    uvicorn.run(app, host=listen_host, port=port)
+    listen_host = ["::", "0.0.0.0"] if not args.disable_ipv6 else "0.0.0.0"
+    print(f"Listening on {listen_host}:{args.port}")
+    uvicorn.run(app, host=listen_host, port=args.port)


 if __name__ == "__main__":
-    fire.Fire(main)
+    main()
--- a/llama_stack/distribution/stack.py
+++ b/llama_stack/distribution/stack.py
@ -4,8 +4,13 @@
 # This source code is licensed under the terms described in the LICENSE file in
 # the root directory of this source tree.

+import os
+from pathlib import Path
 from typing import Any, Dict

+import pkg_resources
+import yaml
+
 from termcolor import colored

 from llama_models.llama3.api.datatypes import *  # noqa: F403
@ -35,6 +40,9 @@ from llama_stack.distribution.store.registry import create_dist_registry
 from llama_stack.providers.datatypes import Api


+LLAMA_STACK_API_VERSION = "alpha"
+
+
 class LlamaStack(
    MemoryBanks,
    Inference,
@ -92,6 +100,77 @@ async def register_resources(run_config: StackRunConfig, impls: Dict[Api, Any]):
    print("")


+class EnvVarError(Exception):
+    def __init__(self, var_name: str, path: str = ""):
+        self.var_name = var_name
+        self.path = path
+        super().__init__(
+            f"Environment variable '{var_name}' not set or empty{f' at {path}' if path else ''}"
+        )
+
+
+def replace_env_vars(config: Any, path: str = "") -> Any:
+    if isinstance(config, dict):
+        result = {}
+        for k, v in config.items():
+            try:
+                result[k] = replace_env_vars(v, f"{path}.{k}" if path else k)
+            except EnvVarError as e:
+                raise EnvVarError(e.var_name, e.path) from None
+        return result
+
+    elif isinstance(config, list):
+        result = []
+        for i, v in enumerate(config):
+            try:
+                result.append(replace_env_vars(v, f"{path}[{i}]"))
+            except EnvVarError as e:
+                raise EnvVarError(e.var_name, e.path) from None
+        return result
+
+    elif isinstance(config, str):
+        pattern = r"\${env\.([A-Z0-9_]+)(?::([^}]*))?}"
+
+        def get_env_var(match):
+            env_var = match.group(1)
+            default_val = match.group(2)
+
+            value = os.environ.get(env_var)
+            if not value:
+                if default_val is None:
+                    raise EnvVarError(env_var, path)
+                else:
+                    value = default_val
+
+            # expand "~" from the values
+            return os.path.expanduser(value)
+
+        try:
+            return re.sub(pattern, get_env_var, config)
+        except EnvVarError as e:
+            raise EnvVarError(e.var_name, e.path) from None
+
+    return config
+
+
+def validate_env_pair(env_pair: str) -> tuple[str, str]:
+    """Validate and split an environment variable key-value pair."""
+    try:
+        key, value = env_pair.split("=", 1)
+        key = key.strip()
+        if not key:
+            raise ValueError(f"Empty key in environment variable pair: {env_pair}")
+        if not all(c.isalnum() or c == "_" for c in key):
+            raise ValueError(
+                f"Key must contain only alphanumeric characters and underscores: {key}"
+            )
+        return key, value
+    except ValueError as e:
+        raise ValueError(
+            f"Invalid environment variable format '{env_pair}': {str(e)}. Expected format: KEY=value"
+        ) from e
+
+
 # Produces a stack of providers for the given run config. Not all APIs may be
 # asked for in the run config.
 async def construct_stack(
@ -105,3 +184,17 @@ async def construct_stack(
    )
    await register_resources(run_config, impls)
    return impls
+
+
+def get_stack_run_config_from_template(template: str) -> StackRunConfig:
+    template_path = pkg_resources.resource_filename(
+        "llama_stack", f"templates/{template}/run.yaml"
+    )
+
+    if not Path(template_path).exists():
+        raise ValueError(f"Template '{template}' not found at {template_path}")
+
+    with open(template_path) as f:
+        run_config = yaml.safe_load(f)
+
+    return StackRunConfig(**replace_env_vars(run_config))
--- a/llama_stack/distribution/start_conda_env.sh
+++ b/llama_stack/distribution/start_conda_env.sh
@ -33,10 +33,33 @@ shift
 port="$1"
 shift

+# Process environment variables from --env arguments
+env_vars=""
+while [[ $# -gt 0 ]]; do
+  case "$1" in
+  --env)
+
+    if [[ -n "$2" ]]; then
+      # collect environment variables so we can set them after activating the conda env
+      env_vars="$env_vars --env $2"
+      shift 2
+    else
+      echo -e "${RED}Error: --env requires a KEY=VALUE argument${NC}" >&2
+      exit 1
+    fi
+    ;;
+  *)
+    shift
+    ;;
+  esac
+done
+
 eval "$(conda shell.bash hook)"
 conda deactivate && conda activate "$env_name"

+set -x
 $CONDA_PREFIX/bin/python \
  -m llama_stack.distribution.server.server \
-  --yaml_config "$yaml_config" \
-  --port "$port" "$@"
+  --yaml-config "$yaml_config" \
+  --port "$port" \
+  $env_vars
--- a/llama_stack/distribution/start_container.sh
+++ b/llama_stack/distribution/start_container.sh
@ -31,7 +31,7 @@ if [ $# -lt 3 ]; then
 fi

 build_name="$1"
-docker_image="distribution-$build_name"
+docker_image="localhost/distribution-$build_name"
 shift

 yaml_config="$1"
@ -40,6 +40,26 @@ shift
 port="$1"
 shift

+# Process environment variables from --env arguments
+env_vars=""
+while [[ $# -gt 0 ]]; do
+    case "$1" in
+        --env)
+            echo "env = $2"
+            if [[ -n "$2" ]]; then
+                env_vars="$env_vars -e $2"
+                shift 2
+            else
+                echo -e "${RED}Error: --env requires a KEY=VALUE argument${NC}" >&2
+                exit 1
+            fi
+            ;;
+        *)
+            shift
+            ;;
+    esac
+done
+
 set -x

 if command -v selinuxenabled &> /dev/null && selinuxenabled; then
@ -59,15 +79,18 @@ 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"
 fi

 $DOCKER_BINARY run $DOCKER_OPTS -it \
  -p $port:$port \
+  $env_vars \
  -v "$yaml_config:/app/config.yaml" \
  $mounts \
  $docker_image:$version_tag \
  python -m llama_stack.distribution.server.server \
-  --yaml_config /app/config.yaml \
-  --port $port "$@"
+  --yaml-config /app/config.yaml \
+  --port "$port"
--- a/llama_stack/providers/inline/agents/meta_reference/config.py
+++ b/llama_stack/providers/inline/agents/meta_reference/config.py
@ -4,11 +4,22 @@
 # This source code is licensed under the terms described in the LICENSE file in
 # the root directory of this source tree.

-from pydantic import BaseModel, Field
+from typing import Any, Dict
+
+from pydantic import BaseModel

 from llama_stack.providers.utils.kvstore import KVStoreConfig
 from llama_stack.providers.utils.kvstore.config import SqliteKVStoreConfig


 class MetaReferenceAgentsImplConfig(BaseModel):
-    persistence_store: KVStoreConfig = Field(default=SqliteKVStoreConfig())
+    persistence_store: KVStoreConfig
+
+    @classmethod
+    def sample_run_config(cls, __distro_dir__: str) -> Dict[str, Any]:
+        return {
+            "persistence_store": SqliteKVStoreConfig.sample_run_config(
+                __distro_dir__=__distro_dir__,
+                db_name="agents_store.db",
+            )
+        }
--- a/llama_stack/providers/inline/eval/meta_reference/init.py
+++ b/llama_stack/providers/inline/eval/meta_reference/init.py
@ -22,6 +22,7 @@ async def get_provider_impl(
        deps[Api.datasets],
        deps[Api.scoring],
        deps[Api.inference],
+        deps[Api.agents],
    )
    await impl.initialize()
    return impl
--- a/llama_stack/providers/inline/eval/meta_reference/eval.py
+++ b/llama_stack/providers/inline/eval/meta_reference/eval.py
@ -9,6 +9,7 @@ from llama_models.llama3.api.datatypes import *  # noqa: F403
 from .....apis.common.job_types import Job
 from .....apis.eval.eval import Eval, EvalTaskConfig, EvaluateResponse, JobStatus
 from llama_stack.apis.common.type_system import *  # noqa: F403
+from llama_stack.apis.agents import Agents
 from llama_stack.apis.datasetio import DatasetIO
 from llama_stack.apis.datasets import Datasets
 from llama_stack.apis.eval_tasks import EvalTask
@ -39,12 +40,14 @@ class MetaReferenceEvalImpl(Eval, EvalTasksProtocolPrivate):
        datasets_api: Datasets,
        scoring_api: Scoring,
        inference_api: Inference,
+        agents_api: Agents,
    ) -> None:
        self.config = config
        self.datasetio_api = datasetio_api
        self.datasets_api = datasets_api
        self.scoring_api = scoring_api
        self.inference_api = inference_api
+        self.agents_api = agents_api

        # TODO: assume sync job, will need jobs API for async scheduling
        self.jobs = {}
@ -126,18 +129,50 @@ class MetaReferenceEvalImpl(Eval, EvalTasksProtocolPrivate):
        self.jobs[job_id] = res
        return Job(job_id=job_id)

-    async def evaluate_rows(
-        self,
-        task_id: str,
-        input_rows: List[Dict[str, Any]],
-        scoring_functions: List[str],
-        task_config: EvalTaskConfig,
-    ) -> EvaluateResponse:
+    async def _run_agent_generation(
+        self, input_rows: List[Dict[str, Any]], task_config: EvalTaskConfig
+    ) -> List[Dict[str, Any]]:
        candidate = task_config.eval_candidate
-        if candidate.type == "agent":
-            raise NotImplementedError(
-                "Evaluation with generation has not been implemented for agents"
+        create_response = await self.agents_api.create_agent(candidate.config)
+        agent_id = create_response.agent_id
+
+        generations = []
+        for i, x in tqdm(enumerate(input_rows)):
+            assert ColumnName.chat_completion_input.value in x, "Invalid input row"
+            input_messages = eval(str(x[ColumnName.chat_completion_input.value]))
+            input_messages = [UserMessage(**x) for x in input_messages]
+
+            # NOTE: only single-turn agent generation is supported. Create a new session for each input row
+            session_create_response = await self.agents_api.create_agent_session(
+                agent_id, f"session-{i}"
            )
+            session_id = session_create_response.session_id
+
+            turn_request = dict(
+                agent_id=agent_id,
+                session_id=session_id,
+                messages=input_messages,
+                stream=True,
+            )
+            turn_response = [
+                chunk
+                async for chunk in await self.agents_api.create_agent_turn(
+                    **turn_request
+                )
+            ]
+            final_event = turn_response[-1].event.payload
+            generations.append(
+                {
+                    ColumnName.generated_answer.value: final_event.turn.output_message.content
+                }
+            )
+
+        return generations
+
+    async def _run_model_generation(
+        self, input_rows: List[Dict[str, Any]], task_config: EvalTaskConfig
+    ) -> List[Dict[str, Any]]:
+        candidate = task_config.eval_candidate
        assert (
            candidate.sampling_params.max_tokens is not None
        ), "SamplingParams.max_tokens must be provided"
@ -179,6 +214,23 @@ class MetaReferenceEvalImpl(Eval, EvalTasksProtocolPrivate):
            else:
                raise ValueError("Invalid input row")

+        return generations
+
+    async def evaluate_rows(
+        self,
+        task_id: str,
+        input_rows: List[Dict[str, Any]],
+        scoring_functions: List[str],
+        task_config: EvalTaskConfig,
+    ) -> EvaluateResponse:
+        candidate = task_config.eval_candidate
+        if candidate.type == "agent":
+            generations = await self._run_agent_generation(input_rows, task_config)
+        elif candidate.type == "model":
+            generations = await self._run_model_generation(input_rows, task_config)
+        else:
+            raise ValueError(f"Invalid candidate type: {candidate.type}")
+
        # scoring with generated_answer
        score_input_rows = [
            input_r | generated_r
--- a/llama_stack/providers/inline/inference/meta_reference/config.py
+++ b/llama_stack/providers/inline/inference/meta_reference/config.py
@ -49,6 +49,18 @@ class MetaReferenceInferenceConfig(BaseModel):
        resolved = resolve_model(self.model)
        return resolved.pth_file_count

+    @classmethod
+    def sample_run_config(
+        cls,
+        model: str = "Llama3.2-3B-Instruct",
+        checkpoint_dir: str = "${env.CHECKPOINT_DIR:null}",
+    ) -> Dict[str, Any]:
+        return {
+            "model": model,
+            "max_seq_len": 4096,
+            "checkpoint_dir": checkpoint_dir,
+        }
+

 class MetaReferenceQuantizedInferenceConfig(MetaReferenceInferenceConfig):
    quantization: QuantizationConfig
--- a/llama_stack/providers/inline/inference/meta_reference/generation.py
+++ b/llama_stack/providers/inline/inference/meta_reference/generation.py
@ -107,7 +107,7 @@ class Llama:
            sys.stdout = open(os.devnull, "w")

        start_time = time.time()
-        if config.checkpoint_dir:
+        if config.checkpoint_dir and config.checkpoint_dir != "null":
            ckpt_dir = config.checkpoint_dir
        else:
            ckpt_dir = model_checkpoint_dir(model)
@ -137,7 +137,6 @@ class Llama:
        ), f"model_args vocab = {model_args.vocab_size} but tokenizer vocab = {tokenizer.n_words}"

        if isinstance(config, MetaReferenceQuantizedInferenceConfig):
-
            if isinstance(config.quantization, Fp8QuantizationConfig):
                from .quantization.loader import convert_to_fp8_quantized_model

--- a/llama_stack/providers/inline/inference/vllm/config.py
+++ b/llama_stack/providers/inline/inference/vllm/config.py
@ -34,6 +34,16 @@ class VLLMConfig(BaseModel):
        default=0.3,
    )

+    @classmethod
+    def sample_run_config(cls):
+        return {
+            "model": "${env.VLLM_INFERENCE_MODEL:Llama3.2-3B-Instruct}",
+            "tensor_parallel_size": "${env.VLLM_TENSOR_PARALLEL_SIZE:1}",
+            "max_tokens": "${env.VLLM_MAX_TOKENS:4096}",
+            "enforce_eager": "${env.VLLM_ENFORCE_EAGER:False}",
+            "gpu_memory_utilization": "${env.VLLM_GPU_MEMORY_UTILIZATION:0.3}",
+        }
+
    @field_validator("model")
    @classmethod
    def validate_model(cls, model: str) -> str:
--- a/llama_stack/providers/inline/memory/faiss/config.py
+++ b/llama_stack/providers/inline/memory/faiss/config.py
@ -4,10 +4,11 @@
 # This source code is licensed under the terms described in the LICENSE file in
 # the root directory of this source tree.

+from typing import Any, Dict
+
 from llama_models.schema_utils import json_schema_type
 from pydantic import BaseModel

-from llama_stack.distribution.utils.config_dirs import RUNTIME_BASE_DIR
 from llama_stack.providers.utils.kvstore.config import (
    KVStoreConfig,
    SqliteKVStoreConfig,
@ -16,6 +17,13 @@ from llama_stack.providers.utils.kvstore.config import (

@json_schema_type
 class FaissImplConfig(BaseModel):
-    kvstore: KVStoreConfig = SqliteKVStoreConfig(
-        db_path=(RUNTIME_BASE_DIR / "faiss_store.db").as_posix()
-    )  # Uses SQLite config specific to FAISS storage
+    kvstore: KVStoreConfig
+
+    @classmethod
+    def sample_run_config(cls, __distro_dir__: str) -> Dict[str, Any]:
+        return {
+            "kvstore": SqliteKVStoreConfig.sample_run_config(
+                __distro_dir__=__distro_dir__,
+                db_name="faiss_store.db",
+            )
+        }
--- a/llama_stack/providers/inline/safety/llama_guard/llama_guard.py
+++ b/llama_stack/providers/inline/safety/llama_guard/llama_guard.py
@ -73,18 +73,21 @@ DEFAULT_LG_V3_SAFETY_CATEGORIES = [
    CAT_ELECTIONS,
 ]

-LLAMA_GUARD_MODEL_IDS = [
-    CoreModelId.llama_guard_3_8b.value,
-    CoreModelId.llama_guard_3_1b.value,
-    CoreModelId.llama_guard_3_11b_vision.value,
-]
+# accept both CoreModelId and huggingface repo id
+LLAMA_GUARD_MODEL_IDS = {
+    CoreModelId.llama_guard_3_8b.value: "meta-llama/Llama-Guard-3-8B",
+    "meta-llama/Llama-Guard-3-8B": "meta-llama/Llama-Guard-3-8B",
+    CoreModelId.llama_guard_3_1b.value: "meta-llama/Llama-Guard-3-1B",
+    "meta-llama/Llama-Guard-3-1B": "meta-llama/Llama-Guard-3-1B",
+    CoreModelId.llama_guard_3_11b_vision.value: "meta-llama/Llama-Guard-3-11B-Vision",
+    "meta-llama/Llama-Guard-3-11B-Vision": "meta-llama/Llama-Guard-3-11B-Vision",
+}

 MODEL_TO_SAFETY_CATEGORIES_MAP = {
-    CoreModelId.llama_guard_3_8b.value: (
-        DEFAULT_LG_V3_SAFETY_CATEGORIES + [CAT_CODE_INTERPRETER_ABUSE]
-    ),
-    CoreModelId.llama_guard_3_1b.value: DEFAULT_LG_V3_SAFETY_CATEGORIES,
-    CoreModelId.llama_guard_3_11b_vision.value: DEFAULT_LG_V3_SAFETY_CATEGORIES,
+    "meta-llama/Llama-Guard-3-8B": DEFAULT_LG_V3_SAFETY_CATEGORIES
+    + [CAT_CODE_INTERPRETER_ABUSE],
+    "meta-llama/Llama-Guard-3-1B": DEFAULT_LG_V3_SAFETY_CATEGORIES,
+    "meta-llama/Llama-Guard-3-11B-Vision": DEFAULT_LG_V3_SAFETY_CATEGORIES,
 }


@ -150,8 +153,9 @@ class LlamaGuardSafetyImpl(Safety, ShieldsProtocolPrivate):
        if len(messages) > 0 and messages[0].role != Role.user.value:
            messages[0] = UserMessage(content=messages[0].content)

+        model = LLAMA_GUARD_MODEL_IDS[shield.provider_resource_id]
        impl = LlamaGuardShield(
-            model=shield.provider_resource_id,
+            model=model,
            inference_api=self.inference_api,
            excluded_categories=self.config.excluded_categories,
        )
--- a/llama_stack/providers/inline/scoring/llm_as_judge/scoring_fn/fn_defs/llm_as_judge_405b_simpleqa.py
+++ b/llama_stack/providers/inline/scoring/llm_as_judge/scoring_fn/fn_defs/llm_as_judge_405b_simpleqa.py
@ -0,0 +1,91 @@
+# 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.
+
+from llama_stack.apis.common.type_system import NumberType
+from llama_stack.apis.scoring_functions import LLMAsJudgeScoringFnParams, ScoringFn
+
+GRADER_TEMPLATE = """
+Your job is to look at a question, a gold target, and a predicted answer, and then assign a grade of either ["CORRECT", "INCORRECT", "NOT_ATTEMPTED"].
+First, I will give examples of each grade, and then you will grade a new example.
+The following are examples of CORRECT predicted answers.
+```
+Question: What are the names of Barack Obama's children?
+Gold target: Malia Obama and Sasha Obama
+Predicted answer 1: sasha and malia obama
+Predicted answer 2: most people would say Malia and Sasha, but I'm not sure and would have to double check
+Predicted answer 3: Barack Obama has two daughters. Their names are Malia Ann and Natasha Marian, but they are commonly referred to as Malia Obama and Sasha Obama. Malia was born on July 4, 1998, and Sasha was born on June 10, 2001.
+```
+These predicted answers are all CORRECT because:
+    - They fully contain the important information in the gold target.
+    - They do not contain any information that contradicts the gold target.
+    - Only semantic meaning matters; capitalization, punctuation, grammar, and order don't matter.
+    - Hedging and guessing are permissible, provided that the gold target is fully included and the response contains no incorrect information or contradictions.
+The following are examples of INCORRECT predicted answers.
+```
+Question: What are the names of Barack Obama's children?
+Gold target: Malia and Sasha
+Predicted answer 1: Malia.
+Predicted answer 2: Malia, Sasha, and Susan.
+Predicted answer 3: Barack Obama does not have any children.
+Predicted answer 4: I think it's either Malia and Sasha. Or it could be Malia and Jackie. Or it could be Joey and Malia.
+Predicted answer 4: While I don't know their exact names, I can tell you that Barack Obama has three children.
+Predicted answer 5: It's possible you may mean Betsy and Olivia. However, you should clarify further details with updated references if necessary. Is that the correct answer?
+Predicted answer 6: It may be the case that Obama's child is named James. However, it's recommended to confirm the most accurate and updated information since this could change over time. This model may not always reflect the most current information.
+```
+These predicted answers are all INCORRECT because:
+    - A factual statement in the answer contradicts the gold target. Incorrect statements that have some hedging (e.g., "it is possible that", "although i'm not sure, i think") are also considered incorrect.
+The following are examples of NOT_ATTEMPTED predicted answers.
+```
+Question: What are the names of Barack Obama's children?
+Gold target: Malia and Sasha
+Predicted answer 1: I don't know.
+Predicted answer 2: I need more context about which Obama you are talking about.
+Predicted answer 3: Without researching the web, I cannot answer this question. However, I can tell you that Barack Obama has two children.
+Predicted answer 4: Barack Obama has two children. I know that one of them is Malia, but I'm not sure about the other one.
+```
+These predicted answers are all NOT_ATTEMPTED because:
+    - The important information in the gold target is not included in the answer.
+    - No statements in the answer contradict the gold target.
+Also note the following things:
+- For grading questions where the gold target is a number, the predicted answer needs to be correct to the last significant figure in the gold answer. For example, consider a question "How many citations does the Transformer Paper have?" with gold target "120k".
+    - Predicted answers "120k", "124k", and 115k" are all CORRECT.
+    - Predicted answers "100k" and "113k" are INCORRECT.
+    - Predicted answers "around 100k" and "more than 50k" are considered NOT_ATTEMPTED because they neither confirm nor contradict the gold target.
+- The gold target may contain more information than the question. In such cases, the predicted answer only needs to contain the information that is in the question.
+    - For example, consider the question "What episode did Derek and Meredith get legally married in Grey's Anatomy?" with gold target "Season 7, Episode 20: White Wedding". Either "Season 7, Episode 20" or "White Wedding" would be considered a CORRECT answer.
+- Do not punish predicted answers if they omit information that would be clearly inferred from the question.
+    - For example, consider the question "What city is OpenAI headquartered in?" and the gold target "San Francisco, California". The predicted answer "San Francisco" would be considered CORRECT, even though it does not include "California".
+    - Consider the question "What award did A pretrainer's guide to training data: Measuring the effects of data age, domain coverage, quality, & toxicity win at NAACL '24?", the gold target is "Outstanding Paper Award". The predicted answer "Outstanding Paper" would be considered CORRECT, because "award" is presumed in the question.
+    - For the question "What is the height of Jason Wei in meters?", the gold target is "1.73 m". The predicted answer "1.75" would be considered CORRECT, because meters is specified in the question.
+    - For the question "What is the name of Barack Obama's wife?", the gold target is "Michelle Obama". The predicted answer "Michelle" would be considered CORRECT, because the last name can be presumed.
+- Do not punish for typos in people's name if it's clearly the same name.
+    - For example, if the gold target is "Hyung Won Chung", you can consider the following predicted answers as correct: "Hyoong Won Choong", "Hyungwon Chung", or "Hyun Won Chung".
+Here is a new example. Simply reply with either CORRECT, INCORRECT, NOT ATTEMPTED. Don't apologize or correct yourself if there was a mistake; we are just trying to grade the answer.
+```
+Question: {input_query}
+Gold target: {expected_answer}
+Predicted answer: {generated_answer}
+```
+Grade the predicted answer of this new question as one of:
+A: CORRECT
+B: INCORRECT
+C: NOT_ATTEMPTED
+Just return the letters "A", "B", or "C", with no text around it.
+""".strip()
+
+
+llm_as_judge_405b_simpleqa = ScoringFn(
+    identifier="llm-as-judge::405b-simpleqa",
+    description="Llm As Judge Scoring Function for SimpleQA Benchmark (https://github.com/openai/simple-evals/blob/main/simpleqa_eval.py)",
+    return_type=NumberType(),
+    provider_id="llm-as-judge",
+    provider_resource_id="llm-as-judge-405b-simpleqa",
+    params=LLMAsJudgeScoringFnParams(
+        judge_model="Llama3.1-405B-Instruct",
+        prompt_template=GRADER_TEMPLATE,
+        judge_score_regexes=[r"(A|B|C)"],
+    ),
+)
--- a/llama_stack/providers/inline/scoring/llm_as_judge/scoring_fn/fn_defs/llm_as_judge_base.py
+++ b/llama_stack/providers/inline/scoring/llm_as_judge/scoring_fn/fn_defs/llm_as_judge_base.py
@ -9,7 +9,7 @@ from llama_stack.apis.scoring_functions import ScoringFn


 llm_as_judge_base = ScoringFn(
-    identifier="llm-as-judge::llm_as_judge_base",
+    identifier="llm-as-judge::base",
    description="Llm As Judge Scoring Function",
    return_type=NumberType(),
    provider_id="llm-as-judge",
--- a/llama_stack/providers/inline/scoring/llm_as_judge/scoring_fn/llm_as_judge_scoring_fn.py
+++ b/llama_stack/providers/inline/scoring/llm_as_judge/scoring_fn/llm_as_judge_scoring_fn.py
@ -11,6 +11,8 @@ from llama_stack.apis.scoring import *  # noqa: F401, F403
 from llama_stack.apis.common.type_system import *  # noqa: F403
 import re

+from .fn_defs.llm_as_judge_405b_simpleqa import llm_as_judge_405b_simpleqa
+
 from .fn_defs.llm_as_judge_base import llm_as_judge_base


@ -24,6 +26,7 @@ class LlmAsJudgeScoringFn(BaseScoringFn):
        self.inference_api = inference_api
        self.supported_fn_defs_registry = {
            llm_as_judge_base.identifier: llm_as_judge_base,
+            llm_as_judge_405b_simpleqa.identifier: llm_as_judge_405b_simpleqa,
        }

    async def score_row(
--- a/llama_stack/providers/registry/eval.py
+++ b/llama_stack/providers/registry/eval.py
@ -22,6 +22,7 @@ def available_providers() -> List[ProviderSpec]:
                Api.datasets,
                Api.scoring,
                Api.inference,
+                Api.agents,
            ],
        ),
    ]
--- a/llama_stack/providers/remote/inference/fireworks/config.py
+++ b/llama_stack/providers/remote/inference/fireworks/config.py
@ -4,7 +4,7 @@
 # This source code is licensed under the terms described in the LICENSE file in
 # the root directory of this source tree.

-from typing import Optional
+from typing import Any, Dict, Optional

 from llama_models.schema_utils import json_schema_type
 from pydantic import BaseModel, Field
@ -20,3 +20,10 @@ class FireworksImplConfig(BaseModel):
        default=None,
        description="The Fireworks.ai API Key",
    )
+
+    @classmethod
+    def sample_run_config(cls) -> Dict[str, Any]:
+        return {
+            "url": "https://api.fireworks.ai/inference",
+            "api_key": "${env.FIREWORKS_API_KEY}",
+        }
--- a/llama_stack/providers/remote/inference/fireworks/fireworks.py
+++ b/llama_stack/providers/remote/inference/fireworks/fireworks.py
@ -35,7 +35,7 @@ from llama_stack.providers.utils.inference.prompt_adapter import (
 from .config import FireworksImplConfig


-model_aliases = [
+MODEL_ALIASES = [
    build_model_alias(
        "fireworks/llama-v3p1-8b-instruct",
        CoreModelId.llama3_1_8b_instruct.value,
@ -79,7 +79,7 @@ class FireworksInferenceAdapter(
    ModelRegistryHelper, Inference, NeedsRequestProviderData
 ):
    def __init__(self, config: FireworksImplConfig) -> None:
-        ModelRegistryHelper.__init__(self, model_aliases)
+        ModelRegistryHelper.__init__(self, MODEL_ALIASES)
        self.config = config
        self.formatter = ChatFormat(Tokenizer.get_instance())

--- a/llama_stack/providers/remote/inference/nvidia/_nvidia.py
+++ b/llama_stack/providers/remote/inference/nvidia/_nvidia.py
@ -30,7 +30,7 @@ from llama_stack.apis.inference import (
    ResponseFormat,
 )
 from llama_stack.providers.utils.inference.model_registry import (
-    build_model_alias,
+    build_model_alias_with_just_provider_model_id,
    ModelRegistryHelper,
 )

@ -43,39 +43,39 @@ from ._openai_utils import (
 from ._utils import check_health

 _MODEL_ALIASES = [
-    build_model_alias(
+    build_model_alias_with_just_provider_model_id(
        "meta/llama3-8b-instruct",
        CoreModelId.llama3_8b_instruct.value,
    ),
-    build_model_alias(
+    build_model_alias_with_just_provider_model_id(
        "meta/llama3-70b-instruct",
        CoreModelId.llama3_70b_instruct.value,
    ),
-    build_model_alias(
+    build_model_alias_with_just_provider_model_id(
        "meta/llama-3.1-8b-instruct",
        CoreModelId.llama3_1_8b_instruct.value,
    ),
-    build_model_alias(
+    build_model_alias_with_just_provider_model_id(
        "meta/llama-3.1-70b-instruct",
        CoreModelId.llama3_1_70b_instruct.value,
    ),
-    build_model_alias(
+    build_model_alias_with_just_provider_model_id(
        "meta/llama-3.1-405b-instruct",
        CoreModelId.llama3_1_405b_instruct.value,
    ),
-    build_model_alias(
+    build_model_alias_with_just_provider_model_id(
        "meta/llama-3.2-1b-instruct",
        CoreModelId.llama3_2_1b_instruct.value,
    ),
-    build_model_alias(
+    build_model_alias_with_just_provider_model_id(
        "meta/llama-3.2-3b-instruct",
        CoreModelId.llama3_2_3b_instruct.value,
    ),
-    build_model_alias(
+    build_model_alias_with_just_provider_model_id(
        "meta/llama-3.2-11b-vision-instruct",
        CoreModelId.llama3_2_11b_vision_instruct.value,
    ),
-    build_model_alias(
+    build_model_alias_with_just_provider_model_id(
        "meta/llama-3.2-90b-vision-instruct",
        CoreModelId.llama3_2_90b_vision_instruct.value,
    ),
--- a/llama_stack/providers/remote/inference/ollama/init.py
+++ b/llama_stack/providers/remote/inference/ollama/init.py
@ -4,14 +4,10 @@
 # This source code is licensed under the terms described in the LICENSE file in
 # the root directory of this source tree.

-from llama_stack.distribution.datatypes import RemoteProviderConfig
+from .config import OllamaImplConfig


-class OllamaImplConfig(RemoteProviderConfig):
-    port: int = 11434
-
-
-async def get_adapter_impl(config: RemoteProviderConfig, _deps):
+async def get_adapter_impl(config: OllamaImplConfig, _deps):
    from .ollama import OllamaInferenceAdapter

    impl = OllamaInferenceAdapter(config.url)
--- a/llama_stack/providers/remote/inference/ollama/config.py
+++ b/llama_stack/providers/remote/inference/ollama/config.py
@ -0,0 +1,22 @@
+# 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.
+
+from typing import Any, Dict
+
+from pydantic import BaseModel
+
+
+DEFAULT_OLLAMA_URL = "http://localhost:11434"
+
+
+class OllamaImplConfig(BaseModel):
+    url: str = DEFAULT_OLLAMA_URL
+
+    @classmethod
+    def sample_run_config(
+        cls, url: str = "${env.OLLAMA_URL:http://localhost:11434}", **kwargs
+    ) -> Dict[str, Any]:
+        return {"url": url}
--- a/llama_stack/providers/remote/inference/ollama/ollama.py
+++ b/llama_stack/providers/remote/inference/ollama/ollama.py
@ -16,6 +16,7 @@ from ollama import AsyncClient

 from llama_stack.providers.utils.inference.model_registry import (
    build_model_alias,
+    build_model_alias_with_just_provider_model_id,
    ModelRegistryHelper,
 )

@ -44,10 +45,18 @@ model_aliases = [
        "llama3.1:8b-instruct-fp16",
        CoreModelId.llama3_1_8b_instruct.value,
    ),
+    build_model_alias_with_just_provider_model_id(
+        "llama3.1:8b",
+        CoreModelId.llama3_1_8b_instruct.value,
+    ),
    build_model_alias(
        "llama3.1:70b-instruct-fp16",
        CoreModelId.llama3_1_70b_instruct.value,
    ),
+    build_model_alias_with_just_provider_model_id(
+        "llama3.1:70b",
+        CoreModelId.llama3_1_70b_instruct.value,
+    ),
    build_model_alias(
        "llama3.2:1b-instruct-fp16",
        CoreModelId.llama3_2_1b_instruct.value,
@ -56,6 +65,24 @@ model_aliases = [
        "llama3.2:3b-instruct-fp16",
        CoreModelId.llama3_2_3b_instruct.value,
    ),
+    build_model_alias_with_just_provider_model_id(
+        "llama3.2:1b",
+        CoreModelId.llama3_2_1b_instruct.value,
+    ),
+    build_model_alias_with_just_provider_model_id(
+        "llama3.2:3b",
+        CoreModelId.llama3_2_3b_instruct.value,
+    ),
+    build_model_alias(
+        "llama3.2-vision:11b-instruct-fp16",
+        CoreModelId.llama3_2_11b_vision_instruct.value,
+    ),
+    build_model_alias_with_just_provider_model_id(
+        "llama3.2-vision",
+        CoreModelId.llama3_2_11b_vision_instruct.value,
+    ),
+    # The Llama Guard models don't have their full fp16 versions
+    # so we are going to alias their default version to the canonical SKU
    build_model_alias(
        "llama-guard3:8b",
        CoreModelId.llama_guard_3_8b.value,
@ -64,10 +91,6 @@ model_aliases = [
        "llama-guard3:1b",
        CoreModelId.llama_guard_3_1b.value,
    ),
-    build_model_alias(
-        "x/llama3.2-vision:11b-instruct-fp16",
-        CoreModelId.llama3_2_11b_vision_instruct.value,
-    ),
 ]


@ -82,7 +105,7 @@ class OllamaInferenceAdapter(Inference, ModelsProtocolPrivate):
        return AsyncClient(host=self.url)

    async def initialize(self) -> None:
-        print("Initializing Ollama, checking connectivity to server...")
+        print(f"checking connectivity to Ollama at `{self.url}`...")
        try:
            await self.client.ps()
        except httpx.ConnectError as e:
--- a/llama_stack/providers/remote/inference/tgi/config.py
+++ b/llama_stack/providers/remote/inference/tgi/config.py
@ -12,19 +12,20 @@ from pydantic import BaseModel, Field

@json_schema_type
 class TGIImplConfig(BaseModel):
-    host: str = "localhost"
-    port: int = 8080
-    protocol: str = "http"
-
-    @property
-    def url(self) -> str:
-        return f"{self.protocol}://{self.host}:{self.port}"
-
+    url: str = Field(
+        description="The URL for the TGI serving endpoint",
+    )
    api_token: Optional[str] = Field(
        default=None,
        description="A bearer token if your TGI endpoint is protected.",
    )

+    @classmethod
+    def sample_run_config(cls, url: str = "${env.TGI_URL}", **kwargs):
+        return {
+            "url": url,
+        }
+

@json_schema_type
 class InferenceEndpointImplConfig(BaseModel):
--- a/llama_stack/providers/remote/inference/together/config.py
+++ b/llama_stack/providers/remote/inference/together/config.py
@ -4,7 +4,7 @@
 # This source code is licensed under the terms described in the LICENSE file in
 # the root directory of this source tree.

-from typing import Optional
+from typing import Any, Dict, Optional

 from llama_models.schema_utils import json_schema_type
 from pydantic import BaseModel, Field
@ -20,3 +20,10 @@ class TogetherImplConfig(BaseModel):
        default=None,
        description="The Together AI API Key",
    )
+
+    @classmethod
+    def sample_run_config(cls, **kwargs) -> Dict[str, Any]:
+        return {
+            "url": "https://api.together.xyz/v1",
+            "api_key": "${env.TOGETHER_API_KEY}",
+        }
--- a/llama_stack/providers/remote/inference/together/together.py
+++ b/llama_stack/providers/remote/inference/together/together.py
@ -38,7 +38,7 @@ from llama_stack.providers.utils.inference.prompt_adapter import (
 from .config import TogetherImplConfig


-model_aliases = [
+MODEL_ALIASES = [
    build_model_alias(
        "meta-llama/Meta-Llama-3.1-8B-Instruct-Turbo",
        CoreModelId.llama3_1_8b_instruct.value,
@ -78,7 +78,7 @@ class TogetherInferenceAdapter(
    ModelRegistryHelper, Inference, NeedsRequestProviderData
 ):
    def __init__(self, config: TogetherImplConfig) -> None:
-        ModelRegistryHelper.__init__(self, model_aliases)
+        ModelRegistryHelper.__init__(self, MODEL_ALIASES)
        self.config = config
        self.formatter = ChatFormat(Tokenizer.get_instance())

--- a/llama_stack/providers/remote/inference/vllm/config.py
+++ b/llama_stack/providers/remote/inference/vllm/config.py
@ -24,3 +24,15 @@ class VLLMInferenceAdapterConfig(BaseModel):
        default="fake",
        description="The API token",
    )
+
+    @classmethod
+    def sample_run_config(
+        cls,
+        url: str = "${env.VLLM_URL}",
+        **kwargs,
+    ):
+        return {
+            "url": url,
+            "max_tokens": "${env.VLLM_MAX_TOKENS:4096}",
+            "api_token": "${env.VLLM_API_TOKEN:fake}",
+        }
--- a/llama_stack/providers/tests/README.md
+++ b/llama_stack/providers/tests/README.md
@ -44,7 +44,7 @@ Finally, you can override the model completely by doing:
 ```bash
 pytest -s -v llama_stack/providers/tests/inference/test_text_inference.py \
  -m fireworks \
-  --inference-model "Llama3.1-70B-Instruct" \
+  --inference-model "meta-llama/Llama3.1-70B-Instruct" \
  --env FIREWORKS_API_KEY=<...>
 ```

--- a/llama_stack/providers/tests/agents/conftest.py
+++ b/llama_stack/providers/tests/agents/conftest.py
@ -81,13 +81,13 @@ def pytest_addoption(parser):
    parser.addoption(
        "--inference-model",
        action="store",
-        default="Llama3.1-8B-Instruct",
+        default="meta-llama/Llama-3.1-8B-Instruct",
        help="Specify the inference model to use for testing",
    )
    parser.addoption(
        "--safety-shield",
        action="store",
-        default="Llama-Guard-3-8B",
+        default="meta-llama/Llama-Guard-3-8B",
        help="Specify the safety shield to use for testing",
    )

--- a/llama_stack/providers/tests/agents/fixtures.py
+++ b/llama_stack/providers/tests/agents/fixtures.py
@ -83,6 +83,6 @@ async def agents_stack(request, inference_model, safety_shield):
            )
            for model in inference_models
        ],
-        shields=[safety_shield],
+        shields=[safety_shield] if safety_shield else [],
    )
    return test_stack
--- a/llama_stack/providers/tests/eval/conftest.py
+++ b/llama_stack/providers/tests/eval/conftest.py
@ -63,7 +63,7 @@ def pytest_addoption(parser):
    parser.addoption(
        "--inference-model",
        action="store",
-        default="Llama3.2-3B-Instruct",
+        default="meta-llama/Llama-3.2-3B-Instruct",
        help="Specify the inference model to use for testing",
    )

--- a/llama_stack/providers/tests/inference/conftest.py
+++ b/llama_stack/providers/tests/inference/conftest.py
@ -32,8 +32,12 @@ def pytest_configure(config):


 MODEL_PARAMS = [
-    pytest.param("Llama3.1-8B-Instruct", marks=pytest.mark.llama_8b, id="llama_8b"),
-    pytest.param("Llama3.2-3B-Instruct", marks=pytest.mark.llama_3b, id="llama_3b"),
+    pytest.param(
+        "meta-llama/Llama-3.1-8B-Instruct", marks=pytest.mark.llama_8b, id="llama_8b"
+    ),
+    pytest.param(
+        "meta-llama/Llama-3.2-3B-Instruct", marks=pytest.mark.llama_3b, id="llama_3b"
+    ),
 ]

 VISION_MODEL_PARAMS = [
--- a/llama_stack/providers/tests/inference/test_model_registration.py
+++ b/llama_stack/providers/tests/inference/test_model_registration.py
@ -6,7 +6,6 @@

 import pytest

-from llama_models.datatypes import CoreModelId

 # How to run this test:
 #
@ -17,11 +16,22 @@ from llama_models.datatypes import CoreModelId

 class TestModelRegistration:
    @pytest.mark.asyncio
-    async def test_register_unsupported_model(self, inference_stack):
-        _, models_impl = inference_stack
+    async def test_register_unsupported_model(self, inference_stack, inference_model):
+        inference_impl, models_impl = inference_stack
+
+        provider = inference_impl.routing_table.get_provider_impl(inference_model)
+        if provider.__provider_spec__.provider_type not in (
+            "meta-reference",
+            "remote::ollama",
+            "remote::vllm",
+            "remote::tgi",
+        ):
+            pytest.skip(
+                "Skipping test for remote inference providers since they can handle large models like 70B instruct"
+            )

        # Try to register a model that's too large for local inference
-        with pytest.raises(Exception) as exc_info:
+        with pytest.raises(ValueError) as exc_info:
            await models_impl.register_model(
                model_id="Llama3.1-70B-Instruct",
            )
@ -37,21 +47,27 @@ class TestModelRegistration:
            )

    @pytest.mark.asyncio
-    async def test_update_model(self, inference_stack):
+    async def test_register_with_llama_model(self, inference_stack):
        _, models_impl = inference_stack

-        # Register a model to update
-        model_id = CoreModelId.llama3_1_8b_instruct.value
-        old_model = await models_impl.register_model(model_id=model_id)
-
-        # Update the model
-        new_model_id = CoreModelId.llama3_2_3b_instruct.value
-        updated_model = await models_impl.update_model(
-            model_id=model_id, provider_model_id=new_model_id
+        _ = await models_impl.register_model(
+            model_id="custom-model",
+            metadata={"llama_model": "meta-llama/Llama-2-7b"},
        )

-        # Retrieve the updated model to verify changes
-        assert updated_model.provider_resource_id != old_model.provider_resource_id
+        with pytest.raises(ValueError) as exc_info:
+            await models_impl.register_model(
+                model_id="custom-model-2",
+                metadata={"llama_model": "meta-llama/Llama-2-7b"},
+                provider_model_id="custom-model",
+            )

-        # Cleanup
-        await models_impl.unregister_model(model_id=model_id)
+    @pytest.mark.asyncio
+    async def test_register_with_invalid_llama_model(self, inference_stack):
+        _, models_impl = inference_stack
+
+        with pytest.raises(ValueError) as exc_info:
+            await models_impl.register_model(
+                model_id="custom-model-2",
+                metadata={"llama_model": "invalid-llama-model"},
+            )
--- a/llama_stack/providers/tests/resolver.py
+++ b/llama_stack/providers/tests/resolver.py
@ -6,7 +6,6 @@

 import json
 import tempfile
-from datetime import datetime
 from typing import Any, Dict, List, Optional

 from llama_stack.distribution.datatypes import *  # noqa: F403
@ -37,7 +36,6 @@ async def construct_stack_for_test(
 ) -> TestStack:
    sqlite_file = tempfile.NamedTemporaryFile(delete=False, suffix=".db")
    run_config = dict(
-        built_at=datetime.now(),
        image_name="test-fixture",
        apis=apis,
        providers=providers,
--- a/llama_stack/providers/tests/safety/fixtures.py
+++ b/llama_stack/providers/tests/safety/fixtures.py
@ -47,6 +47,9 @@ def safety_shield(request):
    else:
        params = {}

+    if not shield_id:
+        return None
+
    return ShieldInput(
        shield_id=shield_id,
        params=params,
--- a/llama_stack/providers/tests/scoring/conftest.py
+++ b/llama_stack/providers/tests/scoring/conftest.py
@ -58,7 +58,7 @@ def pytest_addoption(parser):
    parser.addoption(
        "--inference-model",
        action="store",
-        default="Llama3.2-3B-Instruct",
+        default="meta-llama/Llama-3.2-3B-Instruct",
        help="Specify the inference model to use for testing",
    )

--- a/llama_stack/providers/utils/inference/init.py
+++ b/llama_stack/providers/utils/inference/init.py
@ -31,3 +31,8 @@ def supported_inference_models() -> List[str]:
            or is_supported_safety_model(m)
        )
    ]
+
+
+ALL_HUGGINGFACE_REPOS_TO_MODEL_DESCRIPTOR = {
+    m.huggingface_repo: m.descriptor() for m in all_registered_models()
+}
--- a/llama_stack/providers/utils/inference/model_registry.py
+++ b/llama_stack/providers/utils/inference/model_registry.py
@ -11,6 +11,10 @@ from llama_models.sku_list import all_registered_models

 from llama_stack.providers.datatypes import Model, ModelsProtocolPrivate

+from llama_stack.providers.utils.inference import (
+    ALL_HUGGINGFACE_REPOS_TO_MODEL_DESCRIPTOR,
+)
+
 ModelAlias = namedtuple("ModelAlias", ["provider_model_id", "aliases", "llama_model"])


@ -32,6 +36,16 @@ def build_model_alias(provider_model_id: str, model_descriptor: str) -> ModelAli
    )


+def build_model_alias_with_just_provider_model_id(
+    provider_model_id: str, model_descriptor: str
+) -> ModelAlias:
+    return ModelAlias(
+        provider_model_id=provider_model_id,
+        aliases=[],
+        llama_model=model_descriptor,
+    )
+
+
 class ModelRegistryHelper(ModelsProtocolPrivate):
    def __init__(self, model_aliases: List[ModelAlias]):
        self.alias_to_provider_id_map = {}
@ -51,7 +65,7 @@ class ModelRegistryHelper(ModelsProtocolPrivate):
        if identifier in self.alias_to_provider_id_map:
            return self.alias_to_provider_id_map[identifier]
        else:
-            raise ValueError(f"Unknown model: `{identifier}`")
+            return None

    def get_llama_model(self, provider_model_id: str) -> str:
        if provider_model_id in self.provider_id_to_llama_model_map:
@ -60,8 +74,34 @@ class ModelRegistryHelper(ModelsProtocolPrivate):
            return None

    async def register_model(self, model: Model) -> Model:
-        model.provider_resource_id = self.get_provider_model_id(
-            model.provider_resource_id
-        )
+        provider_resource_id = self.get_provider_model_id(model.provider_resource_id)
+        if provider_resource_id:
+            model.provider_resource_id = provider_resource_id
+        else:
+            if model.metadata.get("llama_model") is None:
+                raise ValueError(
+                    f"Model '{model.provider_resource_id}' is not available and no llama_model was specified in metadata. "
+                    "Please specify a llama_model in metadata or use a supported model identifier"
+                )
+            existing_llama_model = self.get_llama_model(model.provider_resource_id)
+            if existing_llama_model:
+                if existing_llama_model != model.metadata["llama_model"]:
+                    raise ValueError(
+                        f"Provider model id '{model.provider_resource_id}' is already registered to a different llama model: '{existing_llama_model}'"
+                    )
+            else:
+                if (
+                    model.metadata["llama_model"]
+                    not in ALL_HUGGINGFACE_REPOS_TO_MODEL_DESCRIPTOR
+                ):
+                    raise ValueError(
+                        f"Invalid llama_model '{model.metadata['llama_model']}' specified in metadata. "
+                        f"Must be one of: {', '.join(ALL_HUGGINGFACE_REPOS_TO_MODEL_DESCRIPTOR.keys())}"
+                    )
+                self.provider_id_to_llama_model_map[model.provider_resource_id] = (
+                    ALL_HUGGINGFACE_REPOS_TO_MODEL_DESCRIPTOR[
+                        model.metadata["llama_model"]
+                    ]
+                )

        return model
--- a/llama_stack/providers/utils/kvstore/config.py
+++ b/llama_stack/providers/utils/kvstore/config.py
@ -36,6 +36,15 @@ class RedisKVStoreConfig(CommonConfig):
    def url(self) -> str:
        return f"redis://{self.host}:{self.port}"

+    @classmethod
+    def sample_run_config(cls):
+        return {
+            "type": "redis",
+            "namespace": None,
+            "host": "${env.REDIS_HOST:localhost}",
+            "port": "${env.REDIS_PORT:6379}",
+        }
+

 class SqliteKVStoreConfig(CommonConfig):
    type: Literal[KVStoreType.sqlite.value] = KVStoreType.sqlite.value
@ -44,6 +53,19 @@ class SqliteKVStoreConfig(CommonConfig):
        description="File path for the sqlite database",
    )

+    @classmethod
+    def sample_run_config(
+        cls, __distro_dir__: str = "runtime", db_name: str = "kvstore.db"
+    ):
+        return {
+            "type": "sqlite",
+            "namespace": None,
+            "db_path": "${env.SQLITE_STORE_DIR:~/.llama/"
+            + __distro_dir__
+            + "}/"
+            + db_name,
+        }
+

 class PostgresKVStoreConfig(CommonConfig):
    type: Literal[KVStoreType.postgres.value] = KVStoreType.postgres.value
@ -54,6 +76,19 @@ class PostgresKVStoreConfig(CommonConfig):
    password: Optional[str] = None
    table_name: str = "llamastack_kvstore"

+    @classmethod
+    def sample_run_config(cls, table_name: str = "llamastack_kvstore"):
+        return {
+            "type": "postgres",
+            "namespace": None,
+            "host": "${env.POSTGRES_HOST:localhost}",
+            "port": "${env.POSTGRES_PORT:5432}",
+            "db": "${env.POSTGRES_DB}",
+            "user": "${env.POSTGRES_USER}",
+            "password": "${env.POSTGRES_PASSWORD}",
+            "table_name": "${env.POSTGRES_TABLE_NAME:" + table_name + "}",
+        }
+
    @classmethod
    @field_validator("table_name")
    def validate_table_name(cls, v: str) -> str:
--- a/Show more
+++ b/Show more
				`@ -0,0 +1 @@`
				`../../llama_stack/templates/fireworks/run.yaml`
				`@ -0,0 +1 @@`
				`../../llama_stack/templates/meta-reference-gpu/run-with-safety.yaml`
				`@ -0,0 +1 @@`
				`../../llama_stack/templates/ollama/run-with-safety.yaml`
				`@ -0,0 +1 @@`
				`../../llama_stack/templates/remote-vllm/run-with-safety.yaml`
				`@ -0,0 +1 @@`
				`../../llama_stack/templates/tgi/run-with-safety.yaml`
				`@ -0,0 +1 @@`
				`../../llama_stack/templates/together/run.yaml`