Auto-generate distro yamls + docs (#468)

# What does this PR do?

Automatically generates
- build.yaml
- run.yaml
- run-with-safety.yaml
- parts of markdown docs

for the distributions.

## Test Plan

At this point, this only updates the YAMLs and the docs. Some testing
(especially with ollama and vllm) has been performed but needs to be
much more tested.

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",
+            )
+        }

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

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
 

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:

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",
+            )
+        }

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,
         )