phoenix-oss/llama-stack-mirror
0
0
Fork 1
mirror of https://github.com/meta-llama/llama-stack.git synced 2025-12-06 18:40:57 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

feat(stores)!: use backend storage references instead of configs (#3697)

Browse source 
**This PR changes configurations in a backward incompatible way.**

Run configs today repeat full SQLite/Postgres snippets everywhere a
store is needed, which means duplicated credentials, extra connection
pools, and lots of drift between files. This PR introduces named storage
backends so the stack and providers can share a single catalog and
reference those backends by name.

## Key Changes

- Add `storage.backends` to `StackRunConfig`, register each KV/SQL
backend once at startup, and validate that references point to the right
family.
- Move server stores under `storage.stores` with lightweight references
(backend + namespace/table) instead of full configs.
- Update every provider/config/doc to use the new reference style;
docs/codegen now surface the simplified YAML.

## Migration

Before:
```yaml
metadata_store:
  type: sqlite
  db_path: ~/.llama/distributions/foo/registry.db
inference_store:
  type: postgres
  host: ${env.POSTGRES_HOST}
  port: ${env.POSTGRES_PORT}
  db: ${env.POSTGRES_DB}
  user: ${env.POSTGRES_USER}
  password: ${env.POSTGRES_PASSWORD}
conversations_store:
  type: postgres
  host: ${env.POSTGRES_HOST}
  port: ${env.POSTGRES_PORT}
  db: ${env.POSTGRES_DB}
  user: ${env.POSTGRES_USER}
  password: ${env.POSTGRES_PASSWORD}
```

After:
```yaml
storage:
  backends:
    kv_default:
      type: kv_sqlite
      db_path: ~/.llama/distributions/foo/kvstore.db
    sql_default:
      type: sql_postgres
      host: ${env.POSTGRES_HOST}
      port: ${env.POSTGRES_PORT}
      db: ${env.POSTGRES_DB}
      user: ${env.POSTGRES_USER}
      password: ${env.POSTGRES_PASSWORD}
  stores:
    metadata:
      backend: kv_default
      namespace: registry
    inference:
      backend: sql_default
      table_name: inference_store
      max_write_queue_size: 10000
      num_writers: 4
    conversations:
      backend: sql_default
      table_name: openai_conversations
```

Provider configs follow the same pattern—for example, a Chroma vector
adapter switches from:

```yaml
providers:
  vector_io:
  - provider_id: chromadb
    provider_type: remote::chromadb
    config:
      url: ${env.CHROMADB_URL}
      kvstore:
        type: sqlite
        db_path: ~/.llama/distributions/foo/chroma.db
```

to:

```yaml
providers:
  vector_io:
  - provider_id: chromadb
    provider_type: remote::chromadb
    config:
      url: ${env.CHROMADB_URL}
      persistence:
        backend: kv_default
        namespace: vector_io::chroma_remote
```

Once the backends are declared, everything else just points at them, so
rotating credentials or swapping to Postgres happens in one place and
the stack reuses a single connection pool.
This commit is contained in:
Ashwin Bharambe 2025-10-20 13:20:09 -07:00 • committed by GitHub
parent add64e8e2a
commit 2c43285e22
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
105 changed files with 2290 additions and 1292 deletions
Download patch file Download diff file Expand all files Collapse all files

76
llama_stack/distributions/template.py
View file

@ -29,6 +29,12 @@ from llama_stack.core.datatypes import (
ToolGroupInput,
)
from llama_stack.core.distribution import get_provider_registry
from llama_stack.core.storage.datatypes import (
InferenceStoreReference,
KVStoreReference,
SqlStoreReference,
StorageBackendType,
)
from llama_stack.core.utils.dynamic import instantiate_class_type
from llama_stack.core.utils.image_types import LlamaStackImageType
from llama_stack.providers.utils.inference.model_registry import ProviderModelEntry
@ -180,10 +186,9 @@ class RunConfigSettings(BaseModel):
default_tool_groups: list[ToolGroupInput] | None = None
default_datasets: list[DatasetInput] | None = None
default_benchmarks: list[BenchmarkInput] | None = None
metadata_store: dict | None = None
inference_store: dict | None = None
conversations_store: dict | None = None
telemetry: TelemetryConfig = Field(default_factory=lambda: TelemetryConfig(enabled=True))
storage_backends: dict[str, Any] | None = None
storage_stores: dict[str, Any] | None = None
def run_config(
self,
@ -226,6 +231,37 @@ class RunConfigSettings(BaseModel):
# Get unique set of APIs from providers
apis = sorted(providers.keys())
storage_backends = self.storage_backends or {
"kv_default": SqliteKVStoreConfig.sample_run_config(
__distro_dir__=f"~/.llama/distributions/{name}",
db_name="kvstore.db",
),
"sql_default": SqliteSqlStoreConfig.sample_run_config(
__distro_dir__=f"~/.llama/distributions/{name}",
db_name="sql_store.db",
),
}
storage_stores = self.storage_stores or {
"metadata": KVStoreReference(
backend="kv_default",
namespace="registry",
).model_dump(exclude_none=True),
"inference": InferenceStoreReference(
backend="sql_default",
table_name="inference_store",
).model_dump(exclude_none=True),
"conversations": SqlStoreReference(
backend="sql_default",
table_name="openai_conversations",
).model_dump(exclude_none=True),
}
storage_config = dict(
backends=storage_backends,
stores=storage_stores,
)
# Return a dict that matches StackRunConfig structure
return {
"version": LLAMA_STACK_RUN_CONFIG_VERSION,
@ -233,21 +269,7 @@ class RunConfigSettings(BaseModel):
"container_image": container_image,
"apis": apis,
"providers": provider_configs,
"metadata_store": self.metadata_store
or SqliteKVStoreConfig.sample_run_config(
__distro_dir__=f"~/.llama/distributions/{name}",
db_name="registry.db",
),
"inference_store": self.inference_store
or SqliteSqlStoreConfig.sample_run_config(
__distro_dir__=f"~/.llama/distributions/{name}",
db_name="inference_store.db",
),
"conversations_store": self.conversations_store
or SqliteSqlStoreConfig.sample_run_config(
__distro_dir__=f"~/.llama/distributions/{name}",
db_name="conversations.db",
),
"storage": storage_config,
"models": [m.model_dump(exclude_none=True) for m in (self.default_models or [])],
"shields": [s.model_dump(exclude_none=True) for s in (self.default_shields or [])],
"vector_dbs": [],
@ -297,11 +319,15 @@ class DistributionTemplate(BaseModel):
# We should have a better way to do this by formalizing the concept of "internal" APIs
# and providers, with a way to specify dependencies for them.
if run_config_.get("inference_store"):
additional_pip_packages.extend(get_sql_pip_packages(run_config_["inference_store"]))
if run_config_.get("metadata_store"):
additional_pip_packages.extend(get_kv_pip_packages(run_config_["metadata_store"]))
storage_cfg = run_config_.get("storage", {})
for backend_cfg in storage_cfg.get("backends", {}).values():
store_type = backend_cfg.get("type")
if not store_type:
continue
if str(store_type).startswith("kv_"):
additional_pip_packages.extend(get_kv_pip_packages(backend_cfg))
elif str(store_type).startswith("sql_"):
additional_pip_packages.extend(get_sql_pip_packages(backend_cfg))
if self.additional_pip_packages:
additional_pip_packages.extend(self.additional_pip_packages)
@ -387,11 +413,13 @@ class DistributionTemplate(BaseModel):
def enum_representer(dumper, data):
return dumper.represent_scalar("tag:yaml.org,2002:str", data.value)
# Register YAML representer for ModelType
# Register YAML representer for enums
yaml.add_representer(ModelType, enum_representer)
yaml.add_representer(DatasetPurpose, enum_representer)
yaml.add_representer(StorageBackendType, enum_representer)
yaml.SafeDumper.add_representer(ModelType, enum_representer)
yaml.SafeDumper.add_representer(DatasetPurpose, enum_representer)
yaml.SafeDumper.add_representer(StorageBackendType, enum_representer)
for output_dir in [yaml_output_dir, doc_output_dir]:
output_dir.mkdir(parents=True, exist_ok=True)