llama-stack/llama_stack/distribution/configure.py
Ashwin Bharambe 6bb57e72a7
Remove "routing_table" and "routing_key" concepts for the user (#201)
This PR makes several core changes to the developer experience surrounding Llama Stack.

Background: PR #92 introduced the notion of "routing" to the Llama Stack. It introduces three object types: (1) models, (2) shields and (3) memory banks. Each of these objects can be associated with a distinct provider. So you can get model A to be inferenced locally while model B, C can be inference remotely (e.g.)

However, this had a few drawbacks:

you could not address the provider instances -- i.e., if you configured "meta-reference" with a given model, you could not assign an identifier to this instance which you could re-use later.
the above meant that you could not register a "routing_key" (e.g. model) dynamically and say "please use this existing provider I have already configured" for a new model.
the terms "routing_table" and "routing_key" were exposed directly to the user. in my view, this is way too much overhead for a new user (which almost everyone is.) people come to the stack wanting to do ML and encounter a completely unexpected term.
What this PR does: This PR structures the run config with only a single prominent key:

- providers
Providers are instances of configured provider types. Here's an example which shows two instances of the remote::tgi provider which are serving two different models.

providers:
  inference:
  - provider_id: foo
    provider_type: remote::tgi
    config: { ... }
  - provider_id: bar
    provider_type: remote::tgi
    config: { ... }
Secondly, the PR adds dynamic registration of { models | shields | memory_banks } to the API surface. The distribution still acts like a "routing table" (as previously) except that it asks the backing providers for a listing of these objects. For example it asks a TGI or Ollama inference adapter what models it is serving. Only the models that are being actually served can be requested by the user for inference. Otherwise, the Stack server will throw an error.

When dynamically registering these objects, you can use the provider IDs shown above. Info about providers can be obtained using the Api.inspect set of endpoints (/providers, /routes, etc.)

The above examples shows the correspondence between inference providers and models registry items. Things work similarly for the safety <=> shields and memory <=> memory_banks pairs.

Registry: This PR also makes it so that Providers need to implement additional methods for registering and listing objects. For example, each Inference provider is now expected to implement the ModelsProtocolPrivate protocol (naming is not great!) which consists of two methods

register_model
list_models
The goal is to inform the provider that a certain model needs to be supported so the provider can make any relevant backend changes if needed (or throw an error if the model cannot be supported.)

There are many other cleanups included some of which are detailed in a follow-up comment.
2024-10-10 10:24:13 -07:00

184 lines
6.1 KiB
Python

# 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.
import textwrap
from typing import Any
from llama_stack.distribution.datatypes import * # noqa: F403
from termcolor import cprint
from llama_stack.distribution.distribution import (
builtin_automatically_routed_apis,
get_provider_registry,
)
from llama_stack.distribution.utils.dynamic import instantiate_class_type
from llama_stack.distribution.utils.prompt_for_config import prompt_for_config
from llama_stack.apis.models import * # noqa: F403
from llama_stack.apis.shields import * # noqa: F403
from llama_stack.apis.memory_banks import * # noqa: F403
def configure_single_provider(
registry: Dict[str, ProviderSpec], provider: Provider
) -> Provider:
provider_spec = registry[provider.provider_type]
config_type = instantiate_class_type(provider_spec.config_class)
try:
if provider.config:
existing = config_type(**provider.config)
else:
existing = None
except Exception:
existing = None
cfg = prompt_for_config(config_type, existing)
return Provider(
provider_id=provider.provider_id,
provider_type=provider.provider_type,
config=cfg.dict(),
)
def configure_api_providers(
config: StackRunConfig, build_spec: DistributionSpec
) -> StackRunConfig:
is_nux = len(config.providers) == 0
if is_nux:
print(
textwrap.dedent(
"""
Llama Stack is composed of several APIs working together. For each API served by the Stack,
we need to configure the providers (implementations) you want to use for these APIs.
"""
)
)
provider_registry = get_provider_registry()
builtin_apis = [a.routing_table_api for a in builtin_automatically_routed_apis()]
if config.apis:
apis_to_serve = config.apis
else:
apis_to_serve = [a.value for a in Api if a not in (Api.telemetry, Api.inspect)]
for api_str in apis_to_serve:
api = Api(api_str)
if api in builtin_apis:
continue
if api not in provider_registry:
raise ValueError(f"Unknown API `{api_str}`")
existing_providers = config.providers.get(api_str, [])
if existing_providers:
cprint(
f"Re-configuring existing providers for API `{api_str}`...",
"green",
attrs=["bold"],
)
updated_providers = []
for p in existing_providers:
print(f"> Configuring provider `({p.provider_type})`")
updated_providers.append(
configure_single_provider(provider_registry[api], p)
)
print("")
else:
# we are newly configuring this API
plist = build_spec.providers.get(api_str, [])
plist = plist if isinstance(plist, list) else [plist]
if not plist:
raise ValueError(f"No provider configured for API {api_str}?")
cprint(f"Configuring API `{api_str}`...", "green", attrs=["bold"])
updated_providers = []
for i, provider_type in enumerate(plist):
print(f"> Configuring provider `({provider_type})`")
updated_providers.append(
configure_single_provider(
provider_registry[api],
Provider(
provider_id=(
f"{provider_type}-{i:02d}"
if len(plist) > 1
else provider_type
),
provider_type=provider_type,
config={},
),
)
)
print("")
config.providers[api_str] = updated_providers
return config
def upgrade_from_routing_table(
config_dict: Dict[str, Any],
) -> Dict[str, Any]:
def get_providers(entries):
return [
Provider(
provider_id=(
f"{entry['provider_type']}-{i:02d}"
if len(entries) > 1
else entry["provider_type"]
),
provider_type=entry["provider_type"],
config=entry["config"],
)
for i, entry in enumerate(entries)
]
providers_by_api = {}
routing_table = config_dict.get("routing_table", {})
for api_str, entries in routing_table.items():
providers = get_providers(entries)
providers_by_api[api_str] = providers
provider_map = config_dict.get("api_providers", config_dict.get("provider_map", {}))
if provider_map:
for api_str, provider in provider_map.items():
if isinstance(provider, dict) and "provider_type" in provider:
providers_by_api[api_str] = [
Provider(
provider_id=f"{provider['provider_type']}",
provider_type=provider["provider_type"],
config=provider["config"],
)
]
config_dict["providers"] = providers_by_api
config_dict.pop("routing_table", None)
config_dict.pop("api_providers", None)
config_dict.pop("provider_map", None)
config_dict["apis"] = config_dict["apis_to_serve"]
config_dict.pop("apis_to_serve", None)
return config_dict
def parse_and_maybe_upgrade_config(config_dict: Dict[str, Any]) -> StackRunConfig:
version = config_dict.get("version", None)
if version == LLAMA_STACK_RUN_CONFIG_VERSION:
return StackRunConfig(**config_dict)
if "routing_table" in config_dict:
print("Upgrading config...")
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)