mirror of
https://github.com/katanemo/plano.git
synced 2026-05-30 14:25:15 +02:00
fix(routing): auto-migrate v0.3.0 inline routing_preferences to v0.4.0 top-level
Lift inline routing_preferences under each model_provider into the top-level routing_preferences list with merged models[] and bump version to v0.4.0, with a deprecation warning. Existing v0.3.0 demo configs (Claude Code, Codex, preference_based_routing, etc.) keep working unchanged. Schema flags the inline shape as deprecated but still accepts it. Docs and skills updated to canonical top-level multi-model form.
This commit is contained in:
Spherrrical
parent
b81eb7266c
commit
dde90cae82
11 changed files with 693 additions and 224 deletions
|
|
@ -58,6 +58,104 @@ def get_endpoint_and_port(endpoint, protocol):
|
|||
return endpoint, port
|
||||
|
||||
|
||||
def migrate_inline_routing_preferences(config_yaml):
|
||||
"""Lift v0.3.0-style inline ``routing_preferences`` under each
|
||||
``model_providers`` entry to the v0.4.0 top-level ``routing_preferences``
|
||||
list with ``models: [...]``.
|
||||
|
||||
Preferences with the same ``name`` across multiple providers are merged
|
||||
into a single top-level entry whose ``models`` list contains every
|
||||
provider's full ``<provider>/<model>`` string in declaration order. The
|
||||
first ``description`` encountered wins; conflicts are warned, not
|
||||
errored, so existing v0.3.0 configs keep compiling. Any top-level
|
||||
preference already defined by the user is preserved as-is.
|
||||
|
||||
Also bumps ``version`` to ``v0.4.0`` when migration produced any entry,
|
||||
so brightstaff's v0.4.0 gate for top-level ``routing_preferences``
|
||||
accepts the rendered config.
|
||||
"""
|
||||
model_providers = config_yaml.get("model_providers") or []
|
||||
if not model_providers:
|
||||
return
|
||||
|
||||
migrated = {}
|
||||
for model_provider in model_providers:
|
||||
inline_prefs = model_provider.get("routing_preferences")
|
||||
if not inline_prefs:
|
||||
continue
|
||||
|
||||
full_model_name = model_provider.get("model")
|
||||
if not full_model_name:
|
||||
continue
|
||||
|
||||
if "/" in full_model_name and full_model_name.split("/")[-1].strip() == "*":
|
||||
raise Exception(
|
||||
f"Model {full_model_name} has routing_preferences but uses wildcard (*). Models with routing preferences cannot be wildcards."
|
||||
)
|
||||
|
||||
for pref in inline_prefs:
|
||||
name = pref.get("name")
|
||||
description = pref.get("description", "")
|
||||
if not name:
|
||||
continue
|
||||
if name in migrated:
|
||||
entry = migrated[name]
|
||||
if description and description != entry["description"]:
|
||||
print(
|
||||
f"WARNING: routing preference '{name}' has conflicting descriptions across providers; keeping the first one."
|
||||
)
|
||||
if full_model_name not in entry["models"]:
|
||||
entry["models"].append(full_model_name)
|
||||
else:
|
||||
migrated[name] = {
|
||||
"name": name,
|
||||
"description": description,
|
||||
"models": [full_model_name],
|
||||
}
|
||||
|
||||
if not migrated:
|
||||
return
|
||||
|
||||
for model_provider in model_providers:
|
||||
if "routing_preferences" in model_provider:
|
||||
del model_provider["routing_preferences"]
|
||||
|
||||
existing_top_level = config_yaml.get("routing_preferences") or []
|
||||
existing_names = {entry.get("name") for entry in existing_top_level}
|
||||
merged = list(existing_top_level)
|
||||
for name, entry in migrated.items():
|
||||
if name in existing_names:
|
||||
continue
|
||||
merged.append(entry)
|
||||
config_yaml["routing_preferences"] = merged
|
||||
|
||||
current_version = str(config_yaml.get("version", ""))
|
||||
if _version_tuple(current_version) < (0, 4, 0):
|
||||
config_yaml["version"] = "v0.4.0"
|
||||
|
||||
print(
|
||||
"WARNING: inline routing_preferences under model_providers is deprecated "
|
||||
"and has been auto-migrated to top-level routing_preferences. Update your "
|
||||
"config to v0.4.0 top-level form. See docs/routing-api.md"
|
||||
)
|
||||
|
||||
|
||||
def _version_tuple(version_string):
|
||||
stripped = version_string.strip().lstrip("vV")
|
||||
if not stripped:
|
||||
return (0, 0, 0)
|
||||
parts = stripped.split("-", 1)[0].split(".")
|
||||
out = []
|
||||
for part in parts[:3]:
|
||||
try:
|
||||
out.append(int(part))
|
||||
except ValueError:
|
||||
out.append(0)
|
||||
while len(out) < 3:
|
||||
out.append(0)
|
||||
return tuple(out)
|
||||
|
||||
|
||||
def validate_and_render_schema():
|
||||
ENVOY_CONFIG_TEMPLATE_FILE = os.getenv(
|
||||
"ENVOY_CONFIG_TEMPLATE_FILE", "envoy.template.yaml"
|
||||
|
|
@ -101,6 +199,8 @@ def validate_and_render_schema():
|
|||
config_yaml["model_providers"] = config_yaml["llm_providers"]
|
||||
del config_yaml["llm_providers"]
|
||||
|
||||
migrate_inline_routing_preferences(config_yaml)
|
||||
|
||||
listeners, llm_gateway, prompt_gateway = convert_legacy_listeners(
|
||||
config_yaml.get("listeners"), config_yaml.get("model_providers")
|
||||
)
|
||||
|
|
@ -200,7 +300,16 @@ def validate_and_render_schema():
|
|||
model_provider_name_set = set()
|
||||
llms_with_usage = []
|
||||
model_name_keys = set()
|
||||
model_usage_name_keys = set()
|
||||
|
||||
top_level_preferences = config_yaml.get("routing_preferences") or []
|
||||
seen_pref_names = set()
|
||||
for pref in top_level_preferences:
|
||||
pref_name = pref.get("name")
|
||||
if pref_name in seen_pref_names:
|
||||
raise Exception(
|
||||
f'Duplicate routing preference name "{pref_name}", please provide unique name for each routing preference'
|
||||
)
|
||||
seen_pref_names.add(pref_name)
|
||||
|
||||
print("listeners: ", listeners)
|
||||
|
||||
|
|
@ -259,10 +368,6 @@ def validate_and_render_schema():
|
|||
raise Exception(
|
||||
f"Model {model_name} is configured as default but uses wildcard (*). Default models cannot be wildcards."
|
||||
)
|
||||
if model_provider.get("routing_preferences"):
|
||||
raise Exception(
|
||||
f"Model {model_name} has routing_preferences but uses wildcard (*). Models with routing preferences cannot be wildcards."
|
||||
)
|
||||
|
||||
# Validate azure_openai and ollama provider requires base_url
|
||||
if (provider in SUPPORTED_PROVIDERS_WITH_BASE_URL) and model_provider.get(
|
||||
|
|
@ -311,13 +416,6 @@ def validate_and_render_schema():
|
|||
)
|
||||
model_name_keys.add(model_id)
|
||||
|
||||
for routing_preference in model_provider.get("routing_preferences", []):
|
||||
if routing_preference.get("name") in model_usage_name_keys:
|
||||
raise Exception(
|
||||
f'Duplicate routing preference name "{routing_preference.get("name")}", please provide unique name for each routing preference'
|
||||
)
|
||||
model_usage_name_keys.add(routing_preference.get("name"))
|
||||
|
||||
# Warn if both passthrough_auth and access_key are configured
|
||||
if model_provider.get("passthrough_auth") and model_provider.get(
|
||||
"access_key"
|
||||
|
|
@ -405,7 +503,7 @@ def validate_and_render_schema():
|
|||
router_model_id = (
|
||||
router_model.split("/", 1)[1] if "/" in router_model else router_model
|
||||
)
|
||||
if len(model_usage_name_keys) > 0 and router_model_id not in model_name_set:
|
||||
if len(seen_pref_names) > 0 and router_model_id not in model_name_set:
|
||||
updated_model_providers.append(
|
||||
{
|
||||
"name": "plano-orchestrator",
|
||||
|
|
|
|||
|
|
@ -1,7 +1,11 @@
|
|||
import json
|
||||
import pytest
|
||||
import yaml
|
||||
from unittest import mock
|
||||
from planoai.config_generator import validate_and_render_schema
|
||||
from planoai.config_generator import (
|
||||
validate_and_render_schema,
|
||||
migrate_inline_routing_preferences,
|
||||
)
|
||||
|
||||
|
||||
@pytest.fixture(autouse=True)
|
||||
|
|
@ -295,32 +299,30 @@ model_providers:
|
|||
"id": "duplicate_routeing_preference_name",
|
||||
"expected_error": "Duplicate routing preference name",
|
||||
"plano_config": """
|
||||
version: v0.1.0
|
||||
version: v0.4.0
|
||||
|
||||
listeners:
|
||||
egress_traffic:
|
||||
address: 0.0.0.0
|
||||
- name: llm
|
||||
type: model
|
||||
port: 12000
|
||||
message_format: openai
|
||||
timeout: 30s
|
||||
|
||||
llm_providers:
|
||||
|
||||
model_providers:
|
||||
- model: openai/gpt-4o-mini
|
||||
access_key: $OPENAI_API_KEY
|
||||
default: true
|
||||
|
||||
- model: openai/gpt-4o
|
||||
access_key: $OPENAI_API_KEY
|
||||
routing_preferences:
|
||||
- name: code understanding
|
||||
description: understand and explain existing code snippets, functions, or libraries
|
||||
|
||||
- model: openai/gpt-4.1
|
||||
access_key: $OPENAI_API_KEY
|
||||
routing_preferences:
|
||||
- name: code understanding
|
||||
description: generating new code snippets, functions, or boilerplate based on user prompts or requirements
|
||||
routing_preferences:
|
||||
- name: code understanding
|
||||
description: understand and explain existing code snippets, functions, or libraries
|
||||
models:
|
||||
- openai/gpt-4o
|
||||
- name: code understanding
|
||||
description: generating new code snippets, functions, or boilerplate based on user prompts or requirements
|
||||
models:
|
||||
- openai/gpt-4o-mini
|
||||
|
||||
tracing:
|
||||
random_sampling: 100
|
||||
|
|
@ -501,3 +503,190 @@ def test_convert_legacy_llm_providers_no_prompt_gateway():
|
|||
"port": 12000,
|
||||
"timeout": "30s",
|
||||
}
|
||||
|
||||
|
||||
def test_inline_routing_preferences_migrated_to_top_level():
|
||||
plano_config = """
|
||||
version: v0.3.0
|
||||
|
||||
listeners:
|
||||
- type: model
|
||||
name: model_listener
|
||||
port: 12000
|
||||
|
||||
model_providers:
|
||||
- model: openai/gpt-4o-mini
|
||||
access_key: $OPENAI_API_KEY
|
||||
default: true
|
||||
|
||||
- model: openai/gpt-4o
|
||||
access_key: $OPENAI_API_KEY
|
||||
routing_preferences:
|
||||
- name: code understanding
|
||||
description: understand and explain existing code snippets, functions, or libraries
|
||||
|
||||
- model: anthropic/claude-sonnet-4-20250514
|
||||
access_key: $ANTHROPIC_API_KEY
|
||||
routing_preferences:
|
||||
- name: code generation
|
||||
description: generating new code snippets, functions, or boilerplate based on user prompts or requirements
|
||||
"""
|
||||
config_yaml = yaml.safe_load(plano_config)
|
||||
migrate_inline_routing_preferences(config_yaml)
|
||||
|
||||
assert config_yaml["version"] == "v0.4.0"
|
||||
for provider in config_yaml["model_providers"]:
|
||||
assert "routing_preferences" not in provider
|
||||
|
||||
top_level = config_yaml["routing_preferences"]
|
||||
by_name = {entry["name"]: entry for entry in top_level}
|
||||
assert set(by_name) == {"code understanding", "code generation"}
|
||||
assert by_name["code understanding"]["models"] == ["openai/gpt-4o"]
|
||||
assert by_name["code generation"]["models"] == [
|
||||
"anthropic/claude-sonnet-4-20250514"
|
||||
]
|
||||
assert (
|
||||
by_name["code understanding"]["description"]
|
||||
== "understand and explain existing code snippets, functions, or libraries"
|
||||
)
|
||||
|
||||
|
||||
def test_inline_same_name_across_providers_merges_models():
|
||||
plano_config = """
|
||||
version: v0.3.0
|
||||
|
||||
listeners:
|
||||
- type: model
|
||||
name: model_listener
|
||||
port: 12000
|
||||
|
||||
model_providers:
|
||||
- model: openai/gpt-4o
|
||||
access_key: $OPENAI_API_KEY
|
||||
routing_preferences:
|
||||
- name: code generation
|
||||
description: generating new code snippets, functions, or boilerplate based on user prompts or requirements
|
||||
|
||||
- model: anthropic/claude-sonnet-4-20250514
|
||||
access_key: $ANTHROPIC_API_KEY
|
||||
routing_preferences:
|
||||
- name: code generation
|
||||
description: generating new code snippets, functions, or boilerplate based on user prompts or requirements
|
||||
"""
|
||||
config_yaml = yaml.safe_load(plano_config)
|
||||
migrate_inline_routing_preferences(config_yaml)
|
||||
|
||||
top_level = config_yaml["routing_preferences"]
|
||||
assert len(top_level) == 1
|
||||
entry = top_level[0]
|
||||
assert entry["name"] == "code generation"
|
||||
assert entry["models"] == [
|
||||
"openai/gpt-4o",
|
||||
"anthropic/claude-sonnet-4-20250514",
|
||||
]
|
||||
assert config_yaml["version"] == "v0.4.0"
|
||||
|
||||
|
||||
def test_existing_top_level_routing_preferences_preserved():
|
||||
plano_config = """
|
||||
version: v0.4.0
|
||||
|
||||
listeners:
|
||||
- type: model
|
||||
name: model_listener
|
||||
port: 12000
|
||||
|
||||
model_providers:
|
||||
- model: openai/gpt-4o
|
||||
access_key: $OPENAI_API_KEY
|
||||
- model: anthropic/claude-sonnet-4-20250514
|
||||
access_key: $ANTHROPIC_API_KEY
|
||||
|
||||
routing_preferences:
|
||||
- name: code generation
|
||||
description: generating new code snippets or boilerplate
|
||||
models:
|
||||
- openai/gpt-4o
|
||||
- anthropic/claude-sonnet-4-20250514
|
||||
"""
|
||||
config_yaml = yaml.safe_load(plano_config)
|
||||
before = yaml.safe_dump(config_yaml, sort_keys=True)
|
||||
migrate_inline_routing_preferences(config_yaml)
|
||||
after = yaml.safe_dump(config_yaml, sort_keys=True)
|
||||
|
||||
assert before == after
|
||||
|
||||
|
||||
def test_existing_top_level_wins_over_inline_migration():
|
||||
plano_config = """
|
||||
version: v0.3.0
|
||||
|
||||
listeners:
|
||||
- type: model
|
||||
name: model_listener
|
||||
port: 12000
|
||||
|
||||
model_providers:
|
||||
- model: openai/gpt-4o
|
||||
access_key: $OPENAI_API_KEY
|
||||
routing_preferences:
|
||||
- name: code generation
|
||||
description: inline description should lose
|
||||
|
||||
routing_preferences:
|
||||
- name: code generation
|
||||
description: user-defined top-level description wins
|
||||
models:
|
||||
- openai/gpt-4o
|
||||
"""
|
||||
config_yaml = yaml.safe_load(plano_config)
|
||||
migrate_inline_routing_preferences(config_yaml)
|
||||
|
||||
top_level = config_yaml["routing_preferences"]
|
||||
assert len(top_level) == 1
|
||||
entry = top_level[0]
|
||||
assert entry["description"] == "user-defined top-level description wins"
|
||||
assert entry["models"] == ["openai/gpt-4o"]
|
||||
|
||||
|
||||
def test_wildcard_with_inline_routing_preferences_errors():
|
||||
plano_config = """
|
||||
version: v0.3.0
|
||||
|
||||
listeners:
|
||||
- type: model
|
||||
name: model_listener
|
||||
port: 12000
|
||||
|
||||
model_providers:
|
||||
- model: openrouter/*
|
||||
base_url: https://openrouter.ai/api/v1
|
||||
passthrough_auth: true
|
||||
routing_preferences:
|
||||
- name: code generation
|
||||
description: generating code
|
||||
"""
|
||||
config_yaml = yaml.safe_load(plano_config)
|
||||
with pytest.raises(Exception) as excinfo:
|
||||
migrate_inline_routing_preferences(config_yaml)
|
||||
assert "wildcard" in str(excinfo.value).lower()
|
||||
|
||||
|
||||
def test_migration_noop_when_no_inline_preferences():
|
||||
plano_config = """
|
||||
version: v0.3.0
|
||||
|
||||
listeners:
|
||||
- type: model
|
||||
name: model_listener
|
||||
port: 12000
|
||||
|
||||
model_providers:
|
||||
- model: openai/gpt-4o
|
||||
access_key: $OPENAI_API_KEY
|
||||
"""
|
||||
config_yaml = yaml.safe_load(plano_config)
|
||||
migrate_inline_routing_preferences(config_yaml)
|
||||
|
||||
assert "routing_preferences" not in config_yaml
|
||||
assert config_yaml["version"] == "v0.3.0"
|
||||
|
|
|
|||
Loading…
Add table
Add a link
Reference in a new issue