mirror of
https://github.com/katanemo/plano.git
synced 2026-06-17 15:25:17 +02:00
pending tmo work
This commit is contained in:
Adil Hafeez
parent
bb71d041a0
commit
22c84fb689
17 changed files with 1626 additions and 94 deletions
|
|
@ -98,98 +98,122 @@ def validate_and_render_schema():
|
|||
llms_with_usage = []
|
||||
model_name_keys = set()
|
||||
model_usage_name_keys = set()
|
||||
for llm_provider in config_yaml["llm_providers"]:
|
||||
if llm_provider.get("usage", None):
|
||||
llms_with_usage.append(llm_provider["name"])
|
||||
if llm_provider.get("name") in llm_provider_name_set:
|
||||
raise Exception(
|
||||
f"Duplicate llm_provider name {llm_provider.get('name')}, please provide unique name for each llm_provider"
|
||||
)
|
||||
|
||||
llm_gateway_listener = config_yaml.get("listeners", {}).get("egress_traffic", {})
|
||||
if llm_gateway_listener.get("port") == None:
|
||||
llm_gateway_listener["port"] = 12000
|
||||
|
||||
if llm_gateway_listener and config_yaml["llm_providers_v2"]:
|
||||
raise Exception("Please provide either egress_traffic or llm_providers_v2, not both")
|
||||
|
||||
model_name = llm_provider.get("model")
|
||||
if model_name in model_name_keys:
|
||||
raise Exception(
|
||||
f"Duplicate model name {model_name}, please provide unique model name for each llm_provider"
|
||||
)
|
||||
model_name_keys.add(model_name)
|
||||
if llm_provider.get("name") is None:
|
||||
llm_provider["name"] = model_name
|
||||
if config_yaml["llm_providers"]:
|
||||
if config_yaml["llm_providers_v2"] is not None:
|
||||
raise Exception("Please provide either llm_providers or llm_providers_v2, not both")
|
||||
config_yaml["llm_providers_v2"] = {
|
||||
"default": {
|
||||
"listener": {
|
||||
"port": llm_gateway_listener["port"],
|
||||
"protocol": llm_gateway_listener.get("message_format", "openai")
|
||||
},
|
||||
"providers": config_yaml["llm_providers"]
|
||||
}
|
||||
}
|
||||
|
||||
model_name_tokens = model_name.split("/")
|
||||
if len(model_name_tokens) < 2:
|
||||
raise Exception(
|
||||
f"Invalid model name {model_name}. Please provide model name in the format <provider>/<model_id>."
|
||||
)
|
||||
provider = model_name_tokens[0]
|
||||
model_id = "/".join(model_name_tokens[1:])
|
||||
if provider not in SUPPORTED_PROVIDERS:
|
||||
if (
|
||||
llm_provider.get("base_url", None) is None
|
||||
or llm_provider.get("provider_interface", None) is None
|
||||
):
|
||||
raise Exception(
|
||||
f"Must provide base_url and provider_interface for unsupported provider {provider} for model {model_name}. Supported providers are: {', '.join(SUPPORTED_PROVIDERS)}"
|
||||
)
|
||||
provider = llm_provider.get("provider_interface", None)
|
||||
elif llm_provider.get("provider_interface", None) is not None:
|
||||
raise Exception(
|
||||
f"Please provide provider interface as part of model name {model_name} using the format <provider>/<model_id>. For example, use 'openai/gpt-3.5-turbo' instead of 'gpt-3.5-turbo' "
|
||||
)
|
||||
for llm_provider_name, provider_def in config_yaml["llm_providers_v2"].items():
|
||||
provider_listener = provider_def["listener"]
|
||||
|
||||
for llm_provider in provider_def["providers"]:
|
||||
if llm_provider.get("usage", None):
|
||||
llms_with_usage.append(llm_provider["name"])
|
||||
if llm_provider.get("name") in llm_provider_name_set:
|
||||
raise Exception(
|
||||
f"Duplicate llm_provider name {llm_provider.get('name')}, please provide unique name for each llm_provider"
|
||||
)
|
||||
|
||||
if model_id in model_name_keys:
|
||||
raise Exception(
|
||||
f"Duplicate model_id {model_id}, please provide unique model_id for each llm_provider"
|
||||
)
|
||||
model_name_keys.add(model_id)
|
||||
model_name = llm_provider.get("model")
|
||||
if model_name in model_name_keys:
|
||||
raise Exception(
|
||||
f"Duplicate model name {model_name}, please provide unique model name for each llm_provider"
|
||||
)
|
||||
model_name_keys.add(model_name)
|
||||
if llm_provider.get("name") is None:
|
||||
llm_provider["name"] = model_name
|
||||
|
||||
for routing_preference in llm_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"))
|
||||
model_name_tokens = model_name.split("/")
|
||||
if len(model_name_tokens) < 2:
|
||||
raise Exception(
|
||||
f"Invalid model name {model_name}. Please provide model name in the format <provider>/<model_id>."
|
||||
)
|
||||
provider = model_name_tokens[0]
|
||||
model_id = "/".join(model_name_tokens[1:])
|
||||
if provider not in SUPPORTED_PROVIDERS:
|
||||
if (
|
||||
llm_provider.get("base_url", None) is None
|
||||
or llm_provider.get("provider_interface", None) is None
|
||||
):
|
||||
raise Exception(
|
||||
f"Must provide base_url and provider_interface for unsupported provider {provider} for model {model_name}. Supported providers are: {', '.join(SUPPORTED_PROVIDERS)}"
|
||||
)
|
||||
provider = llm_provider.get("provider_interface", None)
|
||||
elif llm_provider.get("provider_interface", None) is not None:
|
||||
raise Exception(
|
||||
f"Please provide provider interface as part of model name {model_name} using the format <provider>/<model_id>. For example, use 'openai/gpt-3.5-turbo' instead of 'gpt-3.5-turbo' "
|
||||
)
|
||||
|
||||
llm_provider["model"] = model_id
|
||||
llm_provider["provider_interface"] = provider
|
||||
llm_provider_name_set.add(llm_provider.get("name"))
|
||||
provider = None
|
||||
if llm_provider.get("provider") and llm_provider.get("provider_interface"):
|
||||
raise Exception(
|
||||
"Please provide either provider or provider_interface, not both"
|
||||
)
|
||||
if llm_provider.get("provider"):
|
||||
provider = llm_provider["provider"]
|
||||
llm_provider["provider_interface"] = provider
|
||||
del llm_provider["provider"]
|
||||
updated_llm_providers.append(llm_provider)
|
||||
if model_id in model_name_keys:
|
||||
raise Exception(
|
||||
f"Duplicate model_id {model_id}, please provide unique model_id for each llm_provider"
|
||||
)
|
||||
model_name_keys.add(model_id)
|
||||
|
||||
if llm_provider.get("base_url", None):
|
||||
base_url = llm_provider["base_url"]
|
||||
urlparse_result = urlparse(base_url)
|
||||
url_path = urlparse_result.path
|
||||
if url_path and url_path != "/":
|
||||
raise Exception(
|
||||
f"Please provide base_url without path, got {base_url}. Use base_url like 'http://example.com' instead of 'http://example.com/path'."
|
||||
)
|
||||
if urlparse_result.scheme == "" or urlparse_result.scheme not in [
|
||||
"http",
|
||||
"https",
|
||||
]:
|
||||
raise Exception(
|
||||
"Please provide a valid URL with scheme (http/https) in base_url"
|
||||
)
|
||||
protocol = urlparse_result.scheme
|
||||
port = urlparse_result.port
|
||||
if port is None:
|
||||
if protocol == "http":
|
||||
port = 80
|
||||
else:
|
||||
port = 443
|
||||
endpoint = urlparse_result.hostname
|
||||
llm_provider["endpoint"] = endpoint
|
||||
llm_provider["port"] = port
|
||||
llm_provider["protocol"] = protocol
|
||||
llms_with_endpoint.append(llm_provider)
|
||||
for routing_preference in llm_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"))
|
||||
|
||||
llm_provider["model"] = model_id
|
||||
llm_provider["provider_interface"] = provider
|
||||
llm_provider_name_set.add(llm_provider.get("name"))
|
||||
provider = None
|
||||
if llm_provider.get("provider") and llm_provider.get("provider_interface"):
|
||||
raise Exception(
|
||||
"Please provide either provider or provider_interface, not both"
|
||||
)
|
||||
if llm_provider.get("provider"):
|
||||
provider = llm_provider["provider"]
|
||||
llm_provider["provider_interface"] = provider
|
||||
del llm_provider["provider"]
|
||||
updated_llm_providers.append(llm_provider)
|
||||
|
||||
if llm_provider.get("base_url", None):
|
||||
base_url = llm_provider["base_url"]
|
||||
urlparse_result = urlparse(base_url)
|
||||
url_path = urlparse_result.path
|
||||
if url_path and url_path != "/":
|
||||
raise Exception(
|
||||
f"Please provide base_url without path, got {base_url}. Use base_url like 'http://example.com' instead of 'http://example.com/path'."
|
||||
)
|
||||
if urlparse_result.scheme == "" or urlparse_result.scheme not in [
|
||||
"http",
|
||||
"https",
|
||||
]:
|
||||
raise Exception(
|
||||
"Please provide a valid URL with scheme (http/https) in base_url"
|
||||
)
|
||||
protocol = urlparse_result.scheme
|
||||
port = urlparse_result.port
|
||||
if port is None:
|
||||
if protocol == "http":
|
||||
port = 80
|
||||
else:
|
||||
port = 443
|
||||
endpoint = urlparse_result.hostname
|
||||
llm_provider["endpoint"] = endpoint
|
||||
llm_provider["port"] = port
|
||||
llm_provider["protocol"] = protocol
|
||||
llms_with_endpoint.append(llm_provider)
|
||||
|
||||
if len(model_usage_name_keys) > 0:
|
||||
routing_llm_provider = config_yaml.get("routing", {}).get("llm_provider", None)
|
||||
|
|
@ -221,14 +245,6 @@ def validate_and_render_schema():
|
|||
if prompt_gateway_listener.get("timeout") == None:
|
||||
prompt_gateway_listener["timeout"] = "10s"
|
||||
|
||||
llm_gateway_listener = config_yaml.get("listeners", {}).get("egress_traffic", {})
|
||||
if llm_gateway_listener.get("port") == None:
|
||||
llm_gateway_listener["port"] = 12000 # default port for llm gateway
|
||||
if llm_gateway_listener.get("address") == None:
|
||||
llm_gateway_listener["address"] = "127.0.0.1"
|
||||
if llm_gateway_listener.get("timeout") == None:
|
||||
llm_gateway_listener["timeout"] = "10s"
|
||||
|
||||
use_agent_orchestrator = config_yaml.get("overrides", {}).get(
|
||||
"use_agent_orchestrator", False
|
||||
)
|
||||
|
|
|
|||
Loading…
Add table
Add a link
Reference in a new issue