apunkt/plano
1
0
Fork 0
mirror of https://github.com/katanemo/plano.git synced 2026-06-20 15:28:07 +02:00
Code Issues Projects Releases Packages Wiki Activity

config_generator: allow multiple providers for the same model

Browse source 
Remove model_id-level deduplication that prevented different providers
from serving the same model (e.g., custom/claude-opus-4-6 and
custom-aws/claude-opus-4-6). Full model_name dedup is already handled
upstream. Also fix model field to use full provider/model format
instead of stripped model_id, and adjust router model lookup
accordingly.

Signed-off-by: Troy Mitchell <i@troy-y.org>
This commit is contained in:
Troy Mitchell 2026-04-28 16:54:27 +08:00
parent f3e35914e3
commit a295337ef1
1 changed files with 10 additions and 12 deletions
Download patch file Download diff file Expand all files Collapse all files

22
cli/planoai/config_generator.py
View file

@ -415,12 +415,10 @@ def validate_and_render_schema():
) )
# For wildcard models, don't add model_id to the keys since it's "*" # For wildcard models, don't add model_id to the keys since it's "*"
if not is_wildcard: # Note: full model_name dedup is already done above (line 226).
if model_id in model_name_keys: # We no longer dedup on model_id alone, because different providers
raise Exception( # can serve the same model (e.g., custom/claude-opus-4-6 and
f"Duplicate model_id {model_id}, please provide unique model_id for each model_provider" # custom-aws/claude-opus-4-6 share model_id but are distinct providers).
)
model_name_keys.add(model_id)
# Warn if both passthrough_auth and access_key are configured # Warn if both passthrough_auth and access_key are configured
if model_provider.get("passthrough_auth") and model_provider.get( if model_provider.get("passthrough_auth") and model_provider.get(
@ -431,7 +429,7 @@ def validate_and_render_schema():
f"The access_key will be ignored and the client's Authorization header will be forwarded instead." f"The access_key will be ignored and the client's Authorization header will be forwarded instead."
) )
model_provider["model"] = model_id model_provider["model"] = model_name
model_provider["provider_interface"] = provider model_provider["provider_interface"] = provider
model_provider_name_set.add(model_provider.get("name")) model_provider_name_set.add(model_provider.get("name"))
if model_provider.get("provider") and model_provider.get( if model_provider.get("provider") and model_provider.get(
@ -501,15 +499,15 @@ def validate_and_render_schema():
llms_with_endpoint_cluster_names.add(cluster_name) llms_with_endpoint_cluster_names.add(cluster_name)
overrides_config = config_yaml.get("overrides", {}) overrides_config = config_yaml.get("overrides", {})
# Build lookup of model names (already prefix-stripped by config processing) # Build lookup of model names (full provider/model format)
model_name_set = {mp.get("model") for mp in updated_model_providers} model_name_set = {mp.get("model") for mp in updated_model_providers}
# Auto-add plano-orchestrator provider if routing preferences exist and no provider matches the routing model # Auto-add plano-orchestrator provider if routing preferences exist and no provider matches the routing model
router_model = overrides_config.get("llm_routing_model", "Plano-Orchestrator") router_model = overrides_config.get("llm_routing_model", "Plano-Orchestrator")
router_model_id = ( if len(seen_pref_names) > 0 and router_model not in model_name_set:
router_model.split("/", 1)[1] if "/" in router_model else router_model router_model_id = (
) router_model.split("/", 1)[1] if "/" in router_model else router_model
if len(seen_pref_names) > 0 and router_model_id not in model_name_set: )
updated_model_providers.append( updated_model_providers.append(
{ {
"name": "plano-orchestrator", "name": "plano-orchestrator",