Skip to content

The fastest, cheapest, most efficient open-source agent harness. Run more than 500 agents on a $10 VPS.

refactor(inference): update provider discovery logic

Steven Enamakelb6a9ab8e0d58
diff --git a/crates/openhuman-core/src/inference/provider/factory/discovery.rs b/crates/openhuman-core/src/inference/provider/factory/discovery.rs
index 2bd256a670..728f8b52de 100644
--- a/crates/openhuman-core/src/inference/provider/factory/discovery.rs
+++ b/crates/openhuman-core/src/inference/provider/factory/discovery.rs
@@ -5,6 +5,7 @@
use super::*;
use tinyinference_llm::model::discover::DiscoveryRequest;
+use tinyinference_llm::providers::registry::{ProviderBuildOptions, ProviderRegistry};
/// The discovery request for `provider` (a role's resolved provider string)
/// serving `model`, or `None` when the route has no listing to ask:
@@ -83,15 +84,25 @@ pub(crate) fn model_limits_request_for(
return None;
}
let spec = super::cloud_spec::cloud_provider_spec(resolution.entry, &resolution.effective_model);
- super::cloud_spec::cloud_discovery_request(
- &spec,
+ let options = ProviderBuildOptions {
+ auth_style: Some(resolution.entry.auth_style.into()),
+ codex: Some(&resolution.codex),
+ ..ProviderBuildOptions::default()
+ };
+ match ProviderRegistry::new(super::cloud_spec::CloudCredentialResolver(
&resolution.key,
- resolution.entry.auth_style,
- &[],
- &[],
- None,
- Some(&resolution.codex),
- )
+ ))
+ .discovery_request(&spec, &options)
+ {
+ Ok(request) => request,
+ Err(error) => {
+ log::debug!(
+ "[model_limits] registry discovery unavailable slug={}: {error}",
+ spec.provider
+ );
+ None
+ }
+ }
}
/// The managed backend's catalogue listing request, from the registered