Skip to content

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

docs(inference): clarify managed routing and session expiry docs

Steven Enamakel4f574fb3beb9
diff --git a/crates/openhuman-core/src/core/observability.rs b/crates/openhuman-core/src/core/observability.rs
index f4f7f76a13..abb1100545 100644
--- a/crates/openhuman-core/src/core/observability.rs
+++ b/crates/openhuman-core/src/core/observability.rs
@@ -990,8 +990,7 @@ pub fn is_session_expired_message(msg: &str) -> bool {
// TAURI-RUST-N — same OpenHuman backend "Invalid token" envelope
// wrapped by the tinyagents `ProviderError::Display` format
// (`TinyAgentsError::Provider(Box<ProviderError>)`). The crate-native
- // path (`TinyHumansModel` through the host adapter) delegates to
- // TinyInference's `OpenAiModel`
+ // Historical `OpenHumanBackendModel` delegated to `OpenAiModel`
// which formats errors as `"{provider} returned HTTP {status}: {body}"`,
// producing `"OpenHuman returned HTTP 401: ..."` — different from the
// `"OpenHuman API error (401"` prefix the classic `api_error` path emits.
diff --git a/crates/openhuman-core/src/inference/README.md b/crates/openhuman-core/src/inference/README.md
index bc8b402471..b234d02da4 100644
--- a/crates/openhuman-core/src/inference/README.md
+++ b/crates/openhuman-core/src/inference/README.md
@@ -81,10 +81,11 @@ through the same gates and constructors.
Managed chat uses TinyInference's `TinyHumansModel` through the host adapter in
`openhuman-tinyhumans/src/inference/managed_model.rs`. The adapter supplies the live bearer
-resolver and host error observer; TinyInference owns request construction and response parsing.
-The host's typed authentication-failure callback publishes `DomainEvent::SessionExpired` when a
-stored session has expired or the backend rejects it. API-key credentials remain eligible for
-managed inference and do not trigger session demotion.
+resolver and host error observer; TinyInference owns request construction and response parsing,
+including the managed usage envelope. The host checks stored-session expiry before a request and
+publishes `DomainEvent::SessionExpired` locally when that check fails. Its typed authentication-
+failure callback publishes the same event for managed endpoint authentication failures, using the
+owner's explicit provenance rather than inferring ownership from the provider label.
[`provider/openai_codex.rs`](./provider/openai_codex.rs) is related but is not a model. When ChatGPT/Codex OAuth tokens exist
for the `openai` slug, `resolve_openai_codex_routing` re-targets that slug at the Codex backend
diff --git a/crates/openhuman-core/src/inference/provider/factory_crate_native_tests.rs b/crates/openhuman-core/src/inference/provider/factory_crate_native_tests.rs
index 0a37972aee..6db1c76e6e 100644
--- a/crates/openhuman-core/src/inference/provider/factory_crate_native_tests.rs
+++ b/crates/openhuman-core/src/inference/provider/factory_crate_native_tests.rs
@@ -241,9 +241,8 @@ async fn one_shot_chat_models_preserve_factory_temperature_as_request_default()
// ── Motion B (#4727): managed-backend crate-native routing ──────────────────
// `create_chat_model` must route the managed OpenHuman backend through the
-// crate-native `TinyHumansModel`, whose concrete `managed` profile
-// advertises the capabilities that routing previously inferred through the
-// provider adapter.
+// crate-native `TinyHumansModel`, whose profile advertises the capabilities
+// that routing previously inferred through the provider adapter.
#[test]
fn resolves_to_managed_backend_for_default_config_but_not_for_local() {