The fastest, cheapest, most efficient open-source agent harness. Run more than 500 agents on a $10 VPS.
Rust agent-orchestrationai-agentsai-assistantdesktopllmlocal-firstmcppersonal-aiprivacyrustsecond-braintauri
chore: files changed crates/openhuman-embed/src/complete.rs
Steven Enamakel5748ba012394
diff --git a/crates/openhuman-embed/src/complete.rs b/crates/openhuman-embed/src/complete.rsindex e5ec23357d..2c01898791 100644--- a/crates/openhuman-embed/src/complete.rs+++ b/crates/openhuman-embed/src/complete.rs@@ -642,16 +642,13 @@ impl Completer {request.timeout_ms);let wire_request = request.into_wire();- if let Err(message) =- openhuman_core::inference::host_runtime::ops::validate_complete_once_request(- &wire_request,- )- {- return Err(CoreError::Rpc {- method: COMPLETE,- message,- });- }+ let model = endpoint.headers.iter().fold(+ tinyinference_llm::providers::openai::OpenAiModel::new(endpoint.api_key.clone())+ .with_base_url(endpoint.base_url.clone())+ .with_model(model.clone()),+ |model, (name, value)| model.with_header(name.clone(), value.clone()),+ );+ let model: Arc<dyn tinyinference_llm::model::ChatModel<()>> = Arc::new(model);let budget = self.budget.as_ref()@@ -660,24 +657,37 @@ impl Completer {call: budget.call,});// The spend ledger is an explicit argument of the call, not an ambient scope.- let call = openhuman_core::inference::host_runtime::ops::complete_once_controlled(- &endpoint,+ let budget = budget.as_ref().map(|budget| {+ tinyagents_harness::cost::ModelBudget::new(budget.ledger.clone(), budget.call)+ });+ let call = tinyagents_harness::oneshot::complete_once_controlled(+ model,wire_request,- budget.as_ref(),+ tinyagents_harness::oneshot::OneshotOptions { budget },self.observers.physical_dispatch.clone(),self.dispatch_admission.clone(),);let started = std::time::Instant::now();let response = call.await.map_err(|error| {let message = match error {- tinyagents_harness::cost::ControlledDispatchError::Admission(source) => {+ tinyagents_harness::oneshot::CompleteOnceError::InvalidRequest(message) => {+ return CoreError::Rpc {+ method: COMPLETE,+ message,+ };+ }+ tinyagents_harness::oneshot::CompleteOnceError::Dispatch(+ tinyagents_harness::cost::ControlledDispatchError::Admission(source),+ ) => {return CoreError::DispatchAdmission {method: COMPLETE,source,usage: None,};}- tinyagents_harness::cost::ControlledDispatchError::Model(source) => {+ tinyagents_harness::oneshot::CompleteOnceError::Dispatch(+ tinyagents_harness::cost::ControlledDispatchError::Model(source),+ ) => {format!("complete_once: {source}")}};@@ -706,7 +716,7 @@ impl Completer {fn checked_endpoint(&self,request: &CompletionRequest,- ) -> Result<openhuman_core::inference::host_runtime::ops::CompletionEndpoint, CoreError> {+ ) -> Result<CompletionEndpoint, CoreError> {let base_url = self.route.base_url.trim();let api_key = self.route.api_key.trim();if request.model.trim().is_empty() || base_url.is_empty() || api_key.is_empty() {@@ -718,13 +728,11 @@ impl Completer {endpoint: sanitize_url_for_display(base_url),});}- Ok(- openhuman_core::inference::host_runtime::ops::CompletionEndpoint {- base_url: base_url.to_string(),- api_key: api_key.to_string(),- headers: self.headers.clone(),- },- )+ Ok(CompletionEndpoint {+ base_url: base_url.to_string(),+ api_key: api_key.to_string(),+ headers: self.headers.clone(),+ })}}