add configurable orchestrator_model_context_length override (default 8192)

Made-with: Cursor

config/plano_config_schema.yaml

@@ -288,6 +288,9 @@ properties:
       agent_orchestration_model:
         type: string
         description: "Model name for the agent orchestrator (e.g., 'Plano-Orchestrator'). Must match a model in model_providers."
+      orchestrator_model_context_length:
+        type: integer
+        description: "Maximum token length for the orchestrator/routing model context window. Default is 8192."
   system_prompt:
     type: string
   prompt_targets:

crates/brightstaff/src/handlers/agents/selector.rs

@@ -177,6 +177,7 @@ mod tests {
             "http://localhost:8080".to_string(),
             "test-model".to_string(),
             "plano-orchestrator".to_string(),
+            crate::router::orchestrator_model_v1::MAX_TOKEN_LEN,
         ))
     }
 

crates/brightstaff/src/handlers/integration_tests.rs

@@ -23,6 +23,7 @@ mod tests {
             "http://localhost:8080".to_string(),
             "test-model".to_string(),
             "plano-orchestrator".to_string(),
+            crate::router::orchestrator_model_v1::MAX_TOKEN_LEN,
         ))
     }
 

crates/brightstaff/src/main.rs

@@ -303,6 +303,10 @@ async fn init_app_state(
         .map(|p| p.name.clone())
         .unwrap_or_else(|| DEFAULT_ORCHESTRATOR_LLM_PROVIDER.to_string());
 
+    let orchestrator_max_tokens = overrides
+        .orchestrator_model_context_length
+        .unwrap_or(brightstaff::router::orchestrator_model_v1::MAX_TOKEN_LEN);
+
     let orchestrator_service = Arc::new(OrchestratorService::with_routing(
         format!("{llm_provider_url}{CHAT_COMPLETIONS_PATH}"),
         orchestrator_model_name,
@@ -312,6 +316,7 @@ async fn init_app_state(
         session_ttl_seconds,
         session_cache,
         session_tenant_header,
+        orchestrator_max_tokens,
     ));
 
     let state_storage = init_state_storage(config).await?;

crates/brightstaff/src/router/orchestrator.rs

@@ -50,11 +50,12 @@ impl OrchestratorService {
         orchestrator_url: String,
         orchestration_model_name: String,
         orchestrator_provider_name: String,
+        max_token_length: usize,
     ) -> Self {
         let orchestrator_model = Arc::new(orchestrator_model_v1::OrchestratorModelV1::new(
             HashMap::new(),
             orchestration_model_name,
-            orchestrator_model_v1::MAX_TOKEN_LEN,
+            max_token_length,
         ));
 
         OrchestratorService {
@@ -80,6 +81,7 @@ impl OrchestratorService {
         session_ttl_seconds: Option<u64>,
         session_cache: Arc<dyn SessionCache>,
         tenant_header: Option<String>,
+        max_token_length: usize,
     ) -> Self {
         let top_level_preferences: HashMap<String, TopLevelRoutingPreference> = top_level_prefs
             .map_or_else(HashMap::new, |prefs| {
@@ -89,7 +91,7 @@ impl OrchestratorService {
         let orchestrator_model = Arc::new(orchestrator_model_v1::OrchestratorModelV1::new(
             HashMap::new(),
             orchestration_model_name,
-            orchestrator_model_v1::MAX_TOKEN_LEN,
+            max_token_length,
         ));
 
         let session_ttl =
@@ -333,6 +335,7 @@ mod tests {
             Some(ttl_seconds),
             session_cache,
             None,
+            orchestrator_model_v1::MAX_TOKEN_LEN,
         )
     }
 

crates/common/src/configuration.rs

@@ -233,6 +233,7 @@ pub struct Overrides {
     pub use_agent_orchestrator: Option<bool>,
     pub llm_routing_model: Option<String>,
     pub agent_orchestration_model: Option<String>,
+    pub orchestrator_model_context_length: Option<usize>,
 }
 
 #[derive(Debug, Clone, Serialize, Deserialize, Default)]