feat(orchestrator): use reasoning-quick-v1 for low-latency chat (#1761)

src/openhuman/agent/agents/loader.rs

@@ -303,7 +303,7 @@ mod tests {
     #[test]
     fn orchestrator_has_reasoning_hint_and_named_tools() {
         let def = find("orchestrator");
-        assert!(matches!(def.model, ModelSpec::Hint(ref h) if h == "reasoning"));
+        assert!(matches!(def.model, ModelSpec::Hint(ref h) if h == "reasoning-quick"));
         match def.tools {
             ToolScope::Named(tools) => {
                 // spawn_subagent was removed in #1141; spawn_worker_thread is the replacement

src/openhuman/agent/agents/orchestrator/agent.toml

@@ -61,7 +61,14 @@ subagents = [
 ]
 
 [model]
-hint = "reasoning"
+# `reasoning-quick` (Kimi K2.6 Turbo on Fireworks via backend PR #760)
+# is tuned for low time-to-first-token on conversational turns. The
+# orchestrator is a planner/router that mostly picks a delegate and
+# synthesises sub-agent output — workload that doesn't benefit from
+# the slower deep-reasoning tier. Sub-agents that need heavier
+# reasoning can still opt into `reasoning-v1` (DeepSeek V4 Pro) via
+# `ModelSpec::Hint("reasoning")` in their own definitions.
+hint = "reasoning-quick"
 
 [tools]
 # Direct tools — things the orchestrator calls itself rather than

src/openhuman/agent/cost.rs

@@ -69,6 +69,15 @@ pub const PRICING_TABLE: &[ModelPricing] = &[
         cached_input_per_mtok_usd: 1.50,
         output_per_mtok_usd: 75.00,
     },
+    // Quick reasoning tier — Kimi K2.6 Turbo on Fireworks (backend PR
+    // #760). Low TTFT, 128k context, `supportsThinking: false`. Rates
+    // track Fireworks' published Kimi turbo pricing at time of writing.
+    ModelPricing {
+        model: "reasoning-quick-v1",
+        input_per_mtok_usd: 0.60,
+        cached_input_per_mtok_usd: 0.06,
+        output_per_mtok_usd: 2.50,
+    },
     // Agentic tier — maps to Sonnet-class models.
     ModelPricing {
         model: "agentic-v1",
@@ -95,14 +104,21 @@ pub fn lookup_pricing(model: &str) -> ModelPricing {
         return *row;
     }
     let lower = model.to_ascii_lowercase();
+    let by_tier = |tier: &str| {
+        PRICING_TABLE
+            .iter()
+            .find(|row| row.model == tier)
+            .copied()
+            .unwrap_or(FALLBACK_PRICING)
+    };
     if lower.contains("opus") {
-        return PRICING_TABLE[0];
+        return by_tier("reasoning-v1");
     }
     if lower.contains("coding") {
-        return PRICING_TABLE[2];
+        return by_tier("coding-v1");
     }
     if lower.contains("sonnet") || lower.contains("agentic") {
-        return PRICING_TABLE[1];
+        return by_tier("agentic-v1");
     }
     FALLBACK_PRICING
 }

src/openhuman/config/mod.rs

@@ -36,7 +36,8 @@ pub use schema::{
     SecretsConfig, SecurityConfig, SlackConfig, StorageConfig, StorageProviderConfig,
     StorageProviderSection, StreamMode, TelegramConfig, UpdateConfig, UpdateRestartStrategy,
     VoiceActivationMode, VoiceServerConfig, WebSearchConfig, WebhookConfig,
-    DEFAULT_CLOUD_LLM_MODEL, DEFAULT_MODEL, MODEL_AGENTIC_V1, MODEL_CODING_V1, MODEL_REASONING_V1,
+    DEFAULT_CLOUD_LLM_MODEL, DEFAULT_MODEL, MODEL_AGENTIC_V1, MODEL_CODING_V1,
+    MODEL_REASONING_QUICK_V1, MODEL_REASONING_V1,
 };
 pub use schema::{
     clear_active_user, default_root_openhuman_dir, pre_login_user_dir, read_active_user_id,

src/openhuman/config/schema/identity_cost.rs

@@ -68,7 +68,9 @@ impl Default for CostConfig {
 
 /// Default pricing for popular models (USD per 1M tokens)
 fn get_default_pricing() -> HashMap<String, ModelPricing> {
-    use super::types::{MODEL_AGENTIC_V1, MODEL_CODING_V1, MODEL_REASONING_V1};
+    use super::types::{
+        MODEL_AGENTIC_V1, MODEL_CODING_V1, MODEL_REASONING_QUICK_V1, MODEL_REASONING_V1,
+    };
 
     let mut prices = HashMap::new();
 
@@ -79,6 +81,14 @@ fn get_default_pricing() -> HashMap<String, ModelPricing> {
             output: 2.52,
         },
     );
+    // Kimi K2.6 Turbo on Fireworks — see backend PR #760.
+    prices.insert(
+        MODEL_REASONING_QUICK_V1.into(),
+        ModelPricing {
+            input: 0.60,
+            output: 2.50,
+        },
+    );
     prices.insert(
         MODEL_AGENTIC_V1.into(),
         ModelPricing {

src/openhuman/config/schema/types.rs

@@ -9,6 +9,13 @@ use std::path::PathBuf;
 /// Standard model identifiers matching the backend model registry.
 pub const MODEL_AGENTIC_V1: &str = "agentic-v1";
 pub const MODEL_REASONING_V1: &str = "reasoning-v1";
+/// Low-latency reasoning tier. Backend maps this to Kimi K2.6 Turbo on
+/// Fireworks (128k context, `supportsThinking: false`) — tuned for
+/// time-to-first-token on conversational turns. See backend PR #760.
+/// The orchestrator (user-facing front-line agent) rides on this tier
+/// by default so chat responses feel snappy; reach for the slower
+/// `reasoning-v1` (DeepSeek V4 Pro) only when deep reasoning is needed.
+pub const MODEL_REASONING_QUICK_V1: &str = "reasoning-quick-v1";
 pub const MODEL_CODING_V1: &str = "coding-v1";
 /// Default model used when no explicit model is configured.
 ///

src/openhuman/providers/router.rs

@@ -3,12 +3,14 @@ use super::Provider;
 use async_trait::async_trait;
 use std::collections::HashMap;
 
-/// Maps OpenHuman's abstract tier model names (`reasoning-v1`, `agentic-v1`,
-/// `coding-v1`, `summarization-v1`) to the hint slot in `model_routes`.
-/// Returns `None` for any model the router shouldn't rewrite.
+/// Maps OpenHuman's abstract tier model names (`reasoning-v1`,
+/// `reasoning-quick-v1`, `agentic-v1`, `coding-v1`, `summarization-v1`)
+/// to the hint slot in `model_routes`. Returns `None` for any model the
+/// router shouldn't rewrite.
 fn openhuman_tier_to_hint(model: &str) -> Option<&'static str> {
     match model {
         "reasoning-v1" => Some("reasoning"),
+        "reasoning-quick-v1" => Some("reasoning-quick"),
         "agentic-v1" => Some("agentic"),
         "coding-v1" => Some("coding"),
         "summarization-v1" => Some("summarization"),

src/openhuman/routing/provider.rs

@@ -17,7 +17,9 @@ use std::time::Instant;
 use anyhow::Result;
 use async_trait::async_trait;
 
-use crate::openhuman::config::{MODEL_AGENTIC_V1, MODEL_CODING_V1, MODEL_REASONING_V1};
+use crate::openhuman::config::{
+    MODEL_AGENTIC_V1, MODEL_CODING_V1, MODEL_REASONING_QUICK_V1, MODEL_REASONING_V1,
+};
 use crate::openhuman::providers::traits::{
     ChatMessage, ChatRequest, ChatResponse, Provider, ProviderCapabilities, StreamChunk,
     StreamError, StreamOptions, StreamResult, ToolsPayload,
@@ -93,6 +95,7 @@ impl IntelligentRoutingProvider {
         // Keep remote model naming aligned with backend modelRegistry.
         match requested_model.strip_prefix("hint:") {
             Some("reasoning") => MODEL_REASONING_V1.to_string(),
+            Some("reasoning-quick") => MODEL_REASONING_QUICK_V1.to_string(),
             Some("agentic") => MODEL_AGENTIC_V1.to_string(),
             Some("coding") => MODEL_CODING_V1.to_string(),
             _ => requested_model.to_string(),