From c4c0a1d693acf9687f47769e53ad8848d8606ce4 Mon Sep 17 00:00:00 2001
From: "Rodrigo Rodriguez (Pragmatismo)" <me@rodrigorodriguez.com>
Date: Fri, 7 Nov 2025 17:04:33 -0300
Subject: [PATCH] feat(llm): add configurable reasoning-format flag support
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

- Added retrieval of `llm-server-reasoning-format` configuration in `src/llm/local.rs`.
- When the config value is non‑empty, the server start command now includes `--reasoning-format <value>`.
- Updated argument construction to conditionally append the new flag.
- Cleaned up `src/automation/mod.rs` by removing an unused `std::sync::Arc` import, simplifying the module and eliminating a dead dependency.
---
 src/automation/mod.rs | 1 -
 src/llm/local.rs      | 5 +++++
 src/shared/state.rs   | 2 +-
 3 files changed, 6 insertions(+), 2 deletions(-)

diff --git a/src/automation/mod.rs b/src/automation/mod.rs
index 7d5148e7..bd62d4b9 100644
--- a/src/automation/mod.rs
+++ b/src/automation/mod.rs
@@ -186,7 +186,6 @@ pub async fn execute_compact_prompt(state: Arc<crate::shared::state::AppState>)
     use crate::shared::models::system_automations::dsl::{is_active, system_automations};
     use diesel::prelude::*;
     use log::info;
-    use std::sync::Arc;
 
     let state_clone = state.clone();
 let service = AutomationService::new(state_clone);
diff --git a/src/llm/local.rs b/src/llm/local.rs
index 9134ba02..bedcd475 100644
--- a/src/llm/local.rs
+++ b/src/llm/local.rs
@@ -199,6 +199,7 @@ pub async fn start_llm_server(
     let mlock = config_manager.get_config(&default_bot_id, "llm-server-mlock", None).unwrap_or("true".to_string());
     let no_mmap = config_manager.get_config(&default_bot_id, "llm-server-no-mmap", None).unwrap_or("true".to_string());
     let gpu_layers = config_manager.get_config(&default_bot_id, "llm-server-gpu-layers", None).unwrap_or("20".to_string());
+    let reasoning_format = config_manager.get_config(&default_bot_id, "llm-server-reasoning-format", None).unwrap_or("".to_string());
     let n_predict = config_manager.get_config(&default_bot_id, "llm-server-n-predict", None).unwrap_or("50".to_string());
 
     // Build command arguments dynamically
@@ -206,6 +207,9 @@ pub async fn start_llm_server(
         "-m {} --host 0.0.0.0 --port {} --top_p 0.95 --temp 0.6 --repeat-penalty 1.2 -ngl {}",
         model_path, port,  gpu_layers
     );
+    if !reasoning_format.is_empty() {
+        args.push_str(&format!(" --reasoning-format {}", reasoning_format));
+    }
 
     if n_moe != "0" {
         args.push_str(&format!(" --n-cpu-moe {}", n_moe));
@@ -219,6 +223,7 @@ pub async fn start_llm_server(
     if mlock == "true" {
         args.push_str(" --mlock");
     }
+
     if no_mmap == "true" {
         args.push_str(" --no-mmap");
     }
diff --git a/src/shared/state.rs b/src/shared/state.rs
index eb8ff400..4f16187c 100644
--- a/src/shared/state.rs
+++ b/src/shared/state.rs
@@ -2,7 +2,7 @@ use crate::channels::{ChannelAdapter, VoiceAdapter, WebChannelAdapter};
 use crate::config::AppConfig;
 use crate::llm::LLMProvider;
 use crate::session::SessionManager;
-use diesel::{Connection, PgConnection};
+use diesel::{ PgConnection};
 use aws_sdk_s3::Client as S3Client;
 use redis::Client as RedisClient;
 use std::collections::HashMap;