From e15a9d63b086a6445ad23953984f9cd2c9e3b1fe Mon Sep 17 00:00:00 2001
From: Georgi Gerganov <ggerganov@gmail.com>
Date: Sat, 31 Jan 2026 11:39:32 +0200
Subject: [PATCH] inst : improve slot distribution

---
 autoload/llama.vim | 10 ++++++----
 1 file changed, 6 insertions(+), 4 deletions(-)

diff --git a/autoload/llama.vim b/autoload/llama.vim
index b03af6a..18e8c09 100644
--- a/autoload/llama.vim
+++ b/autoload/llama.vim
@@ -1373,10 +1373,15 @@ function! llama#inst(l0, l1)
     let l:l0 = a:l0
     let l:l1 = a:l1
 
+    " create request state
+    let l:req_id = s:inst_req_id
+    let s:inst_req_id += 1
+
     " while the user is providing an instruction, send a warm-up request
     let l:messages = llama#inst_build(l:l0, l:l1, '')
 
     let l:request = {
+        \ 'id_slot':      l:req_id,
         \ 'messages':     l:messages,
         \ 'samplers':     [],
         \ 'n_predict':    0,
@@ -1424,10 +1429,6 @@ function! llama#inst(l0, l1)
 
     call llama#debug_log('inst_send | ' . l:inst)
 
-    " Create request state
-    let l:req_id = s:inst_req_id
-    let s:inst_req_id += 1
-
     let l:bufnr = bufnr('%')
 
     let l:req = {
@@ -1470,6 +1471,7 @@ function! llama#inst_send(req_id, messages)
     call llama#debug_log('inst_send', join(a:messages, "\n"))
 
     let l:request = {
+        \ 'id_slot':      a:req_id,
         \ 'messages':     a:messages,
         \ 'min_p':        0.1,
         \ 'temperature':  0.1,