add prefix caching for deepseek

lua/kznllm/presets.lua

@@ -63,6 +63,35 @@ local function make_data_for_openai_chat(prompt_args, opts)
   return data
 end
 
+local function make_data_for_deepseek_chat(prompt_args, opts)
+  local messages = {
+    {
+      role = 'system',
+      content = kznllm.make_prompt_from_template(opts.template_directory / 'nous_research/fill_mode_system_prompt.xml.jinja', prompt_args),
+    },
+    {
+      role = 'user',
+      content = kznllm.make_prompt_from_template(opts.template_directory / 'nous_research/fill_mode_user_prompt.xml.jinja', prompt_args),
+    },
+  }
+
+  local data = {
+    messages = messages,
+    model = opts.model,
+    stream = true,
+  }
+  if M.PROMPT_ARGS_STATE.replace and opts.prefill and opts.stop_param then
+    table.insert(messages, {
+      role = 'assistant',
+      content = opts.prefill .. prompt_args.current_buffer_filetype .. '\n',
+      prefix = true,
+    })
+  end
+  data = vim.tbl_extend('keep', data, opts.data_params, opts.stop_param or {})
+
+  return data
+end
+
 ---Example implementation of a `make_data_fn` compatible with `kznllm.invoke_llm` for anthropic spec
 ---@param prompt_args any
 ---@param opts any
@@ -308,14 +337,16 @@ local presets = {
   {
     id = 'chat-model',
     provider = 'deepseek',
-    make_data_fn = make_data_for_openai_chat,
+    make_data_fn = make_data_for_deepseek_chat,
     debug_fn = openai_debug_fn,
     opts = {
       model = 'deepseek-coder',
       data_params = {
         max_tokens = 8192,
         temperature = 0,
       },
+      stop_param = { stop = { '```' } },
+      prefill = '```',
       debug_fn = openai_debug_fn,
       base_url = 'https://api.deepseek.com',
       endpoint = '/beta/v1/chat/completions',