From 99b71c068f624521ad977e08e41589e2971fa1c7 Mon Sep 17 00:00:00 2001
From: Xuan Son Nguyen <thichthat@gmail.com>
Date: Wed, 13 Mar 2024 11:39:11 +0100
Subject: Server: Use multi-task for embeddings endpoint (#6001)

* use multitask for embd endpoint

* specify types

* remove redundant {"n_predict", 0}
---
 examples/server/utils.hpp | 12 +++++++++++-
 1 file changed, 11 insertions(+), 1 deletion(-)

(limited to 'examples/server/utils.hpp')

diff --git a/examples/server/utils.hpp b/examples/server/utils.hpp
index 48aeef4e..2ddb2cd2 100644
--- a/examples/server/utils.hpp
+++ b/examples/server/utils.hpp
@@ -529,6 +529,16 @@ static std::vector<json> format_partial_response_oaicompat(json result, const st
 }
 
 static json format_embeddings_response_oaicompat(const json & request, const json & embeddings) {
+    json data = json::array();
+    int i = 0;
+    for (auto & elem : embeddings) {
+        data.push_back(json{
+            {"embedding", json_value(elem, "embedding", json::array())},
+            {"index",     i++},
+            {"object",    "embedding"}
+        });
+    }
+
     json res = json {
         {"model", json_value(request, "model", std::string(DEFAULT_OAICOMPAT_MODEL))},
         {"object", "list"},
@@ -536,7 +546,7 @@ static json format_embeddings_response_oaicompat(const json & request, const jso
             {"prompt_tokens", 0},
             {"total_tokens", 0}
         }},
-        {"data", embeddings}
+        {"data", data}
     };
 
     return res;
-- 
cgit v1.2.3