From f613a91b909d396fa74d87b1fd0973c76069a82e Mon Sep 17 00:00:00 2001
From: tikikun <daogiatuank54@gmail.com>
Date: Fri, 17 Nov 2023 14:21:27 +0700
Subject: [PATCH 1/2] embedding that is compatible with openai

---
 controllers/llamaCPP.cc | 50 ++++++++++++++++++++++++++++++++++++-----
 1 file changed, 45 insertions(+), 5 deletions(-)

diff --git a/controllers/llamaCPP.cc b/controllers/llamaCPP.cc
index e3901df60..e01bd9a46 100644
--- a/controllers/llamaCPP.cc
+++ b/controllers/llamaCPP.cc
@@ -27,6 +27,45 @@ std::shared_ptr<State> createState(int task_id, llamaCPP *instance) {
 
 // --------------------------------------------
 
+#include <ctime>
+#include <json/json.h>
+#include <string>
+#include <vector>
+
+std::string create_embedding_payload(const std::vector<float> &embedding,
+                                     int prompt_tokens) {
+  Json::Value root;
+
+  root["object"] = "list";
+
+  Json::Value dataArray(Json::arrayValue);
+  Json::Value dataItem;
+
+  dataItem["object"] = "embedding";
+
+  Json::Value embeddingArray(Json::arrayValue);
+  for (const auto &value : embedding) {
+    embeddingArray.append(value);
+  }
+  dataItem["embedding"] = embeddingArray;
+  dataItem["index"] = 0;
+
+  dataArray.append(dataItem);
+  root["data"] = dataArray;
+
+  root["model"] = "_";
+
+  Json::Value usage;
+  usage["prompt_tokens"] = prompt_tokens;
+  usage["total_tokens"] = prompt_tokens; // Assuming total tokens equals prompt
+                                         // tokens in this context
+  root["usage"] = usage;
+
+  Json::StreamWriterBuilder writer;
+  writer["indentation"] = ""; // Compact output
+  return Json::writeString(writer, root);
+}
+
 std::string create_full_return_json(const std::string &id,
                                     const std::string &model,
                                     const std::string &content,
@@ -245,17 +284,18 @@ void llamaCPP::embedding(
   const auto &jsonBody = req->getJsonObject();
 
   json prompt;
-  if (jsonBody->isMember("content") != 0) {
-    prompt = (*jsonBody)["content"].asString();
+  if (jsonBody->isMember("input") != 0) {
+    prompt = (*jsonBody)["input"].asString();
   } else {
     prompt = "";
   }
   const int task_id = llama.request_completion(
       {{"prompt", prompt}, {"n_predict", 0}}, false, true);
   task_result result = llama.next_result(task_id);
-  std::string embeddingResp = result.result_json.dump();
+  std::vector<float> embedding_result = result.result_json["embedding"];
   auto resp = nitro_utils::nitroHttpResponse();
-  resp->setBody(embeddingResp);
+  std::string embedding_resp = create_embedding_payload(embedding_result, 0);
+  resp->setBody(embedding_resp);
   resp->setContentTypeString("application/json");
   callback(resp);
   return;
@@ -363,7 +403,7 @@ void llamaCPP::loadModel(
   llama.initialize();
 
   Json::Value jsonResp;
-  jsonResp["message"] = "Failed to load model";
+  jsonResp["message"] = "Model loaded successfully";
   model_loaded = true;
   auto resp = nitro_utils::nitroHttpJsonResponse(jsonResp);
 

From 6aa879bb412190b865fef3284e29f1fe4f0c6386 Mon Sep 17 00:00:00 2001
From: tikikun <daogiatuank54@gmail.com>
Date: Fri, 17 Nov 2023 14:22:52 +0700
Subject: [PATCH 2/2] embedding that is compatible with openai

---
 controllers/llamaCPP.cc | 5 -----
 1 file changed, 5 deletions(-)

diff --git a/controllers/llamaCPP.cc b/controllers/llamaCPP.cc
index e01bd9a46..63c032657 100644
--- a/controllers/llamaCPP.cc
+++ b/controllers/llamaCPP.cc
@@ -27,11 +27,6 @@ std::shared_ptr<State> createState(int task_id, llamaCPP *instance) {
 
 // --------------------------------------------
 
-#include <ctime>
-#include <json/json.h>
-#include <string>
-#include <vector>
-
 std::string create_embedding_payload(const std::vector<float> &embedding,
                                      int prompt_tokens) {
   Json::Value root;