From 66a2a0d47dbff005e520f0634ab628a40c1a247b Mon Sep 17 00:00:00 2001
From: Yingbei <yingbei@acorn.io>
Date: Mon, 17 Jun 2024 17:17:00 -0700
Subject: [PATCH] updated readme, fix function call postprocess

---
 README.md                                | 33 +++++++++++++++++-------
 examples/server/function-call-parser.hpp | 12 +++++----
 examples/server/utils.hpp                |  2 +-
 3 files changed, 32 insertions(+), 15 deletions(-)

diff --git a/README.md b/README.md
index 6c24135d61934..7150d6f00b158 100644
--- a/README.md
+++ b/README.md
@@ -1,17 +1,32 @@
-# llama.cpp
+# tools.cpp
 
-![llama](https://user-images.githubusercontent.com/1991296/230134379-7181e485-c521-4d23-a0d6-f7b3b61ba524.png)
+### tools.cpp quickstart
+1. build from source:
+Mac user
+```
+make
+```
 
-[![License: MIT](https://img.shields.io/badge/license-MIT-blue.svg)](https://opensource.org/licenses/MIT)
-[![Server](https://github.com/ggerganov/llama.cpp/actions/workflows/server.yml/badge.svg?branch=master&event=schedule)](https://github.com/ggerganov/llama.cpp/actions/workflows/server.yml)
-[![Conan Center](https://shields.io/conan/v/llama-cpp)](https://conan.io/center/llama-cpp)
+Nvidia-Cuda user:
+```
+make LLAMA_CUDA=1
+```
 
-[Roadmap](https://github.com/users/ggerganov/projects/7) / [Project status](https://github.com/ggerganov/llama.cpp/discussions/3471) / [Manifesto](https://github.com/ggerganov/llama.cpp/discussions/205) / [ggml](https://github.com/ggerganov/ggml)
+2. Install helper package:
+```
+npm install jsonrepair
+```
 
-Inference of Meta's [LLaMA](https://arxiv.org/abs/2302.13971) model (and others) in pure C/C++
+3. Download a compatible gguf model:
+For example:
+```
+wget https://huggingface.co/sanjay920/Llama-3-8b-function-calling-alpha-v1.gguf/resolve/main/Llama-3-8b-function-calling-alpha-v1.gguf
+```
 
-> [!IMPORTANT]
-[2024 Jun 12] Binaries have been renamed w/ a `llama-` prefix. `main` is now `llama-cli`, `server` is `llama-server`, etc (https://github.com/ggerganov/llama.cpp/pull/7809)
+4. start server:
+```
+./llama-server -ngl 35 -m Llama-3-8b-function-calling-alpha-v1.gguf   --port 1234 --host 0.0.0.0  -c 16000 --chat-template llama3
+```
 
 ### Recent API changes
 
diff --git a/examples/server/function-call-parser.hpp b/examples/server/function-call-parser.hpp
index a9e8cc297910c..e131c96326f59 100644
--- a/examples/server/function-call-parser.hpp
+++ b/examples/server/function-call-parser.hpp
@@ -68,7 +68,6 @@ json clean_json_strings(const std::string& input_str) {
         // json repair here
         std::string fixed_str = jsonrepair(input_str);
         json data = json::parse(fixed_str);
-
         for (auto& [key, value] : data.items()) {
             if (value.is_string()) {
                 std::string val = value.get<std::string>();
@@ -82,6 +81,7 @@ json clean_json_strings(const std::string& input_str) {
                     if (v.is_string()) {
                         v = clean_command_string(v.get<std::string>());
                     }
+                    
                 }
             }
         }
@@ -97,7 +97,7 @@ json clean_json_strings(const std::string& input_str) {
 
 std::vector<json> rubra_fc_json_tool_extractor(const std::string& output_str) {
     std::vector<json> result;
-    printf("OUTPUT STR TO BE PARSED : %s", output_str.c_str());
+    printf("OUTPUT STR TO BE PARSED : %s\n", output_str.c_str());
     if (output_str.find("endtoolcall") == std::string::npos) {
         return result;
     }
@@ -111,7 +111,8 @@ std::vector<json> rubra_fc_json_tool_extractor(const std::string& output_str) {
         size_t pos = segment.find("starttoolcall");
         if (pos != std::string::npos) {
             // Extract substring after "toolcall"
-            listOfStrToParse.push_back(segment.substr(pos + std::string("starttoolcall").length()));
+            std::string ss = segment.substr(pos + std::string("starttoolcall").length());
+            listOfStrToParse.push_back(ss);
         }
         start = end + std::string("endtoolcall").length();  // Move past the "endtoolcall"
     }
@@ -121,9 +122,10 @@ std::vector<json> rubra_fc_json_tool_extractor(const std::string& output_str) {
     try {
         for (const auto & line : listOfStrToParse) {
             // json fc = json::parse(line);
+            
             json fc = clean_json_strings(line);
-            if (fc["arguments"].is_string()) {
-                fc["arguments"] = json::parse(fc["arguments"].get<std::string>());
+            if (!fc["arguments"].is_string()) {
+                fc["arguments"] = fc["arguments"].dump();
             }
             if (!fc.is_null()) {
                 function_call_json.push_back(fc);
diff --git a/examples/server/utils.hpp b/examples/server/utils.hpp
index 4e91eb9e04ecd..75ad38bb061e5 100644
--- a/examples/server/utils.hpp
+++ b/examples/server/utils.hpp
@@ -474,7 +474,7 @@ static json format_final_response_oaicompat(const json & request, json result, c
 
                 tool_call["function"] = json{
                     {"name" , pc["name"]},
-                    {"arguments" , pc["kwargs"].dump()},
+                    {"arguments" , pc["kwargs"]},
                 };
                 oai_format_tool_calls.push_back(tool_call);
             }