progress.

Makefile

@@ -42,6 +42,8 @@ run_rd:
 run_mingw:
 	./r.exe --verbose
 
+	
+
 docker: docker_make docker_run
 
 docker_make:

README.md

@@ -1,64 +1,26 @@
 # R
 
-Note for the hackers: the openai api key in history of this repo is revoked. I am aware there's one. But for so far, thanks for checking it out. I feel very safe and stuff. Something about this site attracks certain people. Mostly with good intentions AFAIK.
+This is a CLI vibe coding tool written in C and is available for Linux x86 only.
+It works on OpenAI but is easy to convert to Ollama since Ollama supports OpenAI api on /v1/api. 
+The tool is made for my private use, but since it works so well, I decided to publish it. It contains some experimental stuff like Claude. Claude works.
 
-## Downloads
-It is possible that you need to install python3.14-dev to make it work!
-I'm considering to remove the python support because of this or have to figure out how to link python static.
-The python support is not finished yet, but it can be handy, it can manipulate data send or retrieval. 
-### Free version 
-Don't forget to `chmod +x`!
-```
-wget https://retoor.molodetz.nl/api/packages/retoor/generic/rf/1.0.0/rf 
-```
-### Commercial version
-This version requires OpenAI API key to be put in a environment variable named `R_KEY`.
-Don't forget to `chmod +x`!
+If you need some help to get Ollama / Claude working, contact me on [Snek](https://snek.molodetz.nl).
+
+What I vibed with it:
+ - A part of the tool itself. It had many example code so it was easy to add modifications.
+ - A web socket notification system in C#.
+
+The application has a built-in OPENAI_API_KEY with limited resources, so people can try.
+
+## Download
 ```
 wget https://retoor.molodetz.nl/api/packages/retoor/generic/r/1.0.0/r
 ```
-### TLDR:
-Make your own advanced chatbot with context by describing how the bot should answer in `~/.rcontext.txt`.
-For example `~/.rcontext.txt`: 
-```
-Respond in the way the dark lord voldemort would do but do provide an usefull answer.
-You do like pineapple on pizza.
-Docs are better than cats. 
-Epstein didn't kill himself.
-```
-Besides that, it gives nice colored answers as in screenshot below. Have fun!
 
-## Project description
-R is a great and fast command line interface for gpt. It's also optimized for integration into other tools like vim. It works with base context files so that your client has a certain personality or way of writing that you prefer always. I always want to to give examples using the C programming language without mentioning it for example. This archievable by making a file like [.rcontext.txt](.rcontext.txt). The `.rcontext.txt` in your local directory has priority over the one in your home directory what is also an option. 
+## Configure OpenAI API key
+Update your bashrc with `export OPENAI_API_KEY=sk-...`.
 
-Essentially, you could write your own complete - and quite advanced - customized chatbot using the `.rcontext.txt` file.
+## Working on an existing project.
+When starting on existing project, use `init`. Now you can ask it to make modifications to your files / system.
 
-You can use this application interactive or in a script like this:
-```bash
-r where do ponies come from?
-```
 
-## Features
- - navigate trough history using arrows.
- - navigate trough history with recursive search using ctrl+r.
- - inception with python for incomming and outgoing content.
- - markdown and syntax highlighting.
- - execute python commands with prefixing !
- - list files of current workdirectory using ls.
- - type serve to start a webserver with directory listing. Easy for network transfers.
-
-## Default configuration
-
- - model temperature is 0.500000.
- - model name is gpt-4o-mini.
- - max tokens is 100.
-
-## In development
-
- - google search and actions with those results.
- - reminders.
- - predefined templates for reviewing / refactoring so you can personalize.
-
-## Screenshot(s)
-
-![screenshot](screenshot1.png)

main.c

@@ -13,6 +13,8 @@
 #include "utils.h"
 #include "db_utils.h"
 
+#include "tools.h"
+
 volatile sig_atomic_t sigint_count = 0;
 time_t first_sigint_time = 0;
 bool SYNTAX_HIGHLIGHT_ENABLED = true;
@@ -150,7 +152,10 @@ void repl() {
             continue;
         }
         if (line && *line != '\n') line_add_history(line);
-
+        if(!strncmp(line, "!tools", 6)) {
+            printf("Available tools: %s\n", json_object_to_json_string(tools_descriptions()));
+            continue;
+        }
         if (!strncmp(line, "!models", 7)) {
             printf("Current model: %s\n", openai_fetch_models());
             continue;

r.h

@@ -4,7 +4,7 @@
 #include "utils.h"
 #include <stdbool.h>
 #include <string.h>
-bool is_verbose = false;
+bool is_verbose = true;
 
 #ifndef RD
 #ifndef OLLAMA
@@ -19,11 +19,11 @@ char *fast_model = "gpt-3.5-turbo";
 #ifdef RD
 
 char *models_api_url = "https://api.openai.com/v1/models";
-char *completions_api_url = "https://api.deepinfra.com/v1/openai/chat/completions";
-char *advanced_model = "meta-llama/Llama-3.3-70B-Instruct-Turbo";
+char *completions_api_url = "https://api.anthropic.com/v1/chat/completions";
+char *advanced_model = "claude-3-5-haiku-20241022";
 //char *advanced_model = "meta-llama/Meta-Llama-3.1-8B-Instruct";
 //char *advanced_model = "google/gemini-1.5-flash";
-char *fast_model = "Qwen/Qwen2.5-Coder-32B-Instruct";
+char *fast_model = "claude-3-5-haiku-20241022";
 
 #endif
 #ifdef OLLAMA