-
Notifications
You must be signed in to change notification settings - Fork 1
/
README.Rmd
154 lines (99 loc) · 7.32 KB
/
README.Rmd
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
---
output: github_document
---
# mergen Studio <img src="man/figures/mergenstudio.png" align="right" height="200" style="float:right; height:200px;"/>
<!-- README.md is generated from README.Rmd. Please edit that file -->
```{r setup, include=FALSE}
knitr::opts_chunk$set(echo = TRUE)
```
mergen Studio is an RStudio Addin wrapper for the [mergen](https://github.com/BIMSBbioinfo/mergen) package.
You can find more information on mergen can be found on its website:\
[https://bioinformatics.mdc-berlin.de/mergen/](https://bioinformatics.mdc-berlin.de/mergen/)
## Installation
Install from the GitHub repository using devtools (with R version 4.0.0 or higher):
``` {r,eval=FALSE}
if (!require("devtools", quietly = TRUE))
install.packages("devtools")
devtools::install_github("BIMSBbioinfo/mergenstudio")
```
#### Configuring your AI API key
To interact with an AI API, you require a valid AI API key. To configure your key so that it is present globally in your environment at all times, you can include it in your .Renviron file. This will ensure that the key is automatically loaded.
For setting up mergen Studio, this variable should be called `AI_API_KEY`.
__Caution:__ If you’re using version control systems like GitHub, include .Renviron in your .gitignore file to prevent exposing your personal API key.
Here is how to open your .Renviron file for modification in your project:
```{r,eval=FALSE}
require(usethis)
edit_r_environ(scope="project")
```
For a persistent loading of your API key, add the following line to your .Renviron file replacing `"your_key"` with your key.
```{r,eva=FALSE}
AI_API_KEY="your_key"
```
__NOTE:__ After setting up your API key in the .Renviron file, either restart the R session or run `readRenviron(".Renviron")` to apply the changes.
If you however wish to set this variable only for a single session, you can use the following command:
```{r,eval=FALSE}
Sys.setenv(AI_API_KEY="your_key")
```
mergen Studio also allows you to set your key once in the app interface. Simply paste your
key into the API key box on the side panel under API options.
<img src="man/figures/setting_up_key.png" align="middle" height="600" style="float:middle; height:600px;"/>
## Running mergen Studio
To run mergen Studio, open an R session and type in the following command:
```{r,eval=FALSE}
mergenstudio::run_chat_app()
```
This will open the chat interface. Press the arrow on the top left of the screen to
open the sidebar.
<img src="man/figures/open_sidebar.png" align="middle" height="600" style="float:middle; height:600px;"/>
After this, you can choose which API service to use,
and change/load in your API key if needed. This can be done in the API options menu.
When using Rstudio, mergen Studio also allows you to open the chat interface in the Viewer.
To do this, press on the Addins button, and select `Chat`.
<img src="man/figures/addin_click.png" align="middle" height="600" style="float:middle; height:600px;"/>
<img src="man/figures/addin_works.png" align="middle" height="600" style="float:middle; height:600px;"/>
When making use of `openai` or `replicate` API
services, after setting your API key, all that is needed is to select a model to chat with, and start chatting!
When setting the API service to `generic`, you are required to fill in an `API URL` and
fill in the chat model you wish to use by pasting in the name of the
model you wish to use into the `Chat Model` box. Please make sure that chat model and API URL
are filled in correctly. Usually you can find more information about the API URL and model name syntax
on the website of your API service of choice.
<img src="man/figures/open_API_service.png" align="middle" height="600" style="float:middle; height:600px;"/>
<img src="man/figures/generic_service.png" align="middle" height="600" style="float:middle; height:600px;"/>
## Saving default settings
In the video below, we show you how to open a chat via mergenstudio, and save some default settings which will persist for subsequent chat sessions.
[](https://www.youtube.com/watch?v=--5NL9NUp6c)
## code execution
Once you have started chatting with your LLM of choice, any code that might be generated
by the LLM can be run directly in the chat interface. To run code from a response, simply press
the 'execute code' button on top of the response. This will run the code, and show you the output.
<img src="man/figures/code_example.png" align="middle" height="600" style="float:middle; height:600px;"/>
<img src="man/figures/code_ran.png" align="middle" height="600" style="float:middle; height:600px;"/>
To automatically run any code without having to press the `execute code` button, turn on the option `Activate auto execution`
from the `Chat options` panel.
<img src="man/figures/auto_execution.png" align="middle" height="600" style="float:middle; height:600px;"/>
Below you can find a video tutorial on how mergenstudio can be used to read in data from a file, and run code on it.
[](https://www.youtube.com/watch?v=AngJxS6ZWSc)
## Advanced settings
In mergen Studio it is possible to make use of some more advanced settings to alter the
prompting strategy and improve responses. These advanced settings include `Select context`, `Activate self correct`
and `Activate file header addition`
### Select context
Selecting a context allows you to send a specific context with every question that is sent to the LLM,
as a way of prompt engineering. Mergen contains the following predefined contexts: Simple, ActAs,rbionfoExp and CoT.
Users can also send their own context by simply typing into the box and clicking add. This will add the user context to
the list of possible contexts to choose from (see pictures below). If you wish to not send any context to the prompt
at all, you can select `No context`. This is not recommended however, since the LLM then might provide code examples in
languages other then R.
<img src="man/figures/context.png" align="middle" height="600" style="float:middle; height:600px;"/>
<img src="man/figures/typing_own_context.png" align="middle" height="600" style="float:middle; height:600px;"/>
### Activate self correct
Activating Self Correct will attempt to correct code that is returned by the agent if it results in errors,
by resending the prompt together with additional information about the error message. To make use of this feature,
you can change the setting to `Yes`. Activating self-correct can also be done after an error message is received. Below you can find a video on how mergenstudio can be used to perform complex data analysis tasks and also can be used to correct for errors in the code.
[](https://www.youtube.com/watch?v=W-8KTS4XnJk)
### Activate file header addition
Activating file header addition will attempt to add the first few lines of files mentioned in your prompt (if there are any) to your prompt. To make use of this feature,
you can change the setting to `Yes`. The directory that will used to search for your files can be set under `Chat options`
`Select directory`. Below you can find a video tutorial on how file header addition works:
[](https://www.youtube.com/watch?v=z7DrSjso89E)