Forked from nbbaier/sqliteWriter
Readme
This val provides a class ReadmeWriter
for generating readmes for vals with OpenAI. It can both draft readmes and update them directly
PRs welcome! See Todos below for some ideas I have.
To draft a readme for a given code, use the draftReadme
method:
import { ReadmeWriter } from "https://esm.town/v/nbbaier/readmeGPT";
const readmeWriter = new ReadmeWriter({});
const val = "https://www.val.town/v/:username/:valname";
const generatedReadme = await readmeWriter.draftReadme(val);
To write and update a readme for a given code, use the writeReadme
method:
import { ReadmeWriter } from "https://esm.town/v/nbbaier/readmeGPT";
const readmeWriter = new ReadmeWriter({});
const val = "https://www.val.town/v/:username/:valname";
const successMessage = await readmeWriter.writeReadme(val);
The ReadmeWriter
class represents a utility for generating and updating README files.
Creates an instance of the ReadmeWriter
class.
model
(optional): The model to be used for generating the readme. Defaults to "gpt-3.5-turbo".apiKey
(optional): An OpenAI API key. Defaults toDeno.env.get("OPENAI_API_KEY")
.
-
draftReadme(val: string): Promise<string>
: Generates a readme for the given val.-
Parameters:
val
: URL of the code repository.
-
Returns:
- A promise that resolves to the generated readme.
-
-
writeReadme(val: string): Promise<string>
: Generates and updates a readme for the given val.-
Parameters:
val
: URL of the code repository.
-
Returns:
- A promise that resolves to a success message if the update is successful.
-
- Additional options to pass to the OpenAI model
- Ability to pass more instructions to the prompt to modify how the readme is constructed
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
import { type WriterOptions } from "https://esm.town/v/nbbaier/WriterOptions";
import { fetch } from "https://esm.town/v/std/fetch?v=4";
import OpenAI, { type ClientOptions } from "npm:openai";
export class ReadmeWriter {
model: string;
openai: OpenAI;
apiKey: string;
valtownKey: string;
constructor(options: WriterOptions) {
const { model, ...openaiOptions } = options;
this.model = model ? model : "gpt-3.5-turbo";
this.openai = new OpenAI(openaiOptions);
this.valtownKey = Deno.env.get("valtown");
}
private createPrompt(code: string, userPrompt?: string) {
return `
You are an AI assistant that writes documentation for code. You output readmes
in GitHub flavored markdown. Usage sections should include a single code snippet
that a user can copy and paste. Never return anything other than documentation for
the code you are provided.
${userPrompt}
Take the below code and return a markdown readme:
${code}
`;
}
private async getVal(username: string, valName: string) {
try {
const res = await fetch(`https://api.val.town/v1/alias/${username}/${valName}`, {
method: "GET",
headers: {
"accept": "*/*",
"Content-Type": "application/json",
"Authorization": `Bearer ${this.valtownKey}`,
},
});
const { id, code } = await res.json();
return { id, code };
} catch (error) {
throw new Error("Error getting val code: " + error.message);
}
}
private async performOpenAICall(prompt: string) {
try {
const response = await this.openai.chat.completions.create({
messages: [{ role: "system", content: prompt }],
model: this.model,
});
if (!response.choices || response.choices.length === 0) {
throw new Error("No response from OpenAI");
}
const readme = response.choices[0].message?.content;
if (!readme) {
throw new Error("No readme returned by OpenAI. Try again.");
}
return readme;
} catch (error) {
throw new Error("Error generating readme: " + error.message);
}
}
private async updateReadme(id: string, readme: string) {
try {
const res = await fetch(`https://api.val.town/v1/vals/${id}`, {
method: "PUT",
headers: {
"accept": "*/*",
"Content-Type": "application/json",
"Authorization": `Bearer ${this.valtownKey}`,
},
body: JSON.stringify({ "readme": readme }),
});
return res.status;
} catch (error) {
throw new Error("Error updating readme: " + error.message);
}
}
private async processRequest(val: string, userPrompt?: string) {
const url = new URL(val);
const [, _, username, valName] = url.pathname.split("/");
const { id, code } = await this.getVal(username, valName);
const prompt = this.createPrompt(code, userPrompt);
const readme = await this.performOpenAICall(prompt);
return { id, readme };
}
async draftReadme(val: string, userPrompt?: string) {
const { readme } = await this.processRequest(val, userPrompt);
return readme;
👆 This is a val. Vals are TypeScript snippets of code, written in the browser and run on our servers. Create scheduled functions, email yourself, and persist small pieces of data — all from the browser.