0.1.8 • Published 3 months ago
finetunedb v0.1.8
FinetuneDB Node API Library
This library wraps TypeScript or Javascript OpenAI API calls and logs additional data to your FinetuneDB account.
It is fully compatible with OpenAI's sdk and logs both streaming and non-streaming requests and responses.
Installation
npm install --save finetunedb
# or
yarn add finetunedb
Usage
- Create a workspace at https://app.finetunedb.com
- Generate an API key from the workspace settings
- Find the project ID you want to log to
- Configure the FinetuneDB client as shown below
// import OpenAI from 'openai'
import OpenAI from "finetunedb/openai";
// Fully compatible with original OpenAI initialization
const openai = new OpenAI({
apiKey: "your api key", // defaults to process.env["OPENAI_API_KEY"]
// Optional: Step 1 - Initialize FinetuneDB to enable logging
finetunedb: {
apiKey: "your api key", // defaults to process.env["FINETUNEDB_API_KEY"]
},
});
async function main() {
// Allows optional finetunedb object
const completion = await openai.chat.completions.create({
messages: [{ role: "user", content: "Hello world!" }],
model: "gpt-3.5-turbo",
// Optional: Step 2 - Log a chat completion
finetunedb: {
// Define the project ID
projectId: "cloio7t90000...",
// Enable/disable data collection. Defaults to true.
logRequest: true,
// Optional: Add custom searchable tags
tags: ["test-prompt"],
},
});
// Optional: Update log
await completion.finetunedb.updateLastLog({
tags: ["test-prompt", "new-tag"]
})
}
main();
FAQ
What is the difference between the apiKey
and the projectId
?
The apiKey
is shared across a workspace, and each workspace contains multiple projects with a unique projectId
.
Imagine your app uses multiple fine-tuned models for specific tasks, each tasks could be defined by a project.
It's good practice to fine-tune task specific models.
Will OpenAI calls fail if the FinetuneDB client is misconfigured or the platform is down?
Your OpenAI calls will continue to function as expected no matter what. The sdk handles logging errors gracefully without affecting OpenAI inference.