Task Classifier Example
Three versions
- 01-no-humans.ts - classify the emails with no human intervention
- 02-humans.ts - classify the emails, checking with a human before saving classifications, then print results
- 03-humans-async.ts - classify the emails, print them out, then start a webserver to listen for human overrides. When a human feedback is received, print the updated list.
Running the examples
npm install
npm run no-humans
npm run human-review-sync
npm run human-review-async
For all three examples, you'll need to set OPENAI_API_KEY in your environment.
For the human-review examples, you'll need to set HUMANLAYER_API_KEY in your environment. You can get one at app.humanlayer.dev.
For the human-review-async example, you will need to configure HumanLayer to send a Response Webhook to your local server using ngrok or similar.
01-no-humans.ts
In this example, we just classify the emails and print the results of the LLM classification.
02-human-review-sync.ts
In this example, each email is sent synchronously to a human for review, and then all results are printed at the end.
03-human-review-async.ts
In this example, all LLM classifications are computed, and then they are all sent to a human for review. When a human review is completed, it will be received by a webhook, and the results will be printed out.
