2023-09-11 19:06:01 -07:00

97 lines
4.1 KiB
YAML

# This YAML file contains regression questions for Danswer.
# The sources mentioned are the same ones to power the DanswerBot for the community's use
# The regression flow assumes the data from the sources listed are already indexed
metadata:
version: v0.0.1
date: 2023-09-10
sources:
- name: web
detail: https://www.danswer.ai/
- name: web
detail: https://docs.danswer.dev/
- name: github issues
detail: danswer-ai/danswer
- name: github pull-requests
detail: danswer-ai/danswer
- name: slack
workspace: danswer.slack.com
- name: file
detail: Markdown files from Danswer repo
questions:
- id: 1
question: "What is Danswer?"
expected_answer: "Danswer is an open source question-answering system."
notes: "This comes directly from the docs, the actual answer should be more informative"
- id: 2
question: "What is Danswer licensed under?"
expected_answer: "Danswer is MIT licensed"
notes: "This info can be found in many places"
- id: 3
question: "What are the required variables to set to use GPT-4?"
expected_answer: "Set the environment variables INTERNAL_MODEL_VERSION=openai-chat-completion and GEN_AI_MODEL_VERSION=gpt-4"
notes: "Two env vars are must have, the third (the key) is optional"
- id: 4
question: "Why might I want to use the deberta model for QnA?"
expected_answer: "This kind of model can run on CPU and are less likely to produce hallucinations"
notes: "https://docs.danswer.dev/gen_ai_configs/transformers, this is a pretty hard question"
- id: 5
question: "What auth related tokens do I need for BookStack?"
expected_answer: "You will need the API Token ID and the API Token Secret"
notes: "https://docs.danswer.dev/connectors/bookstack"
- id: 6
question: "ValueError: invalid literal for int() with base 10"
expected_answer: "This was a bug that was fixed shortly after the issue was filed. Try updating the code."
notes: "This question is in Github Issue #290"
- id: 7
question: "Is there support for knowledge sets or document sets?"
expected_answer: "This was requested and approved however it is not clear if the feature is implemented yet."
notes: "This question is in Github Issue #338"
- id: 8
question: "nginx returning 502"
expected_answer: "Google OAuth must be configured for Danswer backend to work. A PR was created to fix it"
notes: "This question is in Github Issue #260"
- id: 9
question: "Why isn't GPT4All enabled by default"
expected_answer: "There is no recent version of GPT4All that is compatible with M1 Mac."
notes: "This question is in Github Issue #232 but also mentioned in several other places"
- id: 10
question: "Why isn't GPT4All enabled by default"
expected_answer: "There is no recent version of GPT4All that is compatible with M1 Mac."
notes: "This question is in Github Issue #232 but also mentioned in several other places"
- id: 11
question: "Why are the models warmed up on server start"
expected_answer: "This ensures that the first indexing isn't really slow."
notes: "This is in Github PR #333"
- id: 12
question: "Why are the models warmed up on server start"
expected_answer: "This ensures that the first indexing isn't really slow."
notes: "This is in Github PR #333"
- id: 13
question: "What text from the Alation Connector is used to generate the docs?"
expected_answer: "Articles are used with the body contents. Schemas, Tables, and Columns use Description"
notes: "This is in Github PR #161"
- id: 14
question: "Does Danswer support PDFs in Google Drive?"
expected_answer: "Yes"
notes: "This question is in Slack, if the message expires due to using free slack version, the info may be gone as well"
- id: 15
question: "I deleted a connector in Danswer but some deleted docs are still showing in search"
expected_answer: "The issue was fixed via a code change, it should go away after pulling the latest code"
notes: "This question is in Slack, if the message expires due to using free slack version, the info may be gone as well"