Hey there,
I got the inline completions working with my OpenAI API Key.
I tried using multiple models (davinci, 3.5-turbo, 4-turbo) and they all produce really bad completions (usually way too long or what seems like copying from some API documentation).
When trying to use it inside parenthesis it doesn’t suggest anything at all.
I don’t know what’s needed to debug but I’ll provide whatever needed.
Example of completion -
Input code (what I typed then stopped typing) - px.li (I imported plotly.express as px)
Suggested completion -
px.line(data, label="data")
px.scatter(data, label="data")
px.box(data, label="data")
px.histogram(data, label="data")
px.box(data, label="data", color="color")
px.box(data, label="data", color="color", aspect_ratio=0.5)
px.box(data, label="data", color="color", aspect_ratio=0.5, color_range="color")
px.box(data, label="data", color="color", aspect_ratio=0.5, color_range="color", range_x=[0, 1])
px.box(data, label="data", color="color", aspect_ratio=0.5, color_range="color", range_x=[0, 1], range_y=[0, 1])
The above code produces a chart that looks like this:
import matplotlib.pyplot as plt
# Plotting 3 discrete distributions.
# Discrete distributions are performed by px.box.
# range_x=[0, 1] and range_y=[0, 1] allows for plotting the data
# within a range of [0, 1] on the x and y axis.
# color_range="color" and color="color" allow for color selection.
# aspect
When using GitHub copilot it always suggests something reasonable like a line plot with dummy data.