https://www.gnu.org/software/bash/manual/bash.html#Redirections https://www.gnu.org/savannah-checkouts/gnu/bash/manual/bash.html#Process-Substitution

This directory contains scripts for interacting with the Copilot API.

source "$(dirname "$(readlink -f "${BASH_SOURCE[0]}")")/../deps.bash"

This script handles authentication and login with the Copilot API. It should be run to ensure your session is authenticated before using other curlpilot scripts that require API access. Upon successful login, it saves the Personal Access Token (PAT) to $HOME/.config/curlpilot/github_pat.txt and the Copilot session token to $HOME/.config/curlpilot/token.txt.

• Personal Access Token (PAT): A PAT is an alternative to using your password for authentication to GitHub (or other services). It's a string of characters that grants specific permissions to your account.
• Session Token: A session token is a temporary credential that authenticates your current session with the Copilot API, allowing subsequent requests to be made without re-authenticating.

This script is designed to send git diffs to Copilot for review. It takes a literal text message as an argument, which is then passed to Copilot. This message should be a plain, descriptive string.

Example usage:

./chat.bash "Tell me one fun animal fact."

Example Response:

Sea otters hold hands while they sleep so they don’t drift apart from each other in the water!

• Portable conversations: All conversation data is stored locally, allowing portability across systems by simply syncing the relevant files.
• Search previous user prompts: Implement Ctrl+r functionality to search through previous user prompts. User prompts should be saved in a raw, unsummarized format.
• Embedding model for conversation history paths: Utilize an embedding model to generate folder paths for conversation history, enabling more intelligent organization.
• Customizable prompts from file: Allow users to define and customize prompts by reading them from a file located at ~/.config/curlpilot/.
• Easy tool integration: Simplify tool integration by allowing any new script placed under ~/.config/curlpilot/tools/ to automatically function as a tool.

instead of

jq -j -n --arg key "$JQ_FILTER_KEY" 'inputs | try (.choices[0][$key].content? // empty)'

i'm doing

jq -j -n 'inputs | .choices[0] | .delta.content // .message.content'

which might be ok

• read

# The pipe '|' sends the string to the { ... } block's standard input
printf '{"json":"payload"}\0{"status":"ok"}' | {
  # First 'read' consumes everything up to the null byte
  IFS= read -r -d '' part1

  # Second 'read' consumes the rest of the stream
  IFS= read -r part2

  echo "Part 1: $part1"
  echo "Part 2: $part2"
}

• expansion

# The string is first stored entirely in a variable
full_string=$'{"json":"payload"}\0{"status":"ok"}'
# 1. Remove the null byte and everything *after* it
part1="${full_string%$'\0'*}"
# 2. Remove everything *up to and including* the null byte
part2="${full_string#*$'\0'}"

echo "Part 1: $part1"
echo "Part 2: $part2"