Description
Github Copilot follows a request based subscription instead of token based. Normally, a request is deducted with every user message. With this plugin, I've noticed it gets deducted every time the background process calls pings the agent to get its attention.
Reproduction Steps
Steps to reproduce the behavior:
- Configure plugin in opencode.json
- Run opencode
- Activate model with copilot provider
- Ask the model to run a process with the pty_* extension and say "hello" after it's done running.
- The agent will stop itself after launching background process, activate again when it pings on completion, and then say hello
- You will get 2 requests deducted from copilot subscription.
Expected Behavior
Single request should have been deducted for a single user message. The pty_spawn is a tool call and normally these calls don't cause extra request deduction.
Actual Behavior
A single user message can deducts number_of_pty_calls + 1 requests.
Environment
- OS: Windows 10
- OpenCode Version: 1.4.3
- Plugin Version: 0.3.2
- Bun Version: 1.3.12
OpenCode Configuration
opencode.json (sanitized)
Debug Logs
Click to expand debug logs
Additional Context
Add any other context about the problem here.
Description
Github Copilot follows a request based subscription instead of token based. Normally, a request is deducted with every user message. With this plugin, I've noticed it gets deducted every time the background process calls pings the agent to get its attention.
Reproduction Steps
Steps to reproduce the behavior:
Expected Behavior
Single request should have been deducted for a single user message. The pty_spawn is a tool call and normally these calls don't cause extra request deduction.
Actual Behavior
A single user message can deducts
number_of_pty_calls + 1requests.Environment
OpenCode Configuration
opencode.json (sanitized)
Debug Logs
Click to expand debug logs
Additional Context
Add any other context about the problem here.