fix(opencode): improve local OpenAI-compatible backend robustness

[krystophny]

Issue for this PR

Closes #20669

Type of change

• Bug fix
• New feature
• Refactor / code improvement
• Documentation

What does this PR do?

This makes the default agent more tolerant of two local OpenAI-compatible backend quirks that showed up in real runs:

• the bash tool can receive a valid command without a description
• some backends return finish_reason: "tool_calls" with tool_calls: []

The change normalizes missing bash descriptions before validation/execution and treats empty tool_calls arrays as a normal stop instead of a real tool-call turn.

How did you verify your code works?

• bun run --cwd packages/opencode typecheck
• bun test test/tool/bash.test.ts test/provider/copilot/copilot-chat-model.test.ts
• manual local smoke against a llama.cpp OpenAI-compatible backend with the default build agent

Screenshots / recordings

Not applicable. This is a backend/runtime bug fix.

Checklist

• I have tested my changes locally
• I have not included unrelated changes in this PR

[github-actions]

Hey! Your PR title Improve local OpenAI-compatible backend robustness doesn't follow conventional commit format.

Please update it to start with one of:

• feat: or feat(scope): new feature
• fix: or fix(scope): bug fix
• docs: or docs(scope): documentation changes
• chore: or chore(scope): maintenance tasks
• refactor: or refactor(scope): code refactoring
• test: or test(scope): adding or updating tests

Where scope is the package name (e.g., app, desktop, opencode).

See CONTRIBUTING.md for details.

[github-actions]

The following comment was made by an LLM, it may be inaccurate:

Based on my search, I found several potentially related PRs:

Most Relevant:

1. PR refactor(opencode): harden BashTool input validation and environment #19290 - refactor(opencode): harden BashTool input validation and environment

   • refactor(opencode): harden BashTool input validation and environment #19290
   • Related: Also addresses BashTool input validation robustness

2. PR fix(bash): make description parameter optional in bash tool #13832 - fix(bash): make description parameter optional in bash tool

   • fix(bash): make description parameter optional in bash tool #13832
   • Related: Directly addresses making bash description optional, which is one of the changes in PR fix(opencode): improve local OpenAI-compatible backend robustness #20668

3. PR fix(bash): make description optional for tool calls #13288 - fix(bash): make description optional for tool calls

   • fix(bash): make description optional for tool calls #13288
   • Related: Similar fix for bash description parameter

4. PR fix(provider): handle empty content for bedrock/openai-compatible APIs #17396 - fix(provider): handle empty content for bedrock/openai-compatible APIs

   • fix(provider): handle empty content for bedrock/openai-compatible APIs #17396
   • Related: Addresses handling of empty/missing content from OpenAI-compatible providers

5. PR fix(opencode): filter empty text content blocks for all providers #17742 - fix(opencode): filter empty text content blocks for all providers

   • fix(opencode): filter empty text content blocks for all providers #17742
   • Related: Addresses filtering empty responses from providers

6. PR feat: add streaming option to provider config to support non-streaming backends #15931 - feat: add streaming option to provider config to support non-streaming backends

   • feat: add streaming option to provider config to support non-streaming backends #15931
   • Related: Adds support for non-standard backend behaviors

These PRs may have already addressed parts of what PR #20668 is fixing. Consider reviewing them to ensure no duplicate effort.

[github-actions]

Thanks for updating your PR! It now meets our contributing guidelines. 👍