feat(opencode): switch tool truncation to token-based limits

[Skyline-23]

Summary

• switch global tool output truncation from byte limits to token budgets
• update Read tool truncation messaging to token-based limits
• align bash tool help text and truncation tests with token limits

Testing

• bun test test/tool/truncation.test.ts
• bun test test/tool/bash.test.ts

References

• Fixes Codex model response latency increases significantly as conversation grows #9045

[github-actions]

Hey! Your PR title Switch tool truncation to token-based limits 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 results, I found some related PRs that deal with output limits and truncation, but none appear to be direct duplicates of PR #9648. Here are the potentially related PRs:

Related PRs (not duplicates):

1. PR feat(opencode): add context overflow prevention with configurable thresholds #8810 - feat(opencode): add context overflow prevention with configurable thresholds

   • Related to managing tool output and context limits, but appears to be about overflow prevention rather than token-based truncation

2. PR fix(grep): stream ripgrep output to prevent memory exhaustion #5432 - fix(grep): stream ripgrep output to prevent memory exhaustion

   • Related to managing tool output size, but focused on grep/ripgrep specifically

3. PR feat: add file persistence for large tool outputs #6234 - feat: add file persistence for large tool outputs

   • Related to handling large tool outputs, but uses file persistence approach rather than token-based limits

These PRs address related concerns about tool output management, but none appear to be duplicate work on the same token-based truncation feature that PR #9648 implements.

[Copilot]

Pull request overview

This pull request switches the tool output truncation system from byte-based limits to token-based limits for more accurate context management. The token estimation uses a simple 4 characters per token approximation.

Changes:

• Replaced MAX_BYTES (50KB) with MAX_TOKENS (10,000) as the default truncation limit
• Updated truncation logic to use Token.estimate() instead of Buffer.byteLength()
• Updated user-facing messages in Read and Bash tools to reference tokens instead of bytes

Reviewed changes

Copilot reviewed 6 out of 6 changed files in this pull request and generated no comments.

Show a summary per file

File	Description
packages/opencode/src/tool/truncation.ts	Core implementation: renamed constants, options, and variables from byte-based to token-based; replaced byte counting with token estimation
packages/opencode/src/tool/read.ts	Updated Read tool to use token-based truncation and updated user-facing messages to reference tokens
packages/opencode/src/tool/bash.ts	Updated template variable replacement from maxBytes to maxTokens
packages/opencode/src/tool/bash.txt	Updated help text to reference token limits instead of byte limits
packages/opencode/test/tool/truncation.test.ts	Updated test descriptions and assertions to use token-based expectations; added Token import
packages/opencode/test/tool/bash.test.ts	Updated truncation test to generate token-based test data

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Skyline-23]

Updated the PR title to conventional format and added Fixes #9045 in the description.