fix: return truncated content when token limit exceeded in MCP search_code

[waynesun09]

Summary

Fixes token truncation behavior in the MCP search_code tool to return partial content instead of discarding files completely when token limits are exceeded.

Problem

Currently, when a search result file would exceed the maxTokens limit, the entire file is discarded with only a truncation message returned. This provides no useful data to the user, especially problematic when the first file is too large.

Example failure scenario:

• User requests search with maxTokens=6000
• First result file contains 10K tokens
• Current behavior: Returns ONLY truncation message, no file data
• Result: User gets no information at all

Solution

Modified the truncation logic in packages/mcp/src/index.ts to:

1. Calculate remaining token budget before breaking the loop
2. If meaningful space remains (>100 tokens), truncate the file content to fit
3. Append a clear truncation marker: ...[content truncated due to token limit]
4. Add the truncated content to results
5. Continue to add the overall truncation message at the end

Changes

File: packages/mcp/src/index.ts (lines 125-142)

Before:

if ((totalTokens + tokens) > maxTokens) {
    isResponseTruncated = true;
    break;  // Discards the file completely
}

After:

if ((totalTokens + tokens) > maxTokens) {
    const remainingTokens = maxTokens - totalTokens;
    
    if (remainingTokens > 100) {
        const maxLength = Math.floor(remainingTokens * 4);
        const truncatedText = text.substring(0, maxLength) + 
            "\n\n...[content truncated due to token limit]";
        
        content.push({
            type: "text",
            text: truncatedText,
        });
    }
    
    isResponseTruncated = true;
    break;
}

Benefits

✅ Users receive partial data instead of nothing
✅ Better debugging and analysis experience
✅ More useful for AI-powered code analysis workflows
✅ Consistent with expected truncation behavior
✅ Maintains backward compatibility (still includes truncation message)

Example Impact

Scenario: Search returns 50 files, each ~2K tokens, with maxTokens=10000

Before: Returns first 5 complete files (10K tokens), discards file #6 completely
After: Returns first 5 complete files + truncated version of file #6

Testing

• ✅ Verified token calculation logic (chars/4 approximation)
• ✅ Tested with various token limits (100, 1000, 10000, 150000)
• ✅ Confirmed truncation marker appears correctly
• ✅ Validated backward compatibility (truncation message still appended)

Related Issues

This fix specifically addresses issues observed in AI agent workflows where large codebases trigger token limits on early results, providing zero useful data to analysis tasks.

---

Commit: c5b8fda
Branch: fix/mcp-search-truncation

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on this repository.

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

✨ Finishing touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Post copyable unit tests in a comment

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[brendan-kellam]

LGTM, this approach makes sense. Thanks for the contribution 👍 Left one comment, once resolved happy to merge