feat: language-aware context compression summaries

Port from anomalyco/opencode#20581: context compaction now generates summaries in the same language the user was using in the conversation. Previously, summaries were always produced in English regardless of the conversation language, which would confuse multilingual users by injecting English context into non-English conversations. Adds 'Write the summary in the same language the user was using in the conversation.' to both the initial and iterative update summarization prompts in ContextCompressor.
2026-04-02 17:06:48 -07:00
1 changed files with 4 additions and 0 deletions
@@ -301,6 +301,8 @@ Update the summary using this exact structure. PRESERVE all existing information

 Target ~{summary_budget} tokens. Be specific — include file paths, command outputs, error messages, and concrete values rather than vague descriptions.

+Write the summary in the same language the user was using in the conversation.
+
 Write only the summary body. Do not include any preamble or prefix."""
        else:
            # First compaction: summarize from scratch
@@ -339,6 +341,8 @@ Use this exact structure:

 Target ~{summary_budget} tokens. Be specific — include file paths, command outputs, error messages, and concrete values rather than vague descriptions. The goal is to prevent the next assistant from repeating work or losing important details.

+Write the summary in the same language the user was using in the conversation.
+
 Write only the summary body. Do not include any preamble or prefix."""

        try: