Compress tool outputs, logs, files, and RAG chunks before they reach the LLM. 20% fewer tokens for coding agents, 60-95% fewer tokens for JSON, same answers. Library, proxy, MCP server.
Python agentaianthropicclaude-codecompressioncontext-engineeringcontext-windowcursorfastapilangchainllmmcpopenaiprompt-engineeringproxypythonragtoken-optimizationtokenstypescript
fix(bedrock): use floor_char_boundary to avoid UTF-8 slice panic in header preview
SwiftWing2182468e4e99e6
diff --git a/crates/headroom-proxy/src/bedrock/eventstream_to_sse.rs b/crates/headroom-proxy/src/bedrock/eventstream_to_sse.rsindex 5b0c56804..ec7a95d55 100644--- a/crates/headroom-proxy/src/bedrock/eventstream_to_sse.rs+++ b/crates/headroom-proxy/src/bedrock/eventstream_to_sse.rs@@ -238,7 +238,8 @@ pub fn header_value_preview(v: &HeaderValue) -> String {if s.len() <= 64 {s.clone()} else {- format!("{}…", &s[..64])+ let end = s.floor_char_boundary(64);+ format!("{}…", &s[..end])}}HeaderValue::Bytes(b) => format!("[{} bytes]", b.len()),@@ -394,4 +395,23 @@ mod tests {.starts_with("[3 bytes"));}++ #[test]+ fn header_value_preview_truncates_at_char_boundary() {+ // 63 ASCII bytes + a 2-byte UTF-8 char (é = U+00E9) puts a char+ // boundary at byte 63 but NOT at byte 64 — the old `&s[..64]`+ // would panic here. floor_char_boundary(64) must return 63.+ let s = "x".repeat(63) + "éfoo";+ assert!(s.len() > 64);+ let preview = header_value_preview(&HeaderValue::String(s));+ assert!(preview.ends_with('…'), "expected ellipsis suffix: {preview:?}");+ assert!(!preview.contains('é'), "must not include the split codepoint");+ }++ #[test]+ fn header_value_preview_exact_boundary_not_truncated() {+ // A string whose UTF-8 length is exactly 64 must not be truncated.+ let s = "x".repeat(64);+ assert_eq!(header_value_preview(&HeaderValue::String(s.clone())), s);+ }}