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
test(tokenizer): hold the env lock in the invalid-repo Hub error test (#4151)
Kenneth Wong1bbc0ea89e80
diff --git a/crates/headroom-core/src/tokenizer/hf_impl.rs b/crates/headroom-core/src/tokenizer/hf_impl.rsindex efb58c5d9..7b5fc8067 100644--- a/crates/headroom-core/src/tokenizer/hf_impl.rs+++ b/crates/headroom-core/src/tokenizer/hf_impl.rs@@ -510,7 +510,17 @@ mod tests {// request. Locks in that we propagate Hub errors as `Hub`, not// `Load`, so callers can distinguish "couldn't fetch" from// "fetched but malformed".+ //+ // Runs online: hold the env lock so a parallel test's+ // `HEADROOM_OFFLINE=1` can't turn this into `Offline`, and clear an+ // inherited value for the same reason.+ let _guard = crate::test_support::env_lock();+ let inherited = std::env::var_os(crate::offline::OFFLINE_ENV);+ std::env::remove_var(crate::offline::OFFLINE_ENV);let r = HfTokenizer::from_pretrained("");+ if let Some(value) = inherited {+ std::env::set_var(crate::offline::OFFLINE_ENV, value);+ }assert!(matches!(r, Err(HfTokenizerError::Hub { .. })),"expected HfTokenizerError::Hub, got {r:?}"