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 failing test for hidden folders
Garm50203b8b3885
diff --git a/headroom/learn/scanner.py b/headroom/learn/scanner.pyindex cf3566d2d..a43d9535f 100644--- a/headroom/learn/scanner.py+++ b/headroom/learn/scanner.py@@ -471,6 +471,18 @@ def _component_tokenizations(component: str) -> list[list[str]]:tokens = [token for token in component.split(separator) if token]add(tokens)+ # Claude's flattened encoding can turn a leading "." in hidden directory+ # names into an empty token followed by the remaining component tokens.+ if component.startswith(".") and len(component) > 1:+ hidden_component = component[1:]+ add(["", hidden_component])+ for separator in ("-", ".", None):+ if separator is None:+ tokens = [token for token in re.split(r"[-.]", hidden_component) if token]+ else:+ tokens = [token for token in hidden_component.split(separator) if token]+ add(["", *tokens])+return tokenizations