🪢 Open source agent evals & observability: Trace, evaluate, and improve LLM applications with one open platform.
TypeScript analyticsautogenevaluationlangchainlarge-language-modelsllama-indexllmllm-evaluationllm-observabilityllmopsmonitoringobservabilityopen-sourceopenaiplaygroundprompt-engineeringprompt-managementself-hostedycombinator
import { decodeUnicodeEscapesOnly } from "./unicode";
// Replacer that keeps bigints JSON-serializable and decodes \uXXXX escapes so
// that non-ASCII content (e.g. Japanese ingested with Python ensure_ascii=True)
// renders as real characters instead of escape sequences.
const stringifyReplacer = (_key: string, value: unknown) => {
if (typeof value === "bigint") return Number.parseInt(value.toString());
if (typeof value === "string") return decodeUnicodeEscapesOnly(value, true);
return value;
};
export const stringify = (data: any, key?: string, indent?: number): string => {
// For comment fields, use pretty-print formatting for better readability
// Other fields use compact format to reduce file size, unless an explicit
// indent is requested by the caller (e.g. human-readable trace downloads).
const resolvedIndent = indent ?? (key === "comments" ? 2 : undefined);
return JSON.stringify(data, stringifyReplacer, resolvedIndent);
};
/**
* CSV-specific stringify that returns strings as-is instead of JSON-encoding them.
* This avoids double-encoding when string fields (e.g. JSON input/output from ClickHouse)
* are passed through JSON.stringify and then CSV-escaped.
*/
export const stringifyForCsv = (data: any, key?: string): string => {
if (typeof data === "string") return decodeUnicodeEscapesOnly(data, true);
return stringify(data, key);
};