jackwener/OpenCLI · error · CommandExecutionError

Malformed ChatGPT Deep Research ${label}: expected object so

Error message

Malformed ChatGPT Deep Research ${label}: expected object source row.

What it means

Thrown by the addSource helper that normalizes Deep Research citation/source rows. Each row from metadata.safe_urls or metadata.search_result_groups must be a plain object; if a row is null, a primitive (string/number), or an array, the code cannot extract url/title fields and throws this error instead of silently emitting a corrupt source entry.

Source

Thrown at clis/chatgpt/utils.js:1348

    try {
        return JSON.parse(value);
    } catch {
        return null;
    }
}

function extractDeepResearchSourcesFromReportMessage(reportMessage) {
    const metadata = reportMessage?.metadata && typeof reportMessage.metadata === 'object'
        ? reportMessage.metadata
        : {};
    const references = Array.isArray(metadata.content_references) ? metadata.content_references : [];
    const safeUrls = Array.isArray(metadata.safe_urls) ? metadata.safe_urls : [];
    const groups = Array.isArray(metadata.search_result_groups) ? metadata.search_result_groups : [];
    const byUrl = new Map();

    const addSource = (source = {}, label = 'source') => {
        if (!source || typeof source !== 'object') {
            throw new CommandExecutionError(`Malformed ChatGPT Deep Research ${label}: expected object source row.`);
        }
        const rawUrl = String(source.url || source.href || source.safe_url || '').trim();
        const title = String(source.title || source.name || source.text || '').trim();
        if (!rawUrl) {
            if (title || source.matched_text || source.metadata) {
                throw new CommandExecutionError(`Malformed ChatGPT Deep Research ${label}: missing source URL.`);
            }
            return;
        }
        if (!/^https?:\/\//i.test(rawUrl)) {
            throw new CommandExecutionError(`Malformed ChatGPT Deep Research ${label}: invalid source URL.`);
        }
        if (!byUrl.has(rawUrl)) {
            byUrl.set(rawUrl, { title, url: rawUrl });
        } else if (title && !byUrl.get(rawUrl).title) {
            byUrl.get(rawUrl).title = title;
        }
    };

View on GitHub (pinned to 49907e53dc)

Solutions

Inspect the raw metadata.search_result_groups / metadata.safe_urls array and drop non-object entries before calling the extraction helper.
Coerce string URL rows into {url} objects upstream: rows.map(r => typeof r === 'string' ? {url: r} : r).filter(Boolean).
Pin/vendor the ChatGPT client version matching the current backend payload shape, or update the parser for the new row schema.
Wrap extraction in try-catch and fall back to rendering the un-normalized sources.

Example fix

// before
metadata.search_result_groups.forEach(group => collectSources(group, addSource));
// after
const rows = (metadata.search_result_groups || []).filter(s => s && typeof s === 'object' && !Array.isArray(s));
rows.forEach(group => collectSources(group, addSource));

Defensive patterns

Strategy: validation

Validate before calling

const rows = Array.isArray(metadata?.search_result_groups) ? metadata.search_result_groups : [];
const bad = rows.filter(r => !r || typeof r !== 'object' || Array.isArray(r));
if (bad.length) console.warn('dropping non-object source rows', bad);

Type guard

const isSourceRow = (s) => !!s && typeof s === 'object' && !Array.isArray(s);

Try / catch

try {
  extractDeepResearchSources(metadata);
} catch (err) {
  if (String(err.message).includes('expected object source row')) {
    metadata.search_result_groups = metadata.search_result_groups.filter(isSourceRow);
    return extractDeepResearchSources(metadata);
  }
  throw err;
}

Prevention

Filter source arrays with a non-array object guard before extraction
Keep a fixture test mirroring the live Deep Research metadata shape
Never hand-write source rows in tests without url/href keys
Log raw metadata when extraction throws to speed up schema-change diagnosis

When it happens

Trigger: metadata.safe_urls or metadata.search_result_groups contains a null, undefined, string, number, or array element passed to addSource (typeof source !== 'object' or source is falsy).

Common situations: ChatGPT backend schema changes to Deep Research annotation payloads; upstream responses mixing string URLs directly into safe_urls instead of objects; partial/JSON-decoded rows where null entries survived filtering.

Understand the failure class

Background: Schema validation failed / invalid input schema: payload rejected because its shape doesn't match the expected schema — this error's family across 28 libraries.

Parsing and encoding errors: unexpected token, malformed input — why parsers reject input and how to find the real culprit.

Related errors

AI-assisted analysis of jackwener/OpenCLI@49907e53dc (2026-08-29). Data as JSON: /api/errors/c5edba3424b42e43. Report an issue: GitHub.