xtekky/gpt4free · error · RuntimeError

Failed to extract nonce/sig from __data.json. Got: {tap_data

Error message

Failed to extract nonce/sig from __data.json. Got: {tap_data}

What it means

io.ReadAll on the opened embedded archive failed. For an embed.FS file this only happens on a genuine I/O error; in practice it surfaces when the embedded bytes are truncated or the read is interrupted, and it wraps the underlying error with %w for diagnosis.

Source

Thrown at g4f/Provider/BraveSearch.py:273

            data_params = {
                "q": prompt,
                "x-sveltekit-invalidated": "11",
            }
            if is_deep:
                data_params["enable_research"] = "true"

            async with session.get(
                DATA_ENDPOINT, params=data_params, headers={"referer": referer}
            ) as response:
                await raise_for_status(response)
                data_json = await response.json()
                tap_data = cls._parse_sveltekit_tap_data(data_json)

                nonce = tap_data.get("nonce")
                sig = tap_data.get("sig")
                if not nonce or not sig:
                    raise RuntimeError(
                        "Failed to extract nonce/sig from __data.json. "
                        f"Got: {tap_data}"
                    )
                conversation.nonce = nonce
                conversation.sig = sig
                debug.log(f"BraveSearch: Got nonce={nonce[:8]}... sig={sig[:8]}...")

            # ===== STEP 2: Create new conversation =====
            new_params = {
                "language": "en",
                "country": "us",
                "ui_lang": "en-us",
                "safesearch": "moderate",
                "force_safesearch": "0",
                "units_of_measurement": "metric",
                "use_location": "1",
                "geoloc": "50.457x30.532",
                "premium_cookie_name": "__Secure-sku#brave-search-premium",

View on GitHub (pinned to 973504e177)

Solutions

Inspect the wrapped %w error — it distinguishes I/O failure from truncation
Re-fetch the archive (delete embed cache, rerun fetch-python.sh) verifying the zip's size/checksum, then rebuild
go clean -cache and rebuild to eliminate a corrupted embed cache
If memory limits are the cause (huge archive + ReadAll), stream-extract with zip.NewReader over the embed.FS file instead of buffering all bytes

Example fix

// before
data, err := io.ReadAll(src)
src.Close()
if err != nil {
    return fmt.Errorf("read embedded archive: %w", err)
}
if err := extractZip(bytes.NewReader(data), int64(len(data)), binDir); err != nil {

// after
zr, err := zip.NewReader(src, info.Size()) // src is embed fs.File; use Stat() for size
if err != nil {
    return fmt.Errorf("read embedded archive: %w", err)
}
// extract from zr directly without buffering the whole archive

Defensive patterns

Strategy: try-catch

Try / catch

data, err := io.ReadAll(src)
src.Close()
if err != nil {
    return fmt.Errorf("read embedded archive: %w", err)
}
// Distinguish corruption from environment: fail fast, never retry a bad embed.
// Repair path is always: re-fetch archive + rebuild binary.

Prevention

Verify zip integrity in fetch-python.sh (unzip -t or a checksum) before embedding
Stream-extract with zip.NewReader instead of ReadAll to avoid memory-driven read failures on large runtimes

When it happens

Trigger: Reading a corrupted or zero-length zip that passed the embed step; extremely large archives exhausting memory during ReadAll (OOM-kill surfaces as a read error on some platforms); a corrupted build cache providing bad embedded data.

Common situations: fetch-python.sh downloading a partial zip (interrupted network) that then got embedded; CI runners with tight memory limits embedding a large runtime zip; disk corruption or an aggressive cache cleaner touching GOCACHE.

Related errors

AI-assisted analysis of xtekky/gpt4free@973504e177 (2026-08-14). Data as JSON: /api/errors/9ca59d75fbf52808. Report an issue: GitHub.