xai-org/x-algorithm · warning · anyhow::Error

Strato error code {}: {}

Error message

Strato error code {}: {}

What it means

Raised in VF get_result when the decoded Strato result for an individual tweet is StratoResult::Err — the visibility-filtering Strato backend returned a per-item error (code + message) inside an otherwise successful batch response. The result map stores this as Err for that tweet_id while other tweets may be Ok.

Source

Thrown at visibility-filtering-client/vf_client.rs:183

            .map(|tweet_id| {
                let args = vec![encode(&(*tweet_id, view.clone()))];
                (
                    "visibility/service/homeMixerFilteredReason.Tweet".to_string(),
                    "fetch".to_string(),
                    args,
                )
            })
            .collect::<Vec<(String, String, Vec<Vec<u8>>)>>();
        let result_batch = client.batch_call(calls, context.as_ref()).await;
        let mut result_map: HashMap<u64, Result<Option<FilteredReason>>> = HashMap::new();
        for (tweet_id, bytes_result) in tweet_ids.iter().zip(result_batch) {
            let item_result = match bytes_result {
                Ok(bytes) => {
                    let decoded: StratoResult<StratoValue<FilteredReason>> = decode(&bytes);
                    match decoded {
                        StratoResult::Ok(strato_value) => Ok(strato_value.v),
                        StratoResult::Err(err) => {
                            Err(anyhow!("Strato error code {}: {}", err.code, err.message))
                        }
                    }
                }
                Err(err) => Err(err),
            };
            result_map.insert(*tweet_id, item_result);
        }
        result_map
    }
}

const XAI_VF_DEFAULT_TIMEOUT_MS: u64 = 400;
const XAI_VF_MAX_BATCH_SIZE: usize = 50;
const XAI_VF_APERTURE_SIZE: usize = 16;
const VF_FILTER_TWEETS_DISCOVERY_ENV: &str = "VF_FILTER_TWEETS_DISCOVERY";

pub struct XaiVfClient {
    client: FilterTweetsServiceClient,

View on GitHub (pinned to 24c60942c5)

Solutions

  1. Inspect err.code/err.message: codes like throttle/transient warrant retrying just the failed tweet IDs
  2. Retry only the Err entries from the result_map rather than the whole batch
  3. If a specific code recurs, check VF service logs for that code's meaning; may indicate data-level issues
  4. Degrade gracefully: treat per-item Err as 'unknown' (e.g. allow or conservatively filter the tweet) per product policy

Example fix

// before
let results = vf_client.get_result(...).await;

// after
let results = vf_client.get_result(...).await;
let failed: Vec<u64> = results.iter().filter(|(_, r)| r.is_err()).map(|(id, _)| *id).collect();
let results = if !failed.is_empty() {
    let mut merged = results;
    for (id, r) in vf_client.get_result_for_ids(&failed).await { merged.insert(id, r); }
    merged
} else { results };
Defensive patterns

Strategy: retry

Validate before calling

null

Try / catch

let failed: Vec<u64> = map.iter().filter(|(_, r)| r.is_err()).map(|(k, _)| *k).collect(); retry only failed ids once; then degrade per policy.

Prevention

When it happens

Trigger: Calling filter-style VF batch lookups where one or more tweet IDs produce backend item errors: internal VF Strato errors, invalid/unsupported reason codes, or per-item processing failures.

Common situations: VF backend partial failures during batch filtering; specific corrupt or unsupported tweet records; VF Strato rolling out a new error code that the client surfaces verbatim.

Related errors

AI-assisted analysis of xai-org/x-algorithm@24c60942c5 (2026-08-28). Data as JSON: /api/errors/68a3aa4188338952. Report an issue: GitHub.