Tencent/WeKnora · error

invalid hit object format

Error message

invalid hit object format

What it means

processSingleHit asserts each element of the hits list is a map[string]interface{}; if a hit is not a JSON object, the code logs a warning and returns this error, aborting the batch. It indicates a malformed element in the Elasticsearch search response's hits array during index copy.

Source

Thrown at internal/application/repository/retriever/elasticsearch/v7/repository.go:1158

		}
	}

	return indexInfoList, nil
}

// processSingleHit processes a single hit and creates index information
func (e *elasticsearchRepository) processSingleHit(ctx context.Context,
	hit interface{},
	sourceToTargetKBIDMap map[string]string,
	sourceToTargetChunkIDMap map[string]string,
	targetKnowledgeBaseID string,
) (*typesLocal.IndexInfo, []float32, error) {
	log := logger.GetLogger(ctx)

	hitMap, ok := hit.(map[string]interface{})
	if !ok {
		log.Warnf("[ElasticsearchV7] Invalid hit object format")
		return nil, nil, fmt.Errorf("invalid hit object format")
	}

	// Get document source
	sourceObj, ok := hitMap["_source"].(map[string]interface{})
	if !ok {
		log.Warnf("[ElasticsearchV7] Hit missing _source field")
		return nil, nil, fmt.Errorf("hit missing _source field")
	}

	// Get source ChunkID and corresponding target ChunkID
	sourceChunkID, ok := sourceObj["chunk_id"].(string)
	if !ok {
		log.Warnf("[ElasticsearchV7] Source index data missing chunk_id field")
		return nil, nil, fmt.Errorf("source index data missing chunk_id field")
	}

	targetChunkID, ok := sourceToTargetChunkIDMap[sourceChunkID]
	if !ok {

View on GitHub (pinned to 988cbb0330)

Solutions

1. Dump the raw hits array and locate the malformed element to confirm the corruption source
2. Ensure decoding uses standard json.Unmarshal into map[string]interface{} without custom transforms
3. Verify the source index actually belongs to Elasticsearch and its mapping is intact
4. Skip-and-log bad hits instead of failing the whole batch if partial copy is acceptable

Example fix

// before
info, vec, err := processSingleHit(hit)
if err != nil { return err }
// after: skip malformed hits
info, vec, err := processSingleHit(hit)
if err != nil {
    log.Warnf("skipping malformed hit: %v", err)
    continue
}

null

func isHitObject(hit interface{}) (*map[string]interface{}, bool) {
    m, ok := hit.(map[string]interface{})
    if !ok { return nil, false }
    return &m, true
}

info, vec, err := processSingleHit(hit)
if err != nil {
    log.Warnf("skipping malformed hit: %v", err)
    continue // fail-soft per hit instead of aborting the batch
}

When it happens

Trigger: processSourceBatch → processSingleHit receives a hit that is not an object — e.g. the hits array contains scalars/arrays, or the previous layer mis-parsed the response so document objects were mangled.

Common situations: Corrupted or non-standard response body; a custom middleware/decoder altering JSON types (e.g. numbers-only elements); hitting a non-ES-compatible service whose hits array has unexpected element types.

Related errors

• invalid search result format
• failed to do bulk: %w
• failed to delete by query: %w
• invalid retriever type: %v
• failed to marshal query embedding: %w

AI-assisted analysis of Tencent/WeKnora@988cbb0330 (2026-09-02). Data as JSON: /api/errors/2f49a9ab7c7251f1. Report an issue: GitHub.