RAGTIME Search API

Based on the positive feedback we have received at NeuCLIR Report Generation Pilot last year, we provide a search endpoint using Qwen3 8B Embedding using a FAISS index for participants if you do not want to stand up a search engine by yourself. The endpoint is powered by RoutIR.

Please register at TREC to get the URL and the BEARER token to the endpoint.

Search

You can use the following HTTP POST request to retrieve documents based on your query.

curl -X POST https://get.url.from.trec/pipeline \
    -H "Authorization: ${RAGTIME_BEARER_TOKEN}" \
    -H "Content-Type: application/json" \
    -d '{
        "pipeline": "ragtime2%20",
        "query": "your query"
    }'

Here ragtime2 is an alias for the pipeline {zho%25, spa%25, rus%25, eng%25}ScoreFusion. If you want to retrieve in a single language, you can use one of the languages alone in the pipeline field. Please see the documentation of RoutIR for more information.

If you are working with the NeuCLIR data for development, you can replace the pipeline with ragtime1 to retrieve documents in RAGTIME 2025 languages (i.e., Chinese, Arabic, Russian and English), which is an alias for the pipeline {zho%25, arb%25, rus%25, eng%25}ScoreFusion.

Get Document Content

You can retrieve the document content also from the search endpoint by using the API /content. For example,

curl -X POST https://get.url.from.trec/content \
    -H "Authorization: ${RAGTIME_BEARER_TOKEN}" \
    -H "Content-Type: application/json" \
    -d '{
        "collection": "ragtime2",
        "id": "01f01b3d-89c7-4f97-9d14-bd2dcdff190b"
    }'

Contact

Please send an email to us if you have any question at ragtime-organizers@googlegroups.com or at the TREC Slack.