Skip to the page
Get started free

SpecStream for Tongyi DeepResearch

SpecStream is the live web data engine for teams and agents. Change the Serper host in tool_search.py and the Jina Reader address in tool_visit.py, and Alibaba's deep research agent searches and reads pages through SpecStream, in the reply shapes its tools already parse.

Read the guide

What you get

Two addresses
google.serper.dev/search becomes api.specstream.ai/serper/search, and r.jina.ai becomes api.specstream.ai/jina. The model, prompts and summary model stay as they are.
One key for both
Your SpecStream key goes in SPECSTREAM_API_KEY and JINA_API_KEYS. One balance pays for every search and every page a run reads.
Unreadable pages cost nothing
A page SpecStream cannot read is not charged, however many times the visit tool tries it.

Set it up

  1. Get a SpecStream key on the Search API screen of your workspace.
  2. Change the Serper host and path in inference/tool_search.py and the Jina Reader prefix in inference/tool_visit.py to SpecStream's routes.
  3. Set SPECSTREAM_API_KEY and JINA_API_KEYS to your key in .env and run run_react_infer.sh.

The full setup guide has every step, with what to do if something doesn’t arrive.

Also in Agent frameworks & builders

The same live web data, in the other tools of this kind.

Live web data, in Tongyi DeepResearch.

Ask in your own words for companies, places, people, filings or anything else the web lists.

Read the guideTalk to a human