Serper Search for ZeroThink

ZeroThink uses Serper as a web-search route for literature discovery, topic-expert identification, and staged research plans. Search results are normalized into title, URL, snippet, source, and timestamp fields where available, then stored in the provenance evidence ledger.

Enable Serper

Add a key to the ZeroThink root .env file:

SERPER_API_KEY=your_key_here

Studio users can also paste a key into the Vault under Serper Web Key. If the key is missing, ZeroThink shows a clear setup message rather than crashing.

CLI Examples

zerothink search "generative UI topic expert literature review"
zerothink search the web for "generative UI topic experts"
zerothink do a web search for "AI co-scientist literature"
zerothink research --topic "AI co-scientist literature discovery" --web serper
zerothink research "AI co-scientist literature discovery" with web search
zerothink nottingham --topic "survey extension generative UI" --serper

On this server checkout, the Windows wrapper lives at tools\zerothink.cmd and the PHP command lives at tools/zerothink.php.

Evidence Ledger

Serper output is written to storage/provenance/evidence-ledger.jsonl. Each record includes the query, workflow context, plan stage, collection timestamp, and result entries.

Research Plan Use

Inside staged research plans, Serper should be used before synthesis. The recommended order is search, retrieve papers, process full text where possible, map evidence, validate with a human, then draft.