Global search recommendations

Dev playground for the predictive recommendations path (POST /v3/dev/test/global-search-recommendations). Uses the in-repo fallback prompt directly — no Langfuse fetch — so editing server/ai-agents/student-search/instructions.ts and re-running is the iteration loop. When happy, sync to Langfuse manually.

Query

Query (0–120 chars) · empty → default recommendations from context

Context (top-level page)

Current scope (sub-tab)

CompanyId (positive number)

GIG context (optional)

Type a gigId and "Load from gigId" to prefill the fields below from the DB. Edit them freely before clicking Run — gigId itself is never sent on the run call.

gigId

Title (required for send)

Industry

Locations

Required skills

Required interests

Participants (active workers on the GIG)

Compensation type

Predictive prompt

Seeded from STUDENT_SEARCH_PREDICTIVE_SUGGEST_FALLBACK_INSTRUCTIONS on mount. Edit here to iterate without restarting the server — the textarea contents are sent as instructionsOverride on the next Run. Empty falls back to the in-repo constant.

Instructions

0 chars

Run options

Model

Skip validation (LLM raw output only)

No run yet. Fill the form and click Run.