Signal Canvas
Signal Canvas is the citation-first answer surface for research questions. Every response includes inline citations, an opportunity shortlist, and a portable source_context that chains into downstream actions without reassembling state.
Overview
Signal Canvas synthesizes research from the cs.AI corpus into cited narrative answers. It supports three modes of operation depending on whether you need broad discovery, paper-specific diligence, or workspace-aware continuity.
Citation-first design
Modes
Corpus mode
Broad research questions for ranked opportunities and comparative synthesis across papers.
Paper mode
Anchor the answer to a canonical paper ID for paper-specific diligence and citations.
Workspace mode
Reuse saved thesis context and run history for execution continuity across sessions.
API endpoint
The primary REST endpoint for Signal Canvas queries:
GET /api/research/signal-canvas/{paper_ref}Pass a canonical arXiv reference as paper_ref (e.g. 2401.12345v1). The response includes the synthesized answer, inline citations, and a source_context object for downstream chaining.
Code examples
Query Signal Canvas from any language. All examples use the corpus search endpoint.
curl -X POST https://sciencetostartup.com/api/research/search \
-H "Authorization: Bearer s2s_YOUR_KEY" \
-H "Content-Type: application/json" \
-d '{
"query_text": "What are the highest-viability papers this week?",
"time_scope": "weekly",
"match_count": 10,
"mode": "corpus",
"allow_clarification": true
}'source_context
Every Signal Canvas response returns a source_context object. Pass this object forward when linking into Build Loop, workspace actions, or launch pack generation to maintain citation continuity without reassembling state.
Context chaining
source_context shape drives corpus, paper, and workspace handoffs. Carry it forward across actions to avoid context drift.