Research Mode (RLM)

The RLM Research Engine is an AI agent that autonomously explores the knowledge graph, searches academic databases, ingests papers, and synthesizes a cited answer.

Endpoint

POST /api/v1/knowledge/research/query

Returns a Server-Sent Events (SSE) stream of steps.

Usage

curl -N -X POST -H "X-API-Key: $KEY" \
  -H "Content-Type: application/json" \
  "https://api.marc27.com/api/v1/knowledge/research/query" \
  -d '{"question": "What materials resist erosion in rocket nozzles?", "depth": 1}'

Depth Pricing

Depth	Cost	What it does
0	Free	Local graph + vector search only
1	$0.01	+ Web search (Semantic Scholar, arXiv, PubMed)
2	$0.05	+ Deep search, follows new entities
3	$0.10	+ Cross-referencing and validation

Tools Available to the Agent

Tool	What it does
graph_search	Fulltext search on entity names
get_neighbors	1-hop graph traversal
vector_search	Semantic similarity search
property_search	Find materials by property (e.g., yield strength)
web_search	Semantic Scholar + arXiv + PubMed (depth >= 1)
ingest_paper	Extract entities from a paper into the graph (depth >= 1)
compute_run	Submit DFT/ML job (MACE, QE, LAMMPS)
synthesize	Generate final answer with citations

SSE Event Format

data: {"step": "reasoning", "data": {"turn": 0, "text": "I'll start with vector search..."}}
data: {"step": "tool_call", "data": {"tool": "vector_search", "query": "nozzle erosion"}}
data: {"step": "tool_result", "data": {"count": 5, "tool": "vector_search"}}
data: {"step": "answer", "data": {"text": "Based on the analysis..."}}
data: {"step": "complete", "data": {"metrics": {"graph_queries": 5, "cost_usd": 0.01}}}

Via PRISM CLI

prism query --platform --semantic "What materials resist nozzle erosion?"