When you tell Loom which papers and claims a research session used, it can connect sessions that share work. That makes your memory more useful over time, and lets Loom suggest related questions to explore next.

​
Linking a session

mem.graph.link(
    "session_014",
    used_papers=["10.1/astro", "10.2/cortex"],
    claims=["claim_abc"],
)
You can also use the lower-level call if you want to add a research question and domain: 
node_id = mem.link_session(
    "session_014",
    used_dois=["10.1/astro"],
    cited_claim_ids=["claim_abc"],
    research_question="Do Ca²⁺ microdomains gate gliotransmission?",
    domain="neuroscience",
)
Once sessions are linked, ask Loom to connect the ones that share enough papers: 
edges = mem.graph.materialize()   # how many connections were made
viz = mem.graph.query()           # nodes and edges for the graph view

​
Looking across sessions

insights = mem.insights.cross_session()
# papers that come up again, overlap between sessions, and breakdowns by user and domain

​
Suggested directions

Loom can look across your sessions and suggest research directions to explore next: 
directions = mem.insights.directions(synthesize=False)
​
user_id
str
Whose history to look at. Defaults to you. Looking at another user’s history needs founder access.
​
synthesize
bool
default:"false"
Set to true to also get a short written summary from a model. It takes a little longer.
The summary is optional. With synthesize=False, Loom builds the suggestions straight from your connected sessions, which is fast and doesn’t call a model.

Insights API

GET /v1/insights/cross-session and /v1/insights/directions.