Build graph analytics endpoint for our trust service

[ea1het]

The trust service should be able to evaluate voting eligibility quickly using precomputed attributes. This is where “close trust neighborhood”, “externally anchored” and “confirmed manipulation” elements of the model live.

Therefore, the objectives of this task are this ones:

a) run a periodic job (e.g., every 5–15 minutes) that builds an in-memory graph from Postgres that computes:

• component/community labels
• “externally anchored” status
• clique suspicion scores

b) writes results back to Postgres tables like:

• user_graph_label(user_id, community_id, anchored_bool, suspicion_score, computed_at)
• optionally edge_flags(voter_id, target_id, flags...)