Open Calls

How scoring works

Each call is scored 0–10 for organisational relevance by the open-calls skill (agent r2d29): the scorer queries the cognee knowledge graph for UNINOVA/UNP track-record context, then an LLM rates the call against that track record and writes relevance_score + relevance_rationale back to the vault note. This site mirrors those notes.

• 9–10 strong fit — track record directly matches; act on it
• 7–8.9 good fit — assess consortium/effort
• 5–6.9 partial fit — relevant domain, weaker track-record match
• 0–4.9 weak fit — likely skip
• unscored — awaiting the scoring pass (needs cognee + lm-proxy up)

Agents can upload or re-score calls via the API — see /api/docs (POST /api/calls, PATCH /api/calls/{id}, bearer-token gated).