The 12 Signals

Classifies natural language queries into precise intents: factual, entity, temporal, synthesis, actions, history, or aggregation.

Pulls people, organizations, decisions, action items, dates, and topics from any text. Returns structured JSON.

Auto-classifies incoming content into types (meeting, decision, task, insight) with summaries and tags.

Turns vague queries into precise, comprehensive search queries that capture the user's likely intent.

Generates conversational answers with inline citations [1][2][3], sources section, and follow-up questions.

Produces rich 2-4 sentence standalone summaries preserving every decision, name, and action item.

Classifies tone as positive, negative, neutral, tense, or urgent.

Scores content 1-5 with a reason explaining the score.

Detects relationships between memories across 7 types: same_topic, depends_on, contradicts, continuation_of, same_people, causes, temporal.

Detects contradictions and forgotten commitments by comparing current context against stored memories.

Handles follow-up questions with context from the previous conversation turn.

When the answer is not in the data, says so clearly and suggests where to find it.