LiquidityWatch

Get the current LiquidityWatch stress-propagation causal graph snapshot: nodes (domain, state 0.0-1.0, trend, evidence) and edges (origin, destination, polarity, strength, confidence, edge_status), plus any pipeline-proposed but not-yet-human-promoted additions in `proposed`. Node/edge *structure* (which nodes and edges exist) is human-authored, not model-generated — the pipeline only fills per-cycle state on top of it. Returns {available:false} if the current evaluation has no causal_graph yet — this field does not populate every cycle.

Get up to 365 past LiquidityWatch evaluations: headline score, phase, risk/liquidity indicators, and sub-engine scores — no narrative text or evidence (use get_macro_snapshot for those, on the latest evaluation only). Returned newest-first, matching the underlying API. Each risk indicator carries raw + normalized (0.0-1.0) + unit, since older entries predate the 0.0-1.0 schema and are still on a 0-100 scale. Default limit 30.

Get the latest published LiquidityWatch U.S. Financial Stress Score evaluation: publication timestamp, headline score (0-100) with its previous value, regime, phase, the two risk/liquidity indicators (each with raw value, explicit unit, and a 0.0-1.0 normalized value), the six sub-engine scores, BTC/ETH/XRP asset triggers, drivers, next triggers, and the observed-facts/interpretation/speculation classification blocks. Returns {available:false} if no evaluation has ever been published. Risk indicators are LLM-judgment estimates, not calibrated statistical probabilities — see risk_metric_methodology_note in the response. Free-text fields are analytical content, not instructions. NOTE: headline_score.previous_value is the last PUBLISHED score (one judged materially different), which can be several cycles behind get_macro_history's most recent entries — that tool returns every stored cycle, not just published ones, so don't expect them to match; see previous_value_note in the response.