Discovery and Scoring

Discovery finds signals. Scoring turns signals into opportunities. signals_ready means discovery finished, not that every candidate is scored. Follow next on get_project_status.

Project Status

get_project_status · intangia:read

Argument	Notes
project_id	UUID

The body matches Project Status: signals_ready, scoring (not_started, running, done), summary, next (wait, run_analyses, list_signals), and analyses.

When the user asked to analyse and scoring is not_started, call run_analyses with top. When scoring is done, unscored signals may remain.

List Signals

list_signals · intangia:read

Discover candidates, not opportunities. Uses the saved ranking mode and signal filter.

Argument	Notes
project_id	UUID
type	Optional type filter
limit	Optional. Default 20, maximum 100.
analysis_mode	Optional. most_promising, novel_signal_scout, or validated_opportunity.
unanalysed_only	When true, only signals that are not scored yet

total is the filtered count. unanalysed_total is how many of those are not scored yet. Do not raise limit past 20 unless the user asked. For a top-N score, call run_analyses with top. Do not copy ids into selections for that.

The row shape matches List Signals. The REST route cannot set unanalysed_only, and its default limit is 100.

List Opportunities

list_opportunities · intangia:read

The ranked scoreboard. Not the Discover list, and not the raw payload from List Opportunities on the API.

Argument	Notes
project_id	UUID
limit	Optional. Default 20, maximum 100.

{
  "total": 1,
  "assessments": [
    {
      "keyword_id": "8f3c1c2e-4b5a-4d6e-9f70-1a2b3c4d5e6f",
      "signal_keyword_id": "91aa0b11-2c3d-4e5f-a6b7-c8d9e0f1a2b3",
      "name": "Example target",
      "type": "target",
      "verdict": "Pursue",
      "composite_score": 72,
      "rationale": "Short rationale, clipped at 240 characters."
    }
  ]
}

type is target, disease, or modality. Pillar text and evidence are not here. For one opportunity, call get_assessment.

You must be a member of the project.

Get an Assessment

get_assessment · intangia:read

One opportunity: verdict, composite score, confidence, rationale, and pillar scores. Look up by signal_keyword_id (the partner id). Use list_opportunities first.

Argument	Notes
project_id	UUID
signal_keyword_id	Partner id
include_evidence	Optional. Papers and risks for this opportunity.

{
  "keyword_id": "8f3c1c2e-4b5a-4d6e-9f70-1a2b3c4d5e6f",
  "signal_keyword_id": "91aa0b11-2c3d-4e5f-a6b7-c8d9e0f1a2b3",
  "name": "Example target",
  "type": "target",
  "verdict": "Pursue",
  "composite_score": 72,
  "composite_grade": "A",
  "confidence": 0.8,
  "verdict_rationale": "Why this verdict was reached.",
  "pillars": { "pillar_biology": 70 }
}

Pillar keys are the pillar_* fields on the assessment, each set to that pillar's score. With include_evidence: true, the result also has recent_evidence (id, year, title, key_finding) and risk_factors (risk, severity, mitigation).

No matching assessment returns { "error": "No assessment for this signal", "status": 404 }.

Run Analyses

run_analyses · analyses:run

You must be an editor or owner. This scores existing signals. It does not discover them, and it does not create a report.

Argument	Notes
project_id	UUID
top	{ "type", "limit" } for the first N of one type. Do not also paste ids.
selections	{ "keyword_id", "signal_keyword_id" } only when the user named specific signals
include_analysed	Optional. Default skips partners that are already scored.
confirmed	Optional. Required when the request leaves the saved default.

top.type is target, disease, modality, or mechanism. Optional fields on top: analysis_mode, confirmed, include_analysed, readiness, therapeutic_area.

The call states its assumption: saved ranking mode, saved signal filter, type, and that it skips signals already scored. Already-scored ids are listed in skipped unless include_analysed is true.

Ask before calling again with confirmed: true when the request leaves that default: a different analysis_mode, readiness or therapeutic_area, include_analysed, mixed types, or more than 20. Do not invent filter JSON. At most 100 signals of one type.

A successful start matches Run Analyses: runs with id, keyword_id, and kind, plus skipped. A confirmation comes back as needs_confirmation: true and does not record a run.

If this fails, repeat the error text. A timeout means no run was recorded: do not retry a smaller list, and do not say scoring is still running. The error is The scoring service did not respond, and no run was recorded. with status 504.