feat: score ingest facet suggestions with Jev

src/gewu/adapters/openrouter_jev.py

@@ -8,7 +8,13 @@ from typing import Any, cast
 from urllib.error import HTTPError, URLError
 from urllib.request import HTTPRedirectHandler, Request, build_opener
 
-from gewu.domain.contracts import IngestReviewCreate, ProjectRecord, ResearchLens
+from gewu.domain.contracts import (
+    FacetName,
+    IngestReviewCreate,
+    JevFacetScore,
+    ProjectRecord,
+    ResearchLens,
+)
 
 ENDPOINT = "https://openrouter.ai/api/alpha/decisions"
 MODEL = "typesafe/jev-1.13"
@@ -21,6 +27,7 @@ class JevAssessment:
     lens: ResearchLens
     lens_confidence: float
     model: str
+    facet_scores: tuple[JevFacetScore, ...] = ()
 
 
 class JevUnavailable(RuntimeError):
@@ -44,6 +51,30 @@ class OpenRouterJev:
         self._model = model
 
     def assess(self, item: IngestReviewCreate, project: ProjectRecord) -> JevAssessment:
+        facet_candidates: tuple[tuple[FacetName, tuple[str, ...]], ...] = (
+            ("tags", item.candidate_tags),
+            ("audience", item.candidate_audience),
+            ("use_for", item.candidate_use_for),
+            ("domain", item.candidate_domain),
+        )
+        facet_questions: dict[str, object] = {}
+        facet_question_map: dict[str, tuple[FacetName, str]] = {}
+        for facet, candidates in facet_candidates:
+            for index, value in enumerate(candidates):
+                key = f"facet_{facet}_{index}"
+                facet_question_map[key] = (facet, value)
+                facet_questions[key] = {
+                    "type": "noul",
+                    "instructions": (
+                        f"Does the candidate {facet} label accurately describe the passage "
+                        "in the context of this project? Judge the label itself, not just "
+                        "whether its words appear in the excerpt."
+                    ),
+                    "criteria": {
+                        "true": "This label is a useful, well-supported classification.",
+                        "false": "This label is inaccurate, unsupported, or too broad.",
+                    },
+                }
         state = {
             "project_title": project.title,
             "project_north_star": project.north_star,
@@ -52,6 +83,9 @@ class OpenRouterJev:
             "candidate_summary": item.summary,
             "candidate_relevance_rationale": item.relevance_rationale,
             "source_excerpt": item.excerpt,
+            "candidate_facets": {
+                facet: candidates for facet, candidates in facet_candidates if candidates
+            },
         }
         criteria = {
             "technical": "Engineering, implementation, technology, or operational practice.",
@@ -59,30 +93,32 @@ class OpenRouterJev:
             "cultural-historical": "Cultural, artistic, historical, social, or audience context.",
             "financial-market": "Costs, revenues, business models, markets, or financing.",
         }
+        questions: dict[str, object] = {
+            "project_relation": {
+                "type": "noul",
+                "instructions": (
+                    "Does this passage help answer a project question or explain "
+                    "the project's technical, conceptual, cultural-historical, or "
+                    "financial-market context? Treat indirect context as relevant when "
+                    "the connection is plausible."
+                ),
+                "criteria": {
+                    "true": "A plausible, specific connection to the project exists.",
+                    "false": "The passage has no plausible project connection.",
+                },
+            },
+            "best_lens": {
+                "type": "choice",
+                "instructions": "Which project research lens best describes this passage?",
+                "criteria": criteria,
+            },
+            **facet_questions,
+        }
         payload = json.dumps(
             {
                 "model": self._model,
                 "state": state,
-                "questions": {
-                    "project_relation": {
-                        "type": "noul",
-                        "instructions": (
-                            "Does this passage help answer a project question or explain "
-                            "the project's technical, conceptual, cultural-historical, or "
-                            "financial-market context? Treat indirect context as relevant when "
-                            "the connection is plausible."
-                        ),
-                        "criteria": {
-                            "true": "A plausible, specific connection to the project exists.",
-                            "false": "The passage has no plausible project connection.",
-                        },
-                    },
-                    "best_lens": {
-                        "type": "choice",
-                        "instructions": "Which project research lens best describes this passage?",
-                        "criteria": criteria,
-                    },
-                },
+                "questions": questions,
             },
             ensure_ascii=False,
             separators=(",", ":"),
@@ -113,6 +149,14 @@ class OpenRouterJev:
             lens = answers["best_lens"]["choice"]
             confidence = float(answers["best_lens"]["confidence"])
             served_model = str(result["model"])
+            facet_scores = tuple(
+                JevFacetScore(
+                    facet=facet,
+                    value=value,
+                    probability=float(answers[key]["noul"]),
+                )
+                for key, (facet, value) in facet_question_map.items()
+            )
         except (KeyError, TypeError, ValueError, json.JSONDecodeError) as error:
             raise JevUnavailable("Jev response did not match its typed decision schema") from error
         if (
@@ -122,4 +166,6 @@ class OpenRouterJev:
             or len(served_model) > 100
         ):
             raise JevUnavailable("Jev response contained an invalid decision")
-        return JevAssessment(relation, cast(ResearchLens, lens), confidence, served_model)
+        return JevAssessment(
+            relation, cast(ResearchLens, lens), confidence, served_model, facet_scores
+        )