Improved candidate feature naming

This commit is contained in:
2026-04-26 02:54:52 +02:00
parent ba0a3eab17
commit 0cf3f9cb15
3 changed files with 79 additions and 1 deletions

View File

@@ -1,5 +1,6 @@
from __future__ import annotations
import re
from dataclasses import dataclass, field
from repo_registry.core.models import ContentChunk, ObservedFact, Repository, SourceReference
@@ -126,7 +127,7 @@ class CandidateGraphGenerator:
) -> CandidateCapabilityDraft:
features = [
CandidateFeatureDraft(
name=fact.value or fact.name,
name=self._feature_name(fact, chunks),
type=self._feature_type(fact),
location=fact.path,
confidence=0.65 if fact.value else 0.45,
@@ -194,6 +195,40 @@ class CandidateGraphGenerator:
return "API"
return "interface"
def _feature_name(self, fact: ObservedFact, chunks: list[ContentChunk]) -> str:
route_name = self._route_feature_name(fact.value)
if route_name:
return route_name
if self._feature_type(fact) == "CLI":
function_name = self._function_name_near_fact(fact, chunks)
if function_name:
return f"CLI command {function_name}"
return fact.value or fact.name
def _route_feature_name(self, value: str) -> str:
match = re.search(r"@(?:app|router)\.(get|post|put|patch|delete)\((['\"])(.*?)\2", value)
if match is None:
return ""
method = match.group(1).upper()
path = match.group(3)
return f"{method} {path}"
def _function_name_near_fact(
self,
fact: ObservedFact,
chunks: list[ContentChunk],
) -> str:
line = fact.metadata.get("line")
for chunk in chunks:
if chunk.path != fact.path or chunk.kind != "interface":
continue
if isinstance(line, int) and not (chunk.start_line <= line <= chunk.end_line):
continue
match = re.search(r"^\s*def\s+([a-zA-Z_][a-zA-Z0-9_]*)\s*\(", chunk.text, re.MULTILINE)
if match is not None:
return match.group(1)
return ""
def _ability_confidence(
self,
*,

View File

@@ -56,6 +56,7 @@ def test_candidate_generator_builds_review_seed_from_observed_facts():
assert interface_capability.name == "Expose Repository Interface"
assert interface_capability.confidence == 0.75
assert interface_capability.features[0].type == "API"
assert interface_capability.features[0].name == "POST /classify"
assert interface_capability.features[0].location == "app.py"
assert interface_capability.evidence[0].strength == "strong"
@@ -140,3 +141,44 @@ def test_candidate_confidence_scoring_increases_with_supporting_facts():
assert graph[0].confidence == 1.0
assert graph[0].capabilities[0].confidence == 0.85
assert graph[0].capabilities[1].confidence == 0.75
def test_candidate_generator_names_cli_features_from_nearby_function():
repository = Repository(
id=1,
name="CliTool",
url="/tmp/cli-tool",
description=None,
branch="main",
status="analyzed",
)
facts = [
fact(1, "documentation", "README", "README.md"),
ObservedFact(
id=2,
repository_id=1,
analysis_run_id=1,
snapshot_id=1,
kind="interface",
path="cli.py",
name="python CLI command decorator",
value="@click.command()",
metadata={"line": 3},
),
]
chunks = [
chunk(
1,
"interface",
"cli.py",
"@click.command()\ndef import_repositories():\n pass",
start_line=3,
end_line=5,
)
]
graph = CandidateGraphGenerator().generate(repository, facts, chunks)
feature = graph[0].capabilities[0].features[0]
assert feature.type == "CLI"
assert feature.name == "CLI command import_repositories"

View File

@@ -354,6 +354,7 @@ def test_analyze_repository_records_snapshot_and_observed_facts(tmp_path):
assert candidate_graph.abilities
assert "Example" in candidate_graph.abilities[0].description
assert "@app.get" in candidate_graph.abilities[0].capabilities[0].description
assert candidate_graph.abilities[0].capabilities[0].features[0].name == "GET /health"
capability_names = {
capability.name
for ability in candidate_graph.abilities