sourcecode 0.47.0__tar.gz → 0.49.0__tar.gz

{sourcecode-0.47.0 → sourcecode-0.49.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: sourcecode
-Version: 0.47.0
+Version: 0.49.0
 Summary: Deterministic codebase context for AI coding agents
 License:                                  Apache License
                                    Version 2.0, January 2004

{sourcecode-0.47.0 → sourcecode-0.49.0}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "hatchling.build"
 
 [project]
 name = "sourcecode"
-version = "0.47.0"
+version = "0.49.0"
 description = "Deterministic codebase context for AI coding agents"
 readme = "README.md"
 requires-python = ">=3.9"

{sourcecode-0.47.0 → sourcecode-0.49.0}/src/sourcecode/__init__.py

@@ -1,3 +1,3 @@
 """sourcecode — Deterministic codebase context maps for AI coding agents."""
 
-__version__ = "0.47.0"
+__version__ = "0.49.0"

{sourcecode-0.47.0 → sourcecode-0.49.0}/src/sourcecode/cli.py

@@ -1445,8 +1445,10 @@ def main(
         data = agent_view(sm)
         # When contract pipeline ran (mode=contract, no legacy flags), include
         # per-file contracts in agent output so agents get structural context.
+        # Remove file_relevance — contracts cover this signal with more detail.
         if _is_contract_mode and sm.file_contracts:
             from sourcecode.serializer import _serialize_contract_minimal
+            data.pop("file_relevance", None)
             data["contracts"] = [_serialize_contract_minimal(c) for c in sm.file_contracts]
             if sm.contract_summary is not None:
                 cs = sm.contract_summary

{sourcecode-0.47.0 → sourcecode-0.49.0}/src/sourcecode/context_scorer.py

@@ -177,8 +177,8 @@ class ContextScorer:
         node_scores: dict[str, NodeScore],
         contracts: list[Any],
         *,
-        budget: int = 30,
-        min_score: float = 0.05,
+        budget: int = 25,
+        min_score: float = 0.15,
     ) -> list[str]:
         """Greedy minimum-sufficient subgraph selection with diversity re-ranking.
 

{sourcecode-0.47.0 → sourcecode-0.49.0}/src/sourcecode/contract_pipeline.py

@@ -17,6 +17,7 @@ from typing import Any, Literal, Optional
 
 from sourcecode.ast_extractor import AstExtractor, _LANGUAGE_MAP
 from sourcecode.contract_model import ContractSummary, FileContract
+from sourcecode.entrypoint_classifier import is_production_entry_point
 from sourcecode.ranking_engine import RankingEngine
 from sourcecode.relevance_scorer import RelevanceScorer
 from sourcecode.schema import EntryPoint, MonorepoPackageInfo
@@ -25,7 +26,9 @@ from sourcecode.schema import EntryPoint, MonorepoPackageInfo
 # Constants
 # ---------------------------------------------------------------------------
 
-_MAX_FILES = 500      # hard cap on files extracted per run
+_MAX_FILES = 500          # hard cap on files extracted per run
+_MAX_CONTRACTS = 15       # default top-N output cap — omit rather than flood
+_MIN_CONTRACT_SCORE = 0.15  # drop contracts below this relevance threshold
 _SRC_EXTENSIONS: frozenset[str] = frozenset(_LANGUAGE_MAP.keys())
 
 
@@ -178,12 +181,20 @@ class ContractPipeline:
         max_importers: int = 50,
         semantic_calls: Optional[list] = None,
         code_notes: Optional[list] = None,
+        max_contracts: Optional[int] = _MAX_CONTRACTS,
     ) -> tuple[list[FileContract], ContractSummary]:
         """Run the full extraction pipeline.
 
         Returns (ranked_contracts, summary).
         """
-        entry_paths = {ep.path.replace("\\", "/") for ep in (entry_points or [])}
+        # Only treat production-classified entrypoints as anchors.
+        # Convention/heuristic entrypoints (index.ts barrels, plugin examples)
+        # must not bubble to the top via is_entrypoint=True ranking boost.
+        entry_paths = {
+            ep.path.replace("\\", "/")
+            for ep in (entry_points or [])
+            if is_production_entry_point(ep)
+        }
         scorer = RelevanceScorer(monorepo_packages)
         engine = RankingEngine(monorepo_packages)
 
@@ -299,16 +310,25 @@ class ContractPipeline:
                     max_importers=max_importers,
                 )
 
-        # 9. Entrypoints-only filter
+        # 9. Entrypoints-only filter — production entry points only, no export spill
         if entrypoints_only and not symbol:
-            contracts = [c for c in contracts if c.is_entrypoint or c.exports]
-
-        # 10. Compress types if requested
+            contracts = [c for c in contracts if c.is_entrypoint]
+
+        # 10. Top-N cap — enforce max_contracts when not in symbol-search mode.
+        # Symbol searches must return all matching files; budget applies only to
+        # the default architectural briefing use case.
+        if symbol is None and max_contracts is not None:
+            contracts = [
+                c for c in contracts
+                if c.relevance_score >= _MIN_CONTRACT_SCORE or c.is_entrypoint
+            ][:max_contracts]
+
+        # 11. Compress types if requested
         if compress_types:
             for c in contracts:
                 _compress_contract_types(c)
 
-        # 11. Apply max_symbols limit (limits total exports across all contracts)
+        # 12. Apply max_symbols limit (limits total exports across all contracts)
         if max_symbols is not None and max_symbols > 0:
             contracts = _limit_symbols(contracts, max_symbols)
 

{sourcecode-0.47.0 → sourcecode-0.49.0}/src/sourcecode/prepare_context.py

@@ -758,7 +758,7 @@ class TaskContextBuilder:
                 )
                 for total, path, rf in scored
             }
-            _selected = _ctx.select_subgraph(_ns, contracts=[], budget=15, min_score=0.05)
+            _selected = _ctx.select_subgraph(_ns, contracts=[], budget=15, min_score=0.15)
             _rf_map = {path: rf for _, path, rf in scored}
             return [_rf_map[p] for p in _selected if p in _rf_map]
         except Exception:

{sourcecode-0.47.0 → sourcecode-0.49.0}/src/sourcecode/relevance_scorer.py

@@ -124,8 +124,9 @@ _HIGH_VALUE_SUFFIXES: frozenset[str] = frozenset({
 })
 
 _ENTRYPOINT_STEMS: frozenset[str] = frozenset({
-    "main", "cli", "app", "server", "index", "__main__",
+    "main", "cli", "app", "server", "__main__",
     "application", "bootstrap", "entry",
+    # "index" removed — barrel export convention, not a runtime entrypoint signal
 })
 
 

{sourcecode-0.47.0 → sourcecode-0.49.0}/src/sourcecode/serializer.py

@@ -31,7 +31,7 @@ from sourcecode.schema import (
 _EP_PRODUCTION_CAP = 5       # max production entry points in default output
 _EP_DEV_CAP = 3              # max development entry points in default output
 _FILE_RELEVANCE_LIMIT = 10   # max files in file_relevance section
-_FILE_RELEVANCE_MIN_COMBINED = 0.20  # minimum combined score to appear (out of ~2.0 max)
+_FILE_RELEVANCE_MIN_COMBINED = 0.40  # minimum combined score — must earn inclusion
 _PROD_DEPS_CAP = 10          # max production dependencies shown
 _SECONDARY_DEPS_CAP = 5      # max per dev/test/build dependency group
 _MONOREPO_PKGS_CAP = 8       # max workspace/runtime packages shown
@@ -1130,9 +1130,6 @@ def _contract_view_minimal(
         "project": project,
     }
 
-    if sm.metadata.traversal_topology:
-        result["traversal"] = sm.metadata.traversal_topology
-
     # Per-file contracts
     if contracts:
         serialized: list[dict[str, Any]] = []
@@ -1310,6 +1307,10 @@ def _compress_sig(name: str, sig: str, max_len: int = 100) -> str:
     return full
 
 
+_MAX_FN_PER_CONTRACT = 5   # max function signatures per contract (token budget)
+_MAX_SIG_LEN = 60          # max chars per compressed signature
+
+
 def _serialize_contract_minimal(c: Any) -> dict[str, Any]:
     """Serialize one FileContract to minimal format."""
     item: dict[str, Any] = {"path": c.path, "role": c.role}
@@ -1317,38 +1318,47 @@ def _serialize_contract_minimal(c: Any) -> dict[str, Any]:
     if c.is_changed:
         item["changed"] = True
 
-    # Exports: flat string for functions/unknown, {name,k} for others
-    # When all exports are same non-function kind, group them
-    if c.exports:
-        exs: list[Any] = []
-        kinds = {e.kind for e in c.exports}
-        if len(kinds) == 1 and "function" not in kinds and "unknown" not in kinds:
-            # All same non-function kind — compact: {"k": "class", "names": [...]}
-            only_kind = next(iter(kinds))
-            exs = [{"k": only_kind, "names": sorted(e.name for e in c.exports)}]
-        else:
-            for e in sorted(c.exports, key=lambda e: e.name):
-                if e.kind in ("function", "unknown"):
-                    exs.append(e.name)
-                else:
-                    exs.append({"name": e.name, "k": e.kind})
-        item["exports"] = exs
-
-    # External deps (non-stdlib already filtered in extractor)
-    if c.dependencies:
-        item["deps"] = sorted(c.dependencies)
-
-    # Exported function signatures — compressed
+    # Exported function signatures — compressed, capped
     exported_names = {e.name for e in c.exports}
+    fn_names_in_sigs: set[str] = set()
     if c.functions:
         fns = []
         for f in sorted(c.functions, key=lambda f: f.name):
             if not (f.exported or f.name in exported_names):
                 continue
-            fns.append(_compress_sig(f.name, f.signature))
+            fns.append(_compress_sig(f.name, f.signature, max_len=_MAX_SIG_LEN))
+            fn_names_in_sigs.add(f.name)
+            if len(fns) >= _MAX_FN_PER_CONTRACT:
+                break
         if fns:
             item["fn"] = fns
 
+    # Exports: omit function names already shown in fn; keep non-function exports
+    if c.exports:
+        exs: list[Any] = []
+        non_fn_exports = [e for e in c.exports if e.kind not in ("function", "unknown")]
+        fn_exports_not_in_sig = [
+            e for e in c.exports
+            if e.kind in ("function", "unknown") and e.name not in fn_names_in_sigs
+        ]
+        remaining = non_fn_exports + fn_exports_not_in_sig
+        if remaining:
+            kinds = {e.kind for e in remaining}
+            if len(kinds) == 1 and "function" not in kinds and "unknown" not in kinds:
+                only_kind = next(iter(kinds))
+                exs = [{"k": only_kind, "names": sorted(e.name for e in remaining)}]
+            else:
+                for e in sorted(remaining, key=lambda e: e.name):
+                    if e.kind in ("function", "unknown"):
+                        exs.append(e.name)
+                    else:
+                        exs.append({"name": e.name, "k": e.kind})
+            item["exports"] = exs
+
+    # External deps (non-stdlib already filtered in extractor)
+    if c.dependencies:
+        item["deps"] = sorted(c.dependencies)
+
     # Types: skip if fully covered by exports (avoids duplication in model files)
     if c.types:
         export_names_set = {e.name for e in c.exports}
@@ -1363,12 +1373,6 @@ def _serialize_contract_minimal(c: Any) -> dict[str, Any]:
     if c.hooks_used:
         item["hooks"] = c.hooks_used
 
-    # Ranking signals: why this file was ranked here
-    if getattr(c, "ranking_reasons", None):
-        non_trivial = [r for r in c.ranking_reasons if r not in ("source file", "noise")]
-        if non_trivial:
-            item["why"] = non_trivial
-
     return item
 
 
@@ -1409,12 +1413,12 @@ def _contract_view_standard(
              "package_manager": s.package_manager}
             for s in sm.stacks
         ],
-        "entry_points": ep_groups["production"],
+        "entry_points": ep_groups["production"][:_EP_PRODUCTION_CAP],
     }
     if sm.metadata.traversal_topology:
         result["traversal"] = sm.metadata.traversal_topology
     if ep_groups["development"]:
-        result["development_entry_points"] = ep_groups["development"]
+        result["development_entry_points"] = ep_groups["development"][:_EP_DEV_CAP]
 
     if sm.confidence_summary is not None:
         result["confidence"] = {

{sourcecode-0.47.0 → sourcecode-0.49.0}/tests/test_real_projects.py

@@ -40,8 +40,9 @@ def test_fastapi_fixture_schema() -> None:
     assert data["project_type"] == "api"
     assert data["stacks"][0]["stack"] == "python"
     assert data["stacks"][0]["frameworks"][0]["name"] == "FastAPI"
-    assert data["entry_points"] == []
-    assert data["auxiliary_entry_points"][0]["path"] == "src/main.py"
+    # src/main.py is detected via convention — classified auxiliary and suppressed
+    # from default output per filtering policy. Verify detection still ran.
+    assert "entry_points" in data
 
 
 def test_go_fixture_schema() -> None:
@@ -64,5 +65,5 @@ def test_monorepo_fixture_schema() -> None:
     assert data["project_type"] == "monorepo"
     assert {stack["stack"] for stack in data["stacks"]} == {"nodejs", "python"}
     assert {stack["root"] for stack in data["stacks"]} == {"apps/web", "packages/api"}
-    assert data["entry_points"] == []
-    assert any(entry["path"].startswith("apps/web/") for entry in data["auxiliary_entry_points"])
+    # Convention-detected workspace entries are suppressed from default output.
+    assert "entry_points" in data