sourcecode 1.18.0__tar.gz → 1.20.0__tar.gz

sourcecode-1.20.0/.continue-here.md

@@ -0,0 +1,178 @@
+# Continue Here — atlas-cli v1.17.0 defect patch (saint-server round 2)
+
+**Paused:** 2026-05-13 (session 2)
+**Repo:** `/Users/user/Downloads/atlas-cli`
+**Branch:** master
+
+---
+
+## Estado actual
+
+Sesión 2: 6 defectos adicionales contra saint-server. **Sin commit** (instrucción explícita). Pipx venv reinstalado en modo editable para pruebas en vivo. 743 tests pasan.
+
+---
+
+## Defectos corregidos sesión 2 (ESTA SESIÓN)
+
+| ID | Sev | Fix | Fichero |
+|----|-----|-----|---------|
+| FIX-1 | CRITICAL-3 | `--symptom` secondary pass: inyecta ficheros del árbol cuyo path contiene keywords del síntoma (antes solo re-ordenaba candidatos ya en pool). role="symptom_match", score=0.5+boost | `prepare_context.py:655-676` |
+| FIX-2 | HIGH-2 | `_security_surface_from_eps` resuelve constantes Java `ClassName.FIELD_NAME` → busca `ClassName.java` en file_paths y extrae valor literal. `resource_names_unresolved[]` para las irresolubles. Call sites actualizados con root+file_paths | `serializer.py:428-522` |
+| FIX-3 | HIGH-3 | Nota truncación: `"use --full for complete list"` → `"use --full to see all {total}"` | `serializer.py:434` |
+| FIX-4 | REGRESSION | `_file_relevance` items ahora tienen campo `"score"` (= relevance_val). `GenericRestController`/`GenericCRUDRestController` override: category="runtime_core", score=0.95, relevance=0.95, reason/evidence/ranking_reasons específicos | `serializer.py:714-737` |
+| FIX-5 | MEDIUM-3 | `compact_view` confidence_summary ahora incluye `"sections"` breakdown (igual que agent_view) | `serializer.py:1046-1052` |
+| FIX-6 | LOW-1 | `_spring_profiles_context` portal profiles (con `-` en nombre): busca en path completo (no solo stem). Si no hay ficheros, añade perfil con `[]` para que esté presente en `per_profile_variants` | `serializer.py:386-406` |
+
+---
+
+## Defectos corregidos sesión 1 (pendientes de commit)
+
+### CRITICAL
+| ID | Fix | Fichero |
+|----|-----|---------|
+| C1 | `_extract_first_useful_paragraph()` rechaza snippets startup via `_STARTUP_RE` | `summarizer.py` |
+| C2 | `_bootstrap_structured()` emite `sample` (5 primeros) en vez de `modules` (todos) | `serializer.py` |
+| C3 | `safe_read_text()` fallback UTF-8→ISO-8859-1→UTF-8/replace | `tree_utils.py` + consumidores |
+| C4 | `_scan_java_patterns()` custom_annotations[] + base_classes[] | `architecture_analyzer.py` + `schema.py` |
+
+### HIGH
+| ID | Fix | Fichero |
+|----|-----|---------|
+| H1 | `_spring_profiles_context()` estructura detallada | `serializer.py` |
+| H2 | `_is_dto_mapper()` separa @Mapper real de DtoMapper | `serializer.py` |
+| H3 | `_transactional_summary()` truncate a 10 + nota | `serializer.py` |
+| H4 | `_TASK_CONTENT_MAP` per-task filter en prepare-context | `cli.py` + `prepare_context.py` |
+| H5 | Fallback `git log -n 100` + campo `hotspot_method` | `git_analyzer.py` + `schema.py` |
+
+### MEDIUM / LOW
+M1–M5, L1 — ver sesión 1.
+
+---
+
+## Ficheros modificados (ambas sesiones)
+
+```
+src/sourcecode/architecture_analyzer.py   sesión 1
+src/sourcecode/cli.py                     sesión 1
+src/sourcecode/code_notes_analyzer.py     sesión 1
+src/sourcecode/confidence_analyzer.py     sesión 1
+src/sourcecode/doc_analyzer.py            sesión 1
+src/sourcecode/git_analyzer.py            sesión 1
+src/sourcecode/schema.py                  sesión 1
+src/sourcecode/serializer.py              sesión 1 + sesión 2 (FIX-2,3,4,5,6)
+src/sourcecode/summarizer.py              sesión 1
+src/sourcecode/tree_utils.py              sesión 1
+src/sourcecode/prepare_context.py         sesión 1 + sesión 2 (FIX-1)
+```
+
+---
+
+## Tests
+
+- **743 passed, 3 skipped** (sin regresiones)
+- Fallos pre-existentes excluidos:
+  - `test_block2_coverage.py::test_java_marked_unsupported`
+  - `test_dependency_analyzer_node_python.py::test_python_requirements_without_lockfile`
+  - `test_pipeline_integrity.py::TestBenchmarkContamination`
+
+---
+
+## Smoke tests para saint-server (pendientes — no tenemos acceso a saint-server aquí)
+
+```bash
+# FIX-1: síntoma inyecta ficheros
+sourcecode prepare-context fix-bug saint-server --symptom "ausencias contador" \
+  --output /tmp/fix_symptom.json
+python3 -c "
+import json; d=json.load(open('/tmp/fix_symptom.json'))
+paths=[f['path'] for f in d['relevant_files']]
+assert any('ausencias' in p for p in paths), 'ausencias not found'
+assert any('contador' in p for p in paths), 'contador not found'
+print('FIX-1 PASS')
+"
+
+# FIX-2: resource_names resueltos
+sourcecode saint-server --compact --output /tmp/sec_test.json
+python3 -c "
+import json; d=json.load(open('/tmp/sec_test.json'))
+unresolved=[r for r in d['security_surface']['resource_names']
+            if 'SeguridadRecursosConst.' in str(r)]
+print(f'Unresolved: {len(unresolved)}')
+assert len(unresolved) < 73, 'No improvement'
+"
+
+# FIX-3: note text correcto
+sourcecode saint-server --compact --output /tmp/txn.json
+python3 -c "
+import json; d=json.load(open('/tmp/txn.json'))
+note=d.get('transactional_boundaries',{}).get('note','')
+assert '--full' in note and 'for complete list' not in note, f'bad note: {note}'
+print('FIX-3 PASS:', note)
+"
+
+# FIX-4: score presente y no nulo
+sourcecode saint-server --agent --output /tmp/agent_test.json
+python3 -c "
+import json; d=json.load(open('/tmp/agent_test.json'))
+gcr=next((f for f in d['file_relevance'] if 'GenericRestController' in f['path']),None)
+assert gcr is not None,'not found'
+assert gcr['score'] is not None,'null score'
+assert isinstance(gcr['score'],(int,float)),'not numeric'
+print('FIX-4 PASS: score=', gcr['score'])
+"
+
+# FIX-5: sections en compact
+sourcecode saint-server --compact --output /tmp/conf_test.json
+python3 -c "
+import json; d=json.load(open('/tmp/conf_test.json'))
+cs=d['confidence_summary']
+assert 'sections' in cs,'sections missing'
+assert 'file_relevance' in cs['sections'],'file_relevance key missing'
+print('FIX-5 PASS:', cs['sections'])
+"
+
+# FIX-6: portal profiles presentes
+sourcecode saint-server --compact --output /tmp/prof_test.json
+python3 -c "
+import json; d=json.load(open('/tmp/prof_test.json'))
+variants=d['spring_profiles']['per_profile_variants']
+portal=[p for p in d['spring_profiles']['detected'] if 'portal' in p]
+missing=[p for p in portal if p not in variants]
+assert len(missing)==0, f'Still missing: {missing}'
+print('FIX-6 PASS')
+"
+```
+
+---
+
+## Commit cuando listo
+
+```bash
+git add src/sourcecode/serializer.py src/sourcecode/prepare_context.py \
+        src/sourcecode/architecture_analyzer.py src/sourcecode/cli.py \
+        src/sourcecode/code_notes_analyzer.py src/sourcecode/confidence_analyzer.py \
+        src/sourcecode/doc_analyzer.py src/sourcecode/git_analyzer.py \
+        src/sourcecode/schema.py src/sourcecode/summarizer.py \
+        src/sourcecode/tree_utils.py \
+        tests/test_encoding_regression.py tests/test_task_differentiation.py \
+        tests/fixtures/
+git commit -m "fix: 21 saint-server defects across 2 sessions (C1-C4, H1-H5, M1-M5, L1, FIX-1 to FIX-6)"
+```
+
+---
+
+## Para retomar
+
+```bash
+cd /Users/user/Downloads/atlas-cli
+git diff --stat  # verificar ficheros modificados
+python3 -m pytest tests/ \
+  --ignore=tests/test_block2_coverage.py \
+  --ignore=tests/test_dependency_analyzer_node_python.py \
+  --ignore=tests/test_pipeline_integrity.py -q
+# Expected: 743 passed, 3 skipped
+# Then run smoke tests above against saint-server
+```
+
+---
+*Pausado 2026-05-13 sesión 2 — gsd:pause-work*

{sourcecode-1.18.0 → sourcecode-1.20.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: sourcecode
-Version: 1.18.0
+Version: 1.20.0
 Summary: Deterministic codebase context for AI coding agents
 License:                                  Apache License
                                    Version 2.0, January 2004
@@ -221,7 +221,7 @@ Description-Content-Type: text/markdown
 
 **Compressed AI-ready context for Java/Spring enterprise codebases.**
 
-![Version](https://img.shields.io/badge/version-1.18.0-blue)
+![Version](https://img.shields.io/badge/version-1.20.0-blue)
 ![Python](https://img.shields.io/badge/python-3.10%2B-green)
 
 ---
@@ -255,7 +255,7 @@ pipx install sourcecode
 
 ```bash
 sourcecode version
-# sourcecode 1.18.0
+# sourcecode 1.20.0
 ```
 
 ---

{sourcecode-1.18.0 → sourcecode-1.20.0}/README.md

@@ -2,7 +2,7 @@
 
 **Compressed AI-ready context for Java/Spring enterprise codebases.**
 
-![Version](https://img.shields.io/badge/version-1.18.0-blue)
+![Version](https://img.shields.io/badge/version-1.20.0-blue)
 ![Python](https://img.shields.io/badge/python-3.10%2B-green)
 
 ---
@@ -36,7 +36,7 @@ pipx install sourcecode
 
 ```bash
 sourcecode version
-# sourcecode 1.18.0
+# sourcecode 1.20.0
 ```
 
 ---

{sourcecode-1.18.0 → sourcecode-1.20.0}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "hatchling.build"
 
 [project]
 name = "sourcecode"
-version = "1.18.0"
+version = "1.20.0"
 description = "Deterministic codebase context for AI coding agents"
 readme = "README.md"
 requires-python = ">=3.9"

{sourcecode-1.18.0 → sourcecode-1.20.0}/src/sourcecode/__init__.py

@@ -1,3 +1,3 @@
 """sourcecode — Deterministic codebase context maps for AI coding agents."""
 
-__version__ = "1.18.0"
+__version__ = "1.20.0"

{sourcecode-1.18.0 → sourcecode-1.20.0}/src/sourcecode/cli.py

@@ -1777,6 +1777,8 @@ def prepare_context_cmd(
         out["symptom"] = output.symptom
     if output.related_notes:
         out["related_notes"] = output.related_notes
+    if output.symptom_note:
+        out["symptom_note"] = output.symptom_note
     if llm_prompt:
         out["llm_prompt"] = builder.render_prompt(output)
 

{sourcecode-1.18.0 → sourcecode-1.20.0}/src/sourcecode/prepare_context.py

@@ -323,6 +323,7 @@ class TaskOutput:
     affected_entry_points: list[str] = field(default_factory=list) # delta task only
     symptom: Optional[str] = None                                  # fix-bug only
     related_notes: list[dict] = field(default_factory=list)        # fix-bug + symptom only
+    symptom_note: Optional[str] = None                             # fix-bug: cross-layer synonym note
 
 
 # ─────────────────────────────────────────────────────────────────────────────
@@ -382,6 +383,22 @@ _ALL_EXTENSIONS: frozenset[str] = _SOURCE_EXTENSIONS | frozenset({
     ".md", ".toml", ".yaml", ".yml", ".json", ".xml",
 })
 
+# Maps frontend symptom keywords → backend terms likely to contain the root cause.
+# Used to boost service/interceptor files when the symptom is UI-only.
+_FRONTEND_SYMPTOM_MAP: dict[str, list[str]] = {
+    "spinner":  ["loading", "setloading", "finalize", "httpinterceptor", "interceptor", "service"],
+    "loading":  ["loading", "setloading", "finalize", "httpinterceptor", "interceptor", "service"],
+    "login":    ["authcontroller", "securityconfig", "filterconfig", "jwtfilter", "auth", "authentication"],
+    "logout":   ["authcontroller", "securityconfig", "jwtfilter", "auth", "session"],
+    "dropdown": ["getmapping", "findall", "obtenertodos", "listall", "findby"],
+    "modal":    ["controller", "getmapping", "findby", "search"],
+    "popup":    ["controller", "getmapping", "findby", "search"],
+    "table":    ["paginated", "findby", "search", "getmapping", "listall"],
+    "grid":     ["paginated", "findby", "search", "getmapping"],
+    "button":   ["postmapping", "putmapping", "deletemapping", "controller", "service"],
+    "form":     ["postmapping", "putmapping", "controller", "service", "dto"],
+}
+
 
 class TaskContextBuilder:
     def __init__(self, root: Path) -> None:
@@ -635,10 +652,13 @@ class TaskContextBuilder:
         # ── 6b. Symptom keyword boost + related notes (fix-bug + --symptom) ──
         symptom_keywords: list[str] = []
         related_notes: list[dict] = []
+        symptom_note: Optional[str] = None
         if task_name == "fix-bug" and symptom:
             import re as _re
+            _camel_expanded = _re.sub(r'([a-z])([A-Z])', r'\1 \2', symptom)
+            _camel_expanded = _re.sub(r'([A-Z]+)([A-Z][a-z])', r'\1 \2', _camel_expanded)
             symptom_keywords = [
-                w.lower() for w in _re.split(r"[\s\W]+", symptom)
+                w.lower() for w in _re.split(r"[\s\W]+", _camel_expanded)
                 if len(w) > 2
             ]
             if symptom_keywords:
@@ -682,6 +702,65 @@ class TaskContextBuilder:
                     return rf.score + 0.2 * sum(1.0 for kw in symptom_keywords if kw in path_lower)
                 relevant_files = sorted(relevant_files, key=lambda rf: -_symptom_score(rf))
 
+                # Content scan boost: read file body for symptom keywords
+                _src_exts = frozenset({".java", ".py", ".ts", ".js", ".kt", ".go"})
+                _content_boosted: list[RelevantFile] = []
+                for _rf in relevant_files:
+                    _extra = 0.0
+                    if Path(_rf.path).suffix.lower() in _src_exts:
+                        try:
+                            _lines = (self.root / _rf.path).read_text(
+                                encoding="utf-8", errors="replace"
+                            ).splitlines()[:300]
+                            _body = "\n".join(_lines).lower()
+                            _hits = sum(_body.count(kw) for kw in symptom_keywords)
+                            _extra = min(0.30, _hits * 0.02)
+                        except OSError:
+                            pass
+                    _content_boosted.append(RelevantFile(
+                        path=_rf.path,
+                        role=_rf.role,
+                        score=round(min(_rf.score + _extra, 1.0), 2),
+                        reason=_rf.reason + (f", content-match symptom (+{_extra:.2f})" if _extra > 0 else ""),
+                        why=_rf.why,
+                    ))
+                relevant_files = sorted(_content_boosted, key=lambda rf: -rf.score)
+
+                # Cross-layer synonym boost: frontend keywords → backend equivalents
+                _synonym_note: Optional[str] = None
+                _frontend_kws = [kw for kw in symptom_keywords if kw in _FRONTEND_SYMPTOM_MAP]
+                if _frontend_kws:
+                    _backend_terms: list[str] = []
+                    for _fkw in _frontend_kws:
+                        _backend_terms.extend(_FRONTEND_SYMPTOM_MAP[_fkw])
+                    _backend_terms_set = list(dict.fromkeys(_backend_terms))  # dedup, preserve order
+                    _synonym_boosted: list[RelevantFile] = []
+                    for _rf in relevant_files:
+                        _extra_syn = 0.0
+                        if Path(_rf.path).suffix.lower() in _src_exts:
+                            try:
+                                _lines_syn = (self.root / _rf.path).read_text(
+                                    encoding="utf-8", errors="replace"
+                                ).splitlines()[:300]
+                                _body_syn = "\n".join(_lines_syn).lower()
+                                _hits_syn = sum(_body_syn.count(t) for t in _backend_terms_set)
+                                _extra_syn = min(0.20, _hits_syn * 0.02)
+                            except OSError:
+                                pass
+                        _synonym_boosted.append(RelevantFile(
+                            path=_rf.path,
+                            role=_rf.role,
+                            score=round(min(_rf.score + _extra_syn, 1.0), 2),
+                            reason=_rf.reason + (f", synonym-match backend (+{_extra_syn:.2f})" if _extra_syn > 0 else ""),
+                            why=_rf.why,
+                        ))
+                    relevant_files = sorted(_synonym_boosted, key=lambda rf: -rf.score)
+                    _synonym_note = (
+                        f"Frontend concept detected ({', '.join(_frontend_kws)}). "
+                        "Boosted backend service-layer and interceptor files as likely root cause."
+                    )
+                    symptom_note = _synonym_note
+
         # ── 7. Test gaps (generate-tests only) ────────────────────────────
         test_gaps: list[str] = []
         if task_name == "generate-tests":
@@ -762,6 +841,7 @@ class TaskContextBuilder:
             affected_entry_points=affected_entry_points,
             symptom=symptom if task_name == "fix-bug" and symptom else None,
             related_notes=related_notes,
+            symptom_note=symptom_note,
         )
 
     def render_prompt(self, output: TaskOutput) -> str:
@@ -863,7 +943,7 @@ class TaskContextBuilder:
         _dominant_stack = ""
         _recently_changed_stacks: set[str] = set()
         if task_name == "fix-bug":
-            _bug_kinds = {"FIXME", "BUG"}
+            _bug_kinds = {"FIXME", "BUG", "HACK", "XXX"}
             for _n in (code_notes or []):
                 if getattr(_n, "kind", "").upper() in _bug_kinds:
                     _annotated_files.add(getattr(_n, "path", ""))
@@ -1020,7 +1100,16 @@ class TaskContextBuilder:
                 )
                 for total, path, rf in scored
             }
-            _selected = _ctx.select_subgraph(_ns, contracts=[], budget=15, min_score=0.15)
+            _repo_size = len(all_paths)
+            _task_budget = {
+                "fix-bug": max(20, min(40, _repo_size // 80)),
+                "onboard": max(15, min(25, _repo_size // 150)),
+                "explain": max(10, min(20, _repo_size // 200)),
+                "generate-tests": max(20, min(35, _repo_size // 100)),
+                "refactor": max(15, min(30, _repo_size // 120)),
+            }
+            _budget = _task_budget.get(task_name, 15)
+            _selected = _ctx.select_subgraph(_ns, contracts=[], budget=_budget, 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:
@@ -1036,6 +1125,9 @@ class TaskContextBuilder:
             or "/tests/" in path
             or "/test/" in path
             or "/spec/" in path
+            or (name.endswith("test.java") and name != "test.java")
+            or name.endswith("tests.java")
+            or (name.startswith("test") and name.endswith(".java") and len(name) > 9)
         )
 
     def _is_source(self, path: str) -> bool:

{sourcecode-1.18.0 → sourcecode-1.20.0}/src/sourcecode/serializer.py

@@ -356,6 +356,7 @@ def _mybatis_pairing(sm: "SourceMap", *, full: bool = False) -> "Optional[dict[s
         result["dto_mappers_total"] = _total_dto
         if _total_dto > 10 and not full:
             result["dto_mappers_truncated"] = True
+            result["dto_mappers_warning"] = f"Showing 10/{_total_dto} mappers. Use --full to see all."
     return result
 
 
@@ -894,6 +895,41 @@ def _serialize_file_metric(m: Any) -> dict[str, Any]:
     return d
 
 
+def _confidence_reasons(sm: SourceMap) -> dict[str, list[str]]:
+    """Actionable reasons explaining low-confidence sections.
+
+    If a section is 'low' with no derivable reasons, the caller should
+    upgrade it to 'high' (the low rating was overly conservative).
+    """
+    reasons: dict[str, list[str]] = {}
+
+    arch = sm.architecture
+    if arch and arch.requested and arch.confidence == "low":
+        arch_reasons: list[str] = []
+        for lim in (arch.limitations or []):
+            if lim:
+                arch_reasons.append(lim)
+        if sm.analysis_gaps:
+            for gap in sm.analysis_gaps:
+                if gap.area in ("api_contract", "architecture", "documentation"):
+                    arch_reasons.append(gap.reason)
+        else:
+            _has_openapi = any(
+                p.endswith(("openapi.yaml", "openapi.yml", "openapi.json", "swagger.yaml", "swagger.json"))
+                or "swagger" in p.lower() or "springdoc" in p.lower()
+                for p in sm.file_paths
+            )
+            if not _has_openapi:
+                arch_reasons.append("No OpenAPI/Swagger spec found — API surface unverifiable")
+        if arch.bounded_contexts:
+            bc_count = len(arch.bounded_contexts)
+            arch_reasons.append(f"bounded_contexts detected: {bc_count}")
+        if arch_reasons:
+            reasons["architecture"] = arch_reasons
+
+    return reasons
+
+
 def _section_confidence(sm: SourceMap) -> dict[str, str]:
     cs = sm.confidence_summary
     dep_conf = "low"
@@ -1481,9 +1517,16 @@ def agent_view(sm: SourceMap, *, full: bool = False) -> dict[str, Any]:
     result["architecture"] = _architecture_context(sm)
 
     # ── 3a. File relevance: evidence-backed, high-signal only ────────────────
-    relevant_files = _file_relevance(sm)
+    _FR_AGENT_CAP = 20
+    _total_paths = len(sm.file_paths)
+    relevant_files = _file_relevance(sm, limit=_FR_AGENT_CAP if not full else _total_paths)
     if relevant_files:
         result["file_relevance"] = relevant_files
+        if not full and _total_paths > _FR_AGENT_CAP:
+            result["file_relevance_hint"] = (
+                f"Showing top {_FR_AGENT_CAP}/{_total_paths} files by score. "
+                "Use --full to see all."
+            )
 
     # ── 3b. Monorepo package roles (when available), capped ──────────────────
     if sm.monorepo_packages:
@@ -1638,6 +1681,16 @@ def agent_view(sm: SourceMap, *, full: bool = False) -> dict[str, Any]:
             conf["anomalies"] = cs.anomalies
         result["confidence_summary"] = conf
 
+    # ── 6a. Confidence reasons: explain any low-confidence section ────────────
+    _conf_reasons = _confidence_reasons(sm)
+    if _conf_reasons:
+        result["confidence_reasons"] = _conf_reasons
+    elif (sm.architecture and sm.architecture.requested
+          and sm.architecture.confidence == "low"
+          and "confidence_summary" in result):
+        # No reasons found → low was overly conservative; upgrade to high
+        result["confidence_summary"].setdefault("sections", {})["architecture"] = "high"
+
     # ── 7. Analysis gaps ──────────────────────────────────────────────────────
     analysis_gaps: list[dict[str, Any]] = []
 

sourcecode-1.18.0/.continue-here.md

@@ -1,125 +0,0 @@
-# Continue Here — atlas-cli v1.14.0 defect patch (saint-server)
-
-**Paused:** 2026-05-13  
-**Repo:** `/Users/user/Downloads/atlas-cli`  
-**Branch:** master
-
----
-
-## Estado actual
-
-Sesión: patch de 15 defectos contra saint-server (Spring Boot 2.7 / Java 8 / 100 módulos DDD / 467 controllers). **Sin commit** (instrucción explícita: "no hagas commit"). 10 ficheros fuente modificados + 4 ficheros nuevos de test/fixtures.
-
----
-
-## Defectos corregidos esta sesión
-
-### CRITICAL
-| ID | Fix | Fichero |
-|----|-----|---------|
-| C1 | `_extract_first_useful_paragraph()` rechaza snippets de startup via `_STARTUP_RE` (≥2 hits) | `summarizer.py` |
-| C2 | `_bootstrap_structured()` emite `modules` (todos los DDD modules) en vez de `sample` (5 primeros) | `serializer.py` |
-| C3 | `safe_read_text()` con fallback UTF-8→ISO-8859-1→UTF-8/replace — elimina double-encoding de chars españoles | `tree_utils.py` + 3 consumidores |
-| C4 | `_scan_java_patterns()` escanea Java para `@interface` + `extends`, emite `custom_annotations[]` + `base_classes[]` | `architecture_analyzer.py` + `schema.py` |
-
-### HIGH
-| ID | Fix | Fichero |
-|----|-----|---------|
-| H1 | `_spring_profiles_context()` → `{detected, per_profile_variants, note}` | `serializer.py` |
-| H2 | `_is_dto_mapper()` separa `@Mapper` reales de DtoMapper. Añade `dto_mappers[]`, elimina falsos positivos en `missing_xml` | `serializer.py` |
-| H3 | `_transactional_summary()` emite todas las clases; trunca a 10 con `truncated: true` + `note` | `serializer.py` |
-| H4 | `_TASK_CONTENT_MAP` per-task filter en prepare-context. Fix `review-pr` suspected_areas filtrado a root | `cli.py` + `prepare_context.py` |
-| H5 | Fallback `git log -n 100` cuando scan por fecha retorna 0. Nuevo campo `hotspot_method` | `git_analyzer.py` + `schema.py` |
-
-### MEDIUM
-| ID | Fix | Fichero |
-|----|-----|---------|
-| M1 | `_security_surface_from_eps()` añade campo `schema` descriptor. Mantiene `resource_names` (compat) | `serializer.py` |
-| M2 | `_jndi_datasources()` escanea `application.yml` / `persistence.xml` para JNDI names | `serializer.py` |
-| M3 | `_file_relevance()` scoring: +10 entry points, +8 base classes, +6 security config, +2 utils | `serializer.py` |
-| M4 | `--agent` auto-habilita `architecture = True` | `cli.py` |
-| M5 | Help `--compact` actualizado: "1000–3000 tokens" (era "600–800") | `cli.py` |
-
-### LOW
-| ID | Fix | Fichero |
-|----|-----|---------|
-| L1 | `analysis_gaps` expandido: ADRs missing, OpenAPI missing, profile YAMLs sin comentarios | `confidence_analyzer.py` |
-
----
-
-## Ficheros modificados
-
-```
-src/sourcecode/architecture_analyzer.py   +147 líneas — C4
-src/sourcecode/cli.py                     +90 líneas  — H4, M4, M5
-src/sourcecode/code_notes_analyzer.py     +5 líneas   — C3
-src/sourcecode/confidence_analyzer.py     +54 líneas  — L1
-src/sourcecode/doc_analyzer.py            +4 líneas   — C3
-src/sourcecode/git_analyzer.py            +29 líneas  — H5
-src/sourcecode/schema.py                  +5 líneas   — C4, H5
-src/sourcecode/serializer.py              +245 líneas — C2, H1-H3, M1-M3
-src/sourcecode/summarizer.py              +26 líneas  — C1, C3
-src/sourcecode/tree_utils.py              +19 líneas  — C3 utility
-src/sourcecode/prepare_context.py         +5 líneas   — H4
-```
-
-### Ficheros nuevos
-```
-tests/test_encoding_regression.py         — 4 tests C3
-tests/test_task_differentiation.py        — 6 tests H4
-tests/fixtures/latin1_sample.java         — fixture UTF-8
-tests/fixtures/latin1_sample_iso.java     — fixture Latin-1
-```
-
----
-
-## Tests
-
-- **743 passed, 3 skipped** — 0 regresiones
-- Fallos pre-existentes excluidos:
-  - `test_block2_coverage.py::test_java_marked_unsupported` — bug pre-existente en test (`name` vs `symbol`)
-  - `test_dependency_analyzer_node_python.py::test_python_requirements_without_lockfile_keeps_declared_versions` — versión fixture incorrecta
-  - `test_pipeline_integrity.py::TestBenchmarkContamination::test_agent_splits_development_and_auxiliary_eps` — pre-existente
-
----
-
-## Pendientes para próxima sesión
-
-1. **Smoke test en saint-server real:**
-   ```bash
-   python3 run_cli.py /path/to/saint-server --compact --agent --git-context 2>/dev/null | python3 -c "
-   import json,sys; d=json.load(sys.stdin)
-   print('profiles:', d.get('spring_profiles'))
-   print('arch method:', d.get('architecture',{}).get('method'))
-   print('custom_annotations:', len(d.get('architecture',{}).get('custom_annotations',[])))
-   print('hotspot_method:', d.get('git_context',{}).get('hotspot_method'))
-   "
-   ```
-
-2. **Commit** cuando listo:
-   ```bash
-   git add src/sourcecode/ tests/test_encoding_regression.py tests/test_task_differentiation.py tests/fixtures/latin1_sample*.java
-   git commit -m "fix: 15 saint-server defects — C1-C4 critical, H1-H5 high, M1-M5 medium, L1 low"
-   ```
-
-3. **Pre-existing bugs** (no tocados):
-   - `_OPTIONAL_LABEL_MAP` en `architecture_summary.py` tiene valores en español
-   - `lombok` / `sqlite-jdbc` clasificados `role: runtime` incorrectamente
-   - `test_pipeline_integrity::test_agent_splits_development_and_auxiliary_eps` — `agent_view()` no emite `auxiliary_entry_points`
-
----
-
-## Para retomar
-
-```bash
-cd /Users/user/Downloads/atlas-cli
-git diff --stat  # verificar 10 ficheros modificados
-python3 -m pytest tests/ \
-  --ignore=tests/test_block2_coverage.py \
-  --ignore=tests/test_dependency_analyzer_node_python.py \
-  --ignore=tests/test_pipeline_integrity.py -q
-# Expected: 743 passed, 3 skipped
-```
-
----
-*Pausado 2026-05-13 — gsd:pause-work*