code-context-control 2.119.0__tar.gz → 2.121.0__tar.gz

{code_context_control-2.119.0/code_context_control.egg-info → code_context_control-2.121.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: code-context-control
-Version: 2.119.0
+Version: 2.121.0
 Summary: Local MCP code-intelligence for AI coding tools: surgical search/read/edit, agent-config version history, path-level access + masking guards, and a multi-project hub.
 Author-email: Dimitri Tselenchuk <dtselenc@gmail.com>
 License-Expression: Apache-2.0

{code_context_control-2.119.0 → code_context_control-2.121.0}/cli/c3.py

@@ -92,7 +92,7 @@ console = Console() if HAS_RICH else None
 # Config
 CONFIG_DIR = ".c3"
 CONFIG_FILE = ".c3/config.json"
-__version__ = "2.119.0"
+__version__ = "2.121.0"
 
 
 def _compress_file_cli(compressor, path, mode="smart", **kw):
@@ -1570,6 +1570,42 @@ def cmd_shell_eval(args):
         sys.exit(1)
 
 
+def cmd_map_eval(args):
+    """Run the file-map gold suite and compare it to its baseline.
+
+    Renders the hand-annotated fixtures through the renderer the MCP tool
+    uses and grades symbol recall, signatures, ranges and determinism. Exit
+    status 1 when the baseline is violated (a must_pass case failed, an
+    aggregate fell under its floor or rose over its ceiling, a complete map
+    grew). See docs/map-eval.md.
+    """
+    from services.bench import map_eval as me
+
+    try:
+        report = me.run_suite(args.suite, baseline=args.baseline)
+    except FileNotFoundError as e:
+        raise RuntimeError(str(e))
+
+    if args.update_baseline:
+        path = args.baseline or me.BUNDLED_BASELINES.get(report.suite)
+        if not path:
+            raise RuntimeError(
+                f"suite {report.suite!r} has no bundled baseline; pass --baseline PATH")
+        floors = json.loads(args.floors) if args.floors else None
+        ceilings = json.loads(args.ceilings) if args.ceilings else None
+        me.write_baseline(report, path, floors=floors, ceilings=ceilings)
+        report.baseline_violations, report.baseline_warnings = me.compare_to_baseline(
+            report, me.load_baseline(path))
+        print(f"baseline written: {path}")
+
+    if args.json:
+        print(json.dumps(report.to_dict(), indent=2))
+    else:
+        print(report.render())
+    if report.baseline_violations:
+        sys.exit(1)
+
+
 def _benchmark_extract_preview(full_path: Path, compressor: CodeCompressor, pattern: str = "", max_lines: int = 50) -> str:
     """Approximate c3_filter behavior for local benchmarking without MCP startup."""
     import re as _re
@@ -9261,6 +9297,7 @@ def main():
         "stats": cmd_stats,
         "search-eval": cmd_search_eval,
         "shell-eval": cmd_shell_eval,
+        "map-eval": cmd_map_eval,
         "benchmark": cmd_benchmark,
         "session-benchmark": cmd_session_benchmark,
         "benchmark-e2e": cmd_benchmark_e2e,

{code_context_control-2.119.0 → code_context_control-2.121.0}/cli/commands/parser.py

@@ -76,6 +76,19 @@ def build_parser(version: str, parse_cli_ide_arg):
                               help='With --update-baseline: JSON object of metric ceilings, e.g. \'{"bytes_p95": 18432}\'')
     p_shell_eval.add_argument("--json", action="store_true", help="Emit the full report as JSON")
 
+    p_map_eval = subparsers.add_parser(
+        "map-eval", help="Run the file-map gold suite (docs/map-eval.md)")
+    p_map_eval.add_argument("--suite", default="fixture",
+                            help="'fixture' (hand-annotated sources, CI gate) or a .jsonl path")
+    p_map_eval.add_argument("--baseline", default=None, help="Baseline JSON to compare against / write")
+    p_map_eval.add_argument("--update-baseline", action="store_true",
+                            help="Write aggregates + per-case status to the baseline (floors/ceilings are kept)")
+    p_map_eval.add_argument("--floors", default=None,
+                            help='With --update-baseline: JSON object of metric floors, e.g. \'{"symbol_recall_mean": 0.9}\'')
+    p_map_eval.add_argument("--ceilings", default=None,
+                            help='With --update-baseline: JSON object of metric ceilings, e.g. \'{"tokens_p95": 4000}\'')
+    p_map_eval.add_argument("--json", action="store_true", help="Emit the full report as JSON")
+
     p_compress = subparsers.add_parser("compress", help="Compress a file")
     p_compress.add_argument("file", help="File to compress")
     p_compress.add_argument("--mode", choices=["map", "dense_map", "smart", "diff"], default="smart")

{code_context_control-2.119.0 → code_context_control-2.121.0}/cli/mcp_server.py

@@ -85,7 +85,7 @@ def _build_instructions(ide_name: str) -> str:
     return (
         "C3 — local code intelligence. Route by goal:\n"
         "  FIND candidates → c3_search\n"
-        "  MAP file shape → c3_compress(mode='map') then READ content → c3_read(symbols=…)\n"
+        "  MAP a file → c3_read(file_path) with no symbols, then READ → c3_read(symbols=…); batch maps → c3_compress\n"
         "  EDIT code → c3_edit (always; it logs to the ledger automatically)\n"
         "  VALIDATE after every edit → c3_validate\n"
         "  BLAST RADIUS before shared-symbol edits → c3_impact\n"
@@ -583,8 +583,10 @@ async def c3_memory(action: str, query: str = "", fact: str = "",
 @mcp.tool()
 async def c3_read(file_path: str, symbols: Any = None, lines: Any = None,
             include_docstrings: bool = True, ctx: Context = None) -> str:
-    """READ exact content. Use after c3_compress when you need real source (read-only, plan-mode safe).
-    file_path: single or comma-separated. symbols: function/class names. lines: int, [start,end], or list of ranges."""
+    """READ source, or MAP a file when called with only file_path (read-only, plan-mode safe).
+    file_path: single or comma-separated. symbols: names as the map prints them (Class.method ok).
+    lines: int, [start,end], or list of ranges. No symbols/lines = one line per symbol with
+    signature and [La-Lb] range at ~10% of the file's tokens; then read the symbols you need."""
     path_err = validate_file_path(file_path)
     if path_err:
         return f"[c3_read:error] {path_err}"
@@ -597,10 +599,11 @@ async def c3_read(file_path: str, symbols: Any = None, lines: Any = None,
 
 
 @mcp.tool()
-async def c3_compress(file_path: str, mode: str = "smart", ctx: Context = None) -> str:
-    """MAP file shape — classes/functions/imports at 40-70% tokens (read-only, plan-mode safe).
-    Use before c3_read to know which symbols to fetch. Modes: map, dense_map, smart, diff, bug_scan, ast.
-    file_path: single or comma-separated."""
+async def c3_compress(file_path: str, mode: str = "map", ctx: Context = None) -> str:
+    """MAP a file — one line per symbol with full signature and [La-Lb] range, ~10% of the
+    file's tokens (read-only, plan-mode safe). Same map as c3_read(file_path) with no symbols;
+    use this form for a comma-separated batch (up to 10). mode: map (default); other modes
+    are legacy."""
     path_err = validate_file_path(file_path)
     if path_err:
         return f"[c3_compress:error] {path_err}"

{code_context_control-2.119.0 → code_context_control-2.121.0}/cli/tools/agent.py

@@ -877,11 +877,10 @@ def _parallel_compress(files: list, svc, steps_log: list,
     maps = {}
 
     def compress_one(fp):
-        full_path = str(Path(svc.project_path) / fp)
         try:
-            res = svc.compressor.compress_file(full_path, "map")
-            if isinstance(res, dict) and res.get("compressed"):
-                return fp, res["compressed"]
+            fmap = svc.file_memory.get_or_build_map(fp.replace("\\", "/"))
+            if fmap and not fmap.startswith("[file_map]"):
+                return fp, fmap
         except Exception:
             pass
         return fp, None
@@ -936,11 +935,10 @@ def _parallel_validate_and_compress(files: list, svc, steps_log: list):
             return "validate", fp, "ERROR", str(e)
 
     def compress_one(fp):
-        full_path = str(Path(svc.project_path) / fp)
         try:
-            res = svc.compressor.compress_file(full_path, "map")
-            if isinstance(res, dict) and res.get("compressed"):
-                return "compress", fp, res["compressed"]
+            fmap = svc.file_memory.get_or_build_map(fp.replace("\\", "/"))
+            if fmap and not fmap.startswith("[file_map]"):
+                return "compress", fp, fmap
         except Exception:
             pass
         return "compress", fp, None

{code_context_control-2.119.0 → code_context_control-2.121.0}/cli/tools/compress.py

@@ -1,5 +1,7 @@
-"""c3_compress — Token-efficient file summaries (6 modes: map, dense_map, smart, diff, bug_scan, ast).
-Supports comma-separated paths for batch compression with parallel execution."""
+"""c3_compress — the canonical file map (services/file_map.render_map), one file or a batch.
+
+mode='map' (default) is the same text c3_read(file_path) serves. The other
+modes (dense_map, smart, diff, bug_scan, ast) are legacy and retire in 2.122.0."""
 
 import json
 import shutil
@@ -14,6 +16,48 @@ from core import count_tokens
 from services import access_guard
 
 
+def map_detail(svc, rel: str, requested_mode: str, *, cache_hit: bool,
+               symbols: int = 0, fallback: str = "") -> dict:
+    """Telemetry `detail` for a map served from file_memory (C0, 2.120.0).
+
+    backend names the code path, parser the extractor that built the record
+    (tree_sitter | regex | generic, "" for a pre-2.120.0 record), cache_hit
+    whether the record was already fresh. `symbols` is how many the caller
+    asked for and `fallback` why a read served a map instead of source.
+    Flat so telemetry.aggregate_tool_telemetry can fold it without a schema.
+    """
+    record = None
+    try:
+        record = svc.file_memory.get(rel)
+    except Exception:
+        record = None
+    if not isinstance(record, dict):
+        record = None
+    detail = {
+        "requested_mode": requested_mode,
+        "backend": "file_memory",
+        "parser": str((record or {}).get("parser") or ""),
+        "cache_hit": bool(cache_hit),
+        "sections": len((record or {}).get("sections") or []),
+    }
+    if symbols:
+        detail["symbols"] = int(symbols)
+    if fallback:
+        detail["fallback"] = fallback
+    return detail
+
+
+def compressor_detail(res: dict, requested_mode: str) -> dict:
+    """Telemetry `detail` for a CodeCompressor result (smart/diff/bug_scan…)."""
+    backend = "large_fast_path" if res.get("fast_path") else "compressor"
+    return {
+        "requested_mode": requested_mode,
+        "backend": backend,
+        "actual_mode": str(res.get("mode") or ""),
+        "cache_hit": bool(res.get("cache_hit")),
+    }
+
+
 def _run_memory_mcp_cli(args: list, cwd: str, timeout: int = 30) -> tuple:
     """Run codebase-memory-mcp CLI and return (success, output_or_error)."""
     binary = shutil.which("codebase-memory-mcp")
@@ -150,6 +194,8 @@ def _compress_single(file_path: str, mode: str, svc, finalize, maybe_facts) -> s
             return ("[file_map:error] not found. To create a new file use "
                     "c3_edit(file_path=..., old_string='', new_string=<content>).")
         rel = str(full.resolve().relative_to(Path(svc.project_path).resolve())).replace("\\", "/")
+        # C0 measurement: was the record already fresh before this call?
+        cache_hit = not svc.file_memory.needs_update(rel)
         queued = svc.file_memory.drain_queue()
         completed = []
         failed = []
@@ -174,8 +220,14 @@ def _compress_single(file_path: str, mode: str, svc, finalize, maybe_facts) -> s
         raw_tokens = None
         map_tokens = 0
         try:
-            raw_tokens = count_tokens(full.read_text(encoding="utf-8", errors="replace"))
+            raw_text = full.read_text(encoding="utf-8", errors="replace")
+            raw_tokens = count_tokens(raw_text)
             map_tokens = count_tokens(res)
+            if map_tokens >= raw_tokens and not res.startswith("[file_map]"):
+                # A map that costs more than the file is worth nothing
+                # (docs/file-map.md § Small files).
+                res = f"[compress:{file_path}] whole file — smaller than its map\n" + raw_text
+                map_tokens = count_tokens(res)
             summary = mode
         except Exception:
             summary = "mapped"
@@ -183,7 +235,8 @@ def _compress_single(file_path: str, mode: str, svc, finalize, maybe_facts) -> s
             finalize, svc, "c3_compress", {"file_path": file_path, "mode": mode},
             res, summary,
             raw_tokens=raw_tokens, optimized_tokens=map_tokens or None,
-            response_tokens=map_tokens)
+            response_tokens=map_tokens,
+            detail=map_detail(svc, rel, mode, cache_hit=cache_hit))
 
     try:
         res = svc.compressor.compress_file(str(full), mode)
@@ -197,7 +250,8 @@ def _compress_single(file_path: str, mode: str, svc, finalize, maybe_facts) -> s
         finalize, svc, "c3_compress", {"file_path": file_path},
         resp + maybe_facts(svc, Path(file_path).name), mode,
         raw_tokens=res.get('original_tokens'),
-        optimized_tokens=res.get('compressed_tokens'))
+        optimized_tokens=res.get('compressed_tokens'),
+        detail=compressor_detail(res, mode))
 
 
 def _compress_batch(paths: list, mode: str, svc, finalize, maybe_facts) -> str:
@@ -288,4 +342,6 @@ def _compress_batch(paths: list, mode: str, svc, finalize, maybe_facts) -> str:
         {"file_path": ",".join(paths), "mode": mode, "batch": True},
         body, f"batch {total_ok}/{len(paths)}",
         raw_tokens=total_raw if measured else None,
-        optimized_tokens=total_opt if measured else None)
+        optimized_tokens=total_opt if measured else None,
+        detail={"requested_mode": mode, "backend": "batch",
+                "files": len(paths), "ok": total_ok})

{code_context_control-2.119.0 → code_context_control-2.121.0}/cli/tools/read.py

@@ -8,6 +8,7 @@ from typing import Any
 
 from cli.tools import _grants
 from cli.tools._helpers import finalize_with_tokens, maybe_related_facts
+from cli.tools.compress import map_detail
 from core import count_tokens
 from services import access_guard
 
@@ -301,16 +302,25 @@ def handle_read(file_path: str, symbols: Any = None, lines: Any = None,
                 ranges.extend(main_ranges)
 
     if not ranges and symbols:
+        map_cached = not svc.file_memory.needs_update(rel_path)
         file_map = svc.file_memory.get_or_build_map(rel_path)
         resp = f"[read:{file_path}] symbols not found: {symbols}. Showing file map:\n{file_map}"
         map_tok = count_tokens(file_map)
         return finalize_with_tokens(
             finalize, svc, "c3_read", {"file": file_path, "symbols": symbols},
             resp, f"{full_file_tokens()}->{map_tok}tok",
-            raw_tokens=full_file_tokens(), optimized_tokens=map_tok)
+            raw_tokens=full_file_tokens(), optimized_tokens=map_tok,
+            detail=map_detail(svc, rel_path, "read", cache_hit=map_cached,
+                              symbols=len(symbols), fallback="symbols_not_found"))
 
     if not ranges:
+        map_cached = not svc.file_memory.needs_update(rel_path)
         file_map = svc.file_memory.get_or_build_map(rel_path)
+        if count_tokens(file_map) >= full_file_tokens():
+            # A map that costs more than the file is worth nothing: serve the
+            # file (docs/file-map.md § Small files).
+            file_map = (f"[read:{file_path}] whole file — smaller than its map\n"
+                        + raw_text)
         resp = (file_map
                 + "\n[map only — pass lines=[start,end] or symbols=[...] for exact source]"
                 + maybe_related_facts(svc, rel_path, top_k=3, context="read"))
@@ -319,7 +329,9 @@ def handle_read(file_path: str, symbols: Any = None, lines: Any = None,
             finalize, svc, "c3_read", {"file": file_path},
             resp, f"{full_file_tokens()}->{map_tok}tok",
             raw_tokens=full_file_tokens(), optimized_tokens=map_tok,
-            response_tokens=map_tok)
+            response_tokens=map_tok,
+            detail=map_detail(svc, rel_path, "read", cache_hit=map_cached,
+                              fallback="map_only"))
     else:
         # Sort and merge overlapping ranges
         ranges.sort()
@@ -359,7 +371,16 @@ def handle_read(file_path: str, symbols: Any = None, lines: Any = None,
     resp = final_content + maybe_related_facts(svc, rel_path, top_k=3, context="read")
     tokens = count_tokens(resp)
     summary = f"{full_file_tokens()}->{tokens}tok" if tokens < full_file_tokens() else f"{tokens}tok"
+    # C0 measurement: what the read asked for and how much source it took.
+    read_detail = {
+        "backend": "source",
+        "symbols": len(symbols) if symbols else 0,
+        "by_lines": bool(lines),
+        "ranges": len(ranges),
+        "lines_served": sum(e - s_ + 1 for s_, e in ranges),
+        "file_lines": len(content_lines),
+    }
     return finalize_with_tokens(
         finalize, svc, "c3_read", {"file": file_path, "symbols": symbols}, resp, summary,
         raw_tokens=full_file_tokens(), optimized_tokens=tokens,
-        response_tokens=tokens)
+        response_tokens=tokens, detail=read_detail)

{code_context_control-2.119.0 → code_context_control-2.121.0}/cli/tools/search.py

@@ -13,6 +13,7 @@ from pathlib import Path
 from cli.tools._helpers import finalize_with_tokens, show_token_ratios
 from core import count_tokens
 from services import access_guard
+from services.file_map import render_map
 from services.indexer import DEFAULT_CODE_EXTS
 from services.lexical_index import Filters, doc_kind
 from services.scanner import SKIP_DIRS, iter_files
@@ -473,21 +474,15 @@ def _first_file_map(rel: str, svc) -> str:
         watcher_active = (hasattr(svc, "watcher") and svc.watcher._observer.is_alive())
         if not watcher_active and svc.file_memory.needs_update(rel):
             svc.file_memory.update(rel)
-        fmap = svc.file_memory.get_or_build_map(rel)
+        record = svc.file_memory.get(rel)
+        if not record:
+            return ""
+        # Same renderer as c3_read/c3_compress, shortened to the cap by
+        # dropping symbol lines from the end (docs/file-map.md).
+        fmap = render_map(record, max_tokens=_MAP_TOKEN_CAP)
         if not fmap:
             return ""
-        if count_tokens(fmap) <= _MAP_TOKEN_CAP:
-            return f"\n  {fmap.replace(chr(10), chr(10) + '  ')}"
-        # Truncate large maps: keep first N lines
-        truncated = []
-        tok = 0
-        for fl in fmap.split("\n"):
-            tok += count_tokens(fl)
-            if tok > _MAP_TOKEN_CAP:
-                break
-            truncated.append(fl)
-        truncated.append("  [map truncated]")
-        return f"\n  {chr(10).join(truncated).replace(chr(10), chr(10) + '  ')}"
+        return f"\n  {fmap.replace(chr(10), chr(10) + '  ')}"
     except Exception:
         return ""
 
@@ -720,11 +715,13 @@ def _append_prefetch(resp: str, query: str, top_k: int, svc) -> str:
         uncached.append(fp)
 
     def compress_one(fp):
+        # The canonical map from file_memory — the same text c3_read serves
+        # (compressor.compress_file never had a "map" mode; it rendered a
+        # structure-only summary under that name).
         try:
-            full = str(Path(svc.project_path) / fp)
-            result = svc.compressor.compress_file(full, "map")
-            if isinstance(result, dict) and result.get("compressed"):
-                return fp, result["compressed"]
+            fmap = svc.file_memory.get_or_build_map(fp.replace("\\", "/"))
+            if fmap and not fmap.startswith("[file_map]"):
+                return fp, fmap
         except Exception:
             pass
         return fp, None

{code_context_control-2.119.0 → code_context_control-2.121.0/code_context_control.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: code-context-control
-Version: 2.119.0
+Version: 2.121.0
 Summary: Local MCP code-intelligence for AI coding tools: surgical search/read/edit, agent-config version history, path-level access + masking guards, and a multi-project hub.
 Author-email: Dimitri Tselenchuk <dtselenc@gmail.com>
 License-Expression: Apache-2.0

{code_context_control-2.119.0 → code_context_control-2.121.0}/code_context_control.egg-info/SOURCES.txt

@@ -225,6 +225,7 @@ services/edit_ledger.py
 services/embedding_index.py
 services/enforcement_policy.py
 services/error_reporting.py
+services/file_map.py
 services/file_memory.py
 services/git_context.py
 services/hub_service.py
@@ -288,6 +289,7 @@ services/vector_store.py
 services/watcher.py
 services/win_subprocess.py
 services/bench/__init__.py
+services/bench/map_eval.py
 services/bench/search_eval.py
 services/bench/shell_eval.py
 services/bench/external/__init__.py
@@ -356,6 +358,7 @@ tests/test_enforcement_routes.py
 tests/test_env_parser.py
 tests/test_failed_calls_do_not_unlock.py
 tests/test_federated_graph.py
+tests/test_file_map_c1.py
 tests/test_filter_backoff.py
 tests/test_ghost_files.py
 tests/test_ghost_generation.py
@@ -390,6 +393,8 @@ tests/test_jira_links.py
 tests/test_jira_routes.py
 tests/test_jira_tool.py
 tests/test_lazy_store_init.py
+tests/test_map_eval.py
+tests/test_map_telemetry_c0.py
 tests/test_mask_guard.py
 tests/test_mask_routes.py
 tests/test_mask_surfaces.py

{code_context_control-2.119.0 → code_context_control-2.121.0}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "code-context-control"
-version = "2.119.0"
+version = "2.121.0"
 description = "Local MCP code-intelligence for AI coding tools: surgical search/read/edit, agent-config version history, path-level access + masking guards, and a multi-project hub."
 readme = "README.md"
 requires-python = ">=3.10"
@@ -163,6 +163,8 @@ extend-exclude = [
     "commercial",
     "tui/build.bat",
     "tui/build.sh",
+    # map-eval fixtures are parser input, one of them deliberately malformed.
+    "tests/map_eval/fixtures",
 ]
 
 [tool.ruff.lint]

{code_context_control-2.119.0 → code_context_control-2.121.0}/services/agents.py

@@ -923,29 +923,10 @@ class FileMemoryAgent(BackgroundAgent):
         completed = []
         failed = []
         for rel_path in files_to_process:
-            ai_summary = None
-
-            # Generate AI summary if available
-            if self.ai_available:
-                record = self.file_memory.get(rel_path)
-                section_names = ""
-                if record:
-                    names = [s.get("name", "") for s in record.get("sections", [])
-                             if s.get("type") not in ("import", "decorator")]
-                    section_names = ", ".join(names[:10])
-
-                if section_names:
-                    ai_summary = self._ai_generate(
-                        f"File: {rel_path}\n"
-                        f"Symbols: {section_names}\n\n"
-                        f"Describe the purpose of this file and its main symbols in 1-2 concise sentences. "
-                        f"Focus on what they do, not how they are implemented.",
-                        system="You are a senior architect summarizing code for an AI assistant. "
-                               "Be technical, concise, and highlight the main responsibilities of the symbols.",
-                        max_tokens=100,
-                    )
-
-            result = self.file_memory.update(rel_path, ai_summary=ai_summary)
+            # No generated summary: it was produced from symbol names alone
+            # and half of the stored ones were cut mid-sentence (2026-09-06).
+            # The map carries symbols and ranges only (docs/file-map.md).
+            result = self.file_memory.update(rel_path)
             if result:
                 updated_count += 1
                 completed.append(rel_path)