codecortex 0.23.3__tar.gz → 0.24.0__tar.gz

{codecortex-0.23.3/src/codecortex.egg-info → codecortex-0.24.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: codecortex
-Version: 0.23.3
+Version: 0.24.0
 Summary: Local-first, MCP-native code-intelligence server — graph, LSP, and semantic search behind one safe code.query tool for coding agents.
 Author: Shammai Hamilton
 License-Expression: MIT
@@ -31,6 +31,7 @@ Requires-Dist: tree-sitter-language-pack>=1.0
 Provides-Extra: dev
 Requires-Dist: pytest>=8; extra == "dev"
 Requires-Dist: pytest-cov>=5; extra == "dev"
+Requires-Dist: pytest-timeout>=2.3; extra == "dev"
 Requires-Dist: numpy>=1.24; extra == "dev"
 Requires-Dist: ruff>=0.16; extra == "dev"
 Requires-Dist: mypy>=1.11; extra == "dev"
@@ -53,6 +54,10 @@ Dynamic: license-file
 > been pointed at an unfamiliar codebase it has found real bugs. **Use it locally, on a developer
 > machine, for a single user** — that is the case it is built and tested for. Before relying on it
 > for anything beyond that, read **[Project status](#project-status)**.
+>
+> **New here? Read [docs/trust.md](docs/trust.md) first** — what each engine's answer is worth, a
+> five-minute check on a symbol whose callers you already know, and the eight states your
+> repository can be in with one command for each.
 
 ![codeintel's own call graph — an interactive, self-contained HTML view with force / radial / layered / module layouts, complexity-sized nodes, and click-to-inspect metrics.](docs/images/graph-codeintel.png)
 
@@ -432,7 +437,7 @@ Full system docs live in [`docs/`](docs/) — start with the index:
 - **[Install & registration](docs/install.md)** — what each agent host actually reads, why the registered command is an absolute path, and the three levels of proof that registration worked.
 - **[Query flow](docs/query-flow.md)** — request lifecycle, engine selection, fan-out & merge, and why it never throws.
 - **[Map file](docs/map-file.md)** — the static `CODE_INTEL.md` orientation layer for hosts with no MCP support.
-- **[Benchmarks](docs/benchmarks.md)** — real numbers at scale: 25 k chunks indexed in ~8 min, ~235 ms warm queries, 60 MB index.
+- **[Benchmarks](docs/benchmarks.md)** — real numbers at scale: 30 k chunks indexed in ~10 min, ~233 ms warm queries, 68 MB index.
 - Engine references: **[graph](docs/graph.md)** · **[lsp](docs/lsp.md)** · **[semantic](docs/semantic.md)**.
 
 ## CLI reference
@@ -448,7 +453,7 @@ Full system docs live in [`docs/`](docs/) — start with the index:
 | `codeintel serve-http [--host HOST] [--port 8766] [--allow-remote] [--token TOKEN]` | Start the HTTP transport (loopback-only unless `--allow-remote`; `--token` requires a bearer token on every request) |
 | `codeintel query --op OP --target TARGET [--engine auto] [--project-root DIR] [--json]` | Run a single query and print the result |
 | `codeintel status [project_root]` | Show engine availability and index age |
-| `codeintel doctor [project_root] [--deep] [--json]` | Diagnose per-engine health + repo index status, with a fix for each gap |
+| `codeintel doctor [project_root] [--deep] [--json]` | Diagnose per-engine health + repo index status, with a fix for each gap. `--deep` puts one **real query** to each engine and requires an answer with content — a process that starts is not a process that answers, and `runnable` means the second |
 | `codeintel map [project_root]` | Generate the `CODE_INTEL.md` orientation file |
 | `codeintel graph [project_root] [--html] [--out FILE] [--limit N]` | Emit the call graph as `{nodes,edges}` JSON, or `--html` a self-contained interactive viewer — see [docs/graph-viewer.md](docs/graph-viewer.md) |
 | `codeintel c4 [project_root] [--out DIR] [--depth N] [--scope PATH] [--include-tests] [--no-index] [--edges union\|imports] [--json] [--layers] [--suggest-config] [--check] [--layers-from auto\|inferred\|declared]` | Write a LikeC4 architecture model (`.c4`) of the repo's files and import graph. Indexes the repo first if it has no graph index. `--edges imports` drops the lower-confidence edges for a legible diagram; `--layers` reports inferred architectural bands; `--check` gates CI on a declared `[layers]` config, exit 2 on drift. See [An architecture model, as source](#an-architecture-model-as-source) and [docs/c4.md](docs/c4.md) |

{codecortex-0.23.3 → codecortex-0.24.0}/README.md

@@ -13,6 +13,10 @@
 > been pointed at an unfamiliar codebase it has found real bugs. **Use it locally, on a developer
 > machine, for a single user** — that is the case it is built and tested for. Before relying on it
 > for anything beyond that, read **[Project status](#project-status)**.
+>
+> **New here? Read [docs/trust.md](docs/trust.md) first** — what each engine's answer is worth, a
+> five-minute check on a symbol whose callers you already know, and the eight states your
+> repository can be in with one command for each.
 
 ![codeintel's own call graph — an interactive, self-contained HTML view with force / radial / layered / module layouts, complexity-sized nodes, and click-to-inspect metrics.](docs/images/graph-codeintel.png)
 
@@ -392,7 +396,7 @@ Full system docs live in [`docs/`](docs/) — start with the index:
 - **[Install & registration](docs/install.md)** — what each agent host actually reads, why the registered command is an absolute path, and the three levels of proof that registration worked.
 - **[Query flow](docs/query-flow.md)** — request lifecycle, engine selection, fan-out & merge, and why it never throws.
 - **[Map file](docs/map-file.md)** — the static `CODE_INTEL.md` orientation layer for hosts with no MCP support.
-- **[Benchmarks](docs/benchmarks.md)** — real numbers at scale: 25 k chunks indexed in ~8 min, ~235 ms warm queries, 60 MB index.
+- **[Benchmarks](docs/benchmarks.md)** — real numbers at scale: 30 k chunks indexed in ~10 min, ~233 ms warm queries, 68 MB index.
 - Engine references: **[graph](docs/graph.md)** · **[lsp](docs/lsp.md)** · **[semantic](docs/semantic.md)**.
 
 ## CLI reference
@@ -408,7 +412,7 @@ Full system docs live in [`docs/`](docs/) — start with the index:
 | `codeintel serve-http [--host HOST] [--port 8766] [--allow-remote] [--token TOKEN]` | Start the HTTP transport (loopback-only unless `--allow-remote`; `--token` requires a bearer token on every request) |
 | `codeintel query --op OP --target TARGET [--engine auto] [--project-root DIR] [--json]` | Run a single query and print the result |
 | `codeintel status [project_root]` | Show engine availability and index age |
-| `codeintel doctor [project_root] [--deep] [--json]` | Diagnose per-engine health + repo index status, with a fix for each gap |
+| `codeintel doctor [project_root] [--deep] [--json]` | Diagnose per-engine health + repo index status, with a fix for each gap. `--deep` puts one **real query** to each engine and requires an answer with content — a process that starts is not a process that answers, and `runnable` means the second |
 | `codeintel map [project_root]` | Generate the `CODE_INTEL.md` orientation file |
 | `codeintel graph [project_root] [--html] [--out FILE] [--limit N]` | Emit the call graph as `{nodes,edges}` JSON, or `--html` a self-contained interactive viewer — see [docs/graph-viewer.md](docs/graph-viewer.md) |
 | `codeintel c4 [project_root] [--out DIR] [--depth N] [--scope PATH] [--include-tests] [--no-index] [--edges union\|imports] [--json] [--layers] [--suggest-config] [--check] [--layers-from auto\|inferred\|declared]` | Write a LikeC4 architecture model (`.c4`) of the repo's files and import graph. Indexes the repo first if it has no graph index. `--edges imports` drops the lower-confidence edges for a legible diagram; `--layers` reports inferred architectural bands; `--check` gates CI on a declared `[layers]` config, exit 2 on drift. See [An architecture model, as source](#an-architecture-model-as-source) and [docs/c4.md](docs/c4.md) |

{codecortex-0.23.3 → codecortex-0.24.0}/pyproject.toml

@@ -53,7 +53,8 @@ Changelog = "https://github.com/hamilton-sky/codeintel/blob/main/CHANGELOG.md"
 codeintel = "codeintel.__main__:main"
 
 [project.optional-dependencies]
-dev = ["pytest>=8", "pytest-cov>=5", "numpy>=1.24", "ruff>=0.16", "mypy>=1.11"]
+dev = ["pytest>=8", "pytest-cov>=5", "pytest-timeout>=2.3", "numpy>=1.24", "ruff>=0.16",
+       "mypy>=1.11"]
 
 # `bench/agent_bench.py` only. Deliberately NOT in `dev`, and emphatically not a runtime dependency:
 # codeintel's own code makes zero outbound HTTP calls and that claim is load-bearing in the README.
@@ -161,3 +162,34 @@ follow_imports = "skip"
 # case in CI, and that costs ~0.4 points. A floor pinned to today's measurement would go red on a
 # runner's toolchain rather than on a real coverage regression. Ratchet it up deliberately.
 addopts = "--cov=codeintel --cov-report=term-missing:skip-covered --cov-fail-under=83"
+
+# No test may run indefinitely. The ceiling is deliberately far above any legitimate test rather
+# than tuned to the current slowest: its job is to convert a hang into a named failure, not to
+# police performance. Measured worst cases after the background-reindex guard landed —
+# `test_cold_process` spawns real cold interpreters at ~60s, `test_hard_exit` runs nested pytest
+# sessions at ~12s — so 180s leaves room for a slow CI runner and still fails in three minutes
+# instead of never.
+#
+# `signal` rather than `thread`: SIGALRM fails the ONE test that overran and lets the session carry
+# on to a normal report. pytest-timeout's `thread` method kills the whole process, which would
+# reproduce the thing being fixed — an unexplained death partway through — with better provenance
+# but no more information. A test that needs longer says so with
+# `@pytest.mark.timeout(<seconds>)` and, ideally, why.
+timeout = 180
+timeout_method = "signal"
+
+# A leaked socket, pipe or file handle is a defect, not a note. Promoted to an error once the
+# suite's own leaks were closed — sixteen HTTP listeners kept open by `shutdown()` without
+# `server_close()`, a template file, eight bare `open(...).read()` calls in `test_mapper`, and two
+# MCP children killed without being waited on or having their pipes closed. Leaving it at
+# `warning` is what let all of those accumulate: the count was in every run's output and had been
+# for months.
+#
+# `default::ResourceWarning:_pytest.unraisableexception` is NOT an exemption for our own leaks —
+# pytest surfaces a garbage-collected leak through that module rather than from the test that
+# caused it, and as an error it fails whichever unrelated test happened to trigger the collection.
+# Our leaks are caught at their own site by the first rule; this keeps the blame accurate.
+filterwarnings = [
+    "error::ResourceWarning",
+    "default::ResourceWarning:_pytest.unraisableexception",
+]

{codecortex-0.23.3 → codecortex-0.24.0/src/codecortex.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: codecortex
-Version: 0.23.3
+Version: 0.24.0
 Summary: Local-first, MCP-native code-intelligence server — graph, LSP, and semantic search behind one safe code.query tool for coding agents.
 Author: Shammai Hamilton
 License-Expression: MIT
@@ -31,6 +31,7 @@ Requires-Dist: tree-sitter-language-pack>=1.0
 Provides-Extra: dev
 Requires-Dist: pytest>=8; extra == "dev"
 Requires-Dist: pytest-cov>=5; extra == "dev"
+Requires-Dist: pytest-timeout>=2.3; extra == "dev"
 Requires-Dist: numpy>=1.24; extra == "dev"
 Requires-Dist: ruff>=0.16; extra == "dev"
 Requires-Dist: mypy>=1.11; extra == "dev"
@@ -53,6 +54,10 @@ Dynamic: license-file
 > been pointed at an unfamiliar codebase it has found real bugs. **Use it locally, on a developer
 > machine, for a single user** — that is the case it is built and tested for. Before relying on it
 > for anything beyond that, read **[Project status](#project-status)**.
+>
+> **New here? Read [docs/trust.md](docs/trust.md) first** — what each engine's answer is worth, a
+> five-minute check on a symbol whose callers you already know, and the eight states your
+> repository can be in with one command for each.
 
 ![codeintel's own call graph — an interactive, self-contained HTML view with force / radial / layered / module layouts, complexity-sized nodes, and click-to-inspect metrics.](docs/images/graph-codeintel.png)
 
@@ -432,7 +437,7 @@ Full system docs live in [`docs/`](docs/) — start with the index:
 - **[Install & registration](docs/install.md)** — what each agent host actually reads, why the registered command is an absolute path, and the three levels of proof that registration worked.
 - **[Query flow](docs/query-flow.md)** — request lifecycle, engine selection, fan-out & merge, and why it never throws.
 - **[Map file](docs/map-file.md)** — the static `CODE_INTEL.md` orientation layer for hosts with no MCP support.
-- **[Benchmarks](docs/benchmarks.md)** — real numbers at scale: 25 k chunks indexed in ~8 min, ~235 ms warm queries, 60 MB index.
+- **[Benchmarks](docs/benchmarks.md)** — real numbers at scale: 30 k chunks indexed in ~10 min, ~233 ms warm queries, 68 MB index.
 - Engine references: **[graph](docs/graph.md)** · **[lsp](docs/lsp.md)** · **[semantic](docs/semantic.md)**.
 
 ## CLI reference
@@ -448,7 +453,7 @@ Full system docs live in [`docs/`](docs/) — start with the index:
 | `codeintel serve-http [--host HOST] [--port 8766] [--allow-remote] [--token TOKEN]` | Start the HTTP transport (loopback-only unless `--allow-remote`; `--token` requires a bearer token on every request) |
 | `codeintel query --op OP --target TARGET [--engine auto] [--project-root DIR] [--json]` | Run a single query and print the result |
 | `codeintel status [project_root]` | Show engine availability and index age |
-| `codeintel doctor [project_root] [--deep] [--json]` | Diagnose per-engine health + repo index status, with a fix for each gap |
+| `codeintel doctor [project_root] [--deep] [--json]` | Diagnose per-engine health + repo index status, with a fix for each gap. `--deep` puts one **real query** to each engine and requires an answer with content — a process that starts is not a process that answers, and `runnable` means the second |
 | `codeintel map [project_root]` | Generate the `CODE_INTEL.md` orientation file |
 | `codeintel graph [project_root] [--html] [--out FILE] [--limit N]` | Emit the call graph as `{nodes,edges}` JSON, or `--html` a self-contained interactive viewer — see [docs/graph-viewer.md](docs/graph-viewer.md) |
 | `codeintel c4 [project_root] [--out DIR] [--depth N] [--scope PATH] [--include-tests] [--no-index] [--edges union\|imports] [--json] [--layers] [--suggest-config] [--check] [--layers-from auto\|inferred\|declared]` | Write a LikeC4 architecture model (`.c4`) of the repo's files and import graph. Indexes the repo first if it has no graph index. `--edges imports` drops the lower-confidence edges for a legible diagram; `--layers` reports inferred architectural bands; `--check` gates CI on a declared `[layers]` config, exit 2 on drift. See [An architecture model, as source](#an-architecture-model-as-source) and [docs/c4.md](docs/c4.md) |

{codecortex-0.23.3 → codecortex-0.24.0}/src/codecortex.egg-info/SOURCES.txt

@@ -18,10 +18,16 @@ src/codeintel/cache.py
 src/codeintel/config.py
 src/codeintel/containment.py
 src/codeintel/doctor.py
+src/codeintel/filelock.py
 src/codeintel/gateway.py
+src/codeintel/graph_answer.py
 src/codeintel/graph_backend.py
+src/codeintel/graph_confidence.py
+src/codeintel/graph_edges.py
+src/codeintel/graph_ops.py
 src/codeintel/graph_render.py
 src/codeintel/graph_resolution.py
+src/codeintel/graph_targets.py
 src/codeintel/grapher.py
 src/codeintel/http_server.py
 src/codeintel/indexer.py
@@ -35,6 +41,7 @@ src/codeintel/metrics.py
 src/codeintel/onboarding.py
 src/codeintel/outcome.py
 src/codeintel/paths.py
+src/codeintel/per_thread.py
 src/codeintel/policy.py
 src/codeintel/progress.py
 src/codeintel/provider.py
@@ -66,6 +73,7 @@ src/codeintel/commands/serve.py
 src/codeintel/commands/serve_http.py
 src/codeintel/commands/setup.py
 src/codeintel/commands/status.py
+src/codeintel/commands/uninstall.py
 src/codeintel/providers/__init__.py
 src/codeintel/providers/graph.py
 src/codeintel/providers/lsp.py
@@ -76,6 +84,8 @@ src/codeintel/viewer/graph_template.html
 tests/test_agent_prompt.py
 tests/test_backend_reaper.py
 tests/test_background_index_failure.py
+tests/test_bench_corpus_typed.py
+tests/test_bench_gate.py
 tests/test_bench_oracle.py
 tests/test_bench_oracle_ts.py
 tests/test_c4.py
@@ -95,6 +105,7 @@ tests/test_corpus.py
 tests/test_could_not_ask_semantic.py
 tests/test_docs_ci_claims.py
 tests/test_docs_deadcode_withdrawal.py
+tests/test_docs_trust_claims.py
 tests/test_doctor.py
 tests/test_e2e.py
 tests/test_edge_confidence.py
@@ -112,6 +123,7 @@ tests/test_http_auth.py
 tests/test_http_server.py
 tests/test_incompleteness.py
 tests/test_index_failure_reason.py
+tests/test_index_incompleteness.py
 tests/test_index_progress.py
 tests/test_indexer_binary.py
 tests/test_installer.py
@@ -124,21 +136,28 @@ tests/test_lsp_real.py
 tests/test_mapper.py
 tests/test_mcp_handshake.py
 tests/test_mcp_server.py
+tests/test_missing_target.py
 tests/test_model_dimension.py
 tests/test_never_raise.py
 tests/test_null_reason_and_ripple.py
 tests/test_onboarding.py
+tests/test_per_thread.py
+tests/test_query_budget.py
 tests/test_rbac.py
 tests/test_redaction_boundary.py
+tests/test_reindex_lock.py
 tests/test_reindexer.py
 tests/test_rerank.py
 tests/test_reset.py
 tests/test_semantic_provider.py
 tests/test_source_kind.py
 tests/test_staleness.py
+tests/test_summary_integrity.py
 tests/test_symbol_context.py
 tests/test_term.py
 tests/test_treesitter.py
+tests/test_uninstall.py
+tests/test_upgrade_path.py
 tests/test_verify.py
 tests/test_verify_call.py
 tests/test_version_skew.py

{codecortex-0.23.3 → codecortex-0.24.0}/src/codecortex.egg-info/requires.txt

@@ -9,6 +9,7 @@ anthropic>=0.40
 [dev]
 pytest>=8
 pytest-cov>=5
+pytest-timeout>=2.3
 numpy>=1.24
 ruff>=0.16
 mypy>=1.11

codecortex-0.24.0/src/codeintel/__init__.py

@@ -0,0 +1 @@
+__version__ = "0.24.0"

{codecortex-0.23.3 → codecortex-0.24.0}/src/codeintel/__main__.py

@@ -44,6 +44,7 @@ _COMMAND_GROUPS: list[tuple[str, list[tuple[str, str]]]] = [
         ("setup", "Prepare backends + index this repo (--all does it all)"),
         ("index", "Build the index every other command reads (semantic + graph)"),
         ("install", "Register codeintel with the AI agents on this machine"),
+        ("uninstall", "Remove those registrations — the index is left alone"),
         ("prompt", "Print a paste-to-your-agent setup prompt for this machine"),
     ]),
     ("Check health", [
@@ -79,6 +80,7 @@ _MODULES = {
     "setup": "setup",
     "index": "index",
     "install": "install",
+    "uninstall": "uninstall",
     "prompt": "prompt",
     "doctor": "doctor",
     "status": "status",
@@ -333,6 +335,13 @@ The one-shot path on a new machine.""",
 Writes the MCP server config so an agent can call codeintel. Then restart the
 agent (or start a new session) — a running host does not reload its MCP
 config. `codeintel doctor` lists what got registered where.""",
+    "uninstall": """examples:
+  codeintel uninstall
+  codeintel uninstall --dry-run         show what would go, change nothing
+
+Removes only codeintel's own entry from each agent config — neighbouring
+servers and your settings are left alone. The index is NOT deleted; use
+`codeintel reset --all` for that.""",
     "prompt": """examples:
   codeintel prompt
 
@@ -462,7 +471,8 @@ def build_parser() -> argparse.ArgumentParser:
         "--json", action="store_true", help="Emit the structured status and index-age report")
     status_parser.add_argument(
         "--deep", action="store_true",
-        help="Boot LSP and sample indexed source readability (slower, read-only)")
+        help="Put a real query to each engine and require an answer, boot the LSP, and sample "
+             "indexed source readability (slower, read-only)")
 
     # serve-http subcommand
     http_parser = subparsers.add_parser(
@@ -503,6 +513,22 @@ def build_parser() -> argparse.ArgumentParser:
         help="Show what would be registered, and where — write nothing",
     )
 
+    # uninstall subcommand
+    uninstall_parser = subparsers.add_parser(
+        "uninstall", help=_DESCRIPTIONS["uninstall"], description=_DESCRIPTIONS["uninstall"])
+    uninstall_parser.add_argument(
+        "--agent",
+        choices=["auto", "claude", "codex", "gemini", "zed", "all"],
+        default="auto",
+        help="Agent to remove from (default: auto — every agent codeintel is REGISTERED with, "
+             "which is a different set from the agents installed here)",
+    )
+    uninstall_parser.add_argument(
+        "--dry-run",
+        action="store_true",
+        help="Show what would be removed, and from where — change nothing",
+    )
+
     # map subcommand
     map_parser = subparsers.add_parser(
         "map", help=_DESCRIPTIONS["map"], description=_DESCRIPTIONS["map"])
@@ -575,7 +601,9 @@ def build_parser() -> argparse.ArgumentParser:
         "doctor", help=_DESCRIPTIONS["doctor"], description=_DESCRIPTIONS["doctor"])
     doctor_parser.add_argument("project_root", nargs="?", default=None, help="Project root (default: cwd)")
     doctor_parser.add_argument("--deep", action="store_true",
-                               help="Also boot-check serena (slower; first boot pulls it via uvx)")
+                               help="Ask each engine one REAL query and require content — a "
+                                    "process that starts is not one that answers. Also boots "
+                                    "serena (slower; first boot pulls it via uvx)")
     doctor_parser.add_argument("--json", action="store_true",
                                help="Emit the structured JSON report instead of the table")
 

{codecortex-0.23.3 → codecortex-0.24.0}/src/codeintel/c4.py

@@ -495,7 +495,8 @@ def build_c4_payload(project_root: Any, *, depth: int | None = None, scope: tupl
     payload at an auto-fit (or requested) directory depth. Never raises -- returns a flat `reason`
     string and empty lists on any failure."""
     try:
-        from codeintel.providers.graph import GraphProvider, _repo_display_name
+        from codeintel.graph_render import _repo_display_name
+        from codeintel.providers.graph import GraphProvider
 
         p = GraphProvider()
         if not getattr(p, "available", False):

codecortex-0.24.0/src/codeintel/commands/uninstall.py

@@ -0,0 +1,106 @@
+"""`codeintel uninstall` — remove codeintel's entry from the agent configs `install` wrote.
+
+The inverse of `install`, and deliberately only that. It does NOT delete the semantic index: that
+is `reset`'s job, it is expensive to rebuild (ten minutes on a large repo), and a command whose
+name says "uninstall" quietly destroying ten minutes of work is the kind of surprise this project
+spends most of its comments avoiding. What the index costs and how to remove it is printed instead,
+which is the same rule every other disclosure here follows — name what was NOT done.
+
+There is no confirmation prompt, and that is a considered difference from `reset --all`. This edits
+one entry in a config file and `codeintel install` puts it straight back; `reset` destroys an index
+that has to be re-embedded. `--dry-run` covers the reader who wants to look first.
+"""
+
+from typing import Any
+
+from codeintel.commands._common import never_raise
+
+
+def _targets(agent: str, installer_mod: Any) -> tuple[list[str], list[str]]:
+    """`(agents, untouched)` for the requested `--agent` selection.
+
+    `auto` means "wherever codeintel is actually registered", not "agents installed on this
+    machine" — the question `install` asks. They differ in both directions that matter: an agent
+    uninstalled from this machine can still hold our entry, and an agent that is present but never
+    registered has nothing to remove and should not be reported as though it did.
+    """
+    if agent == "auto":
+        registered = installer_mod.registered_agents()
+        return registered, [a for a in installer_mod._AGENTS if a not in registered]
+    if agent == "all":
+        return list(installer_mod._AGENTS), []
+    return [agent], []
+
+
+def _dry_run_line(installer_mod: Any, agent: str) -> str:
+    """What `--dry-run` would do to one agent's config — read-only, through the same lookup
+    `codeintel doctor` uses to notice a registration at all."""
+    spec = installer_mod._CONFIG[agent]
+    path, current = installer_mod.registered_command(spec)
+    if current is None:
+        return f"~ {agent}: no codeintel entry in {path} (no change)"
+    return f"- {agent}: would remove the codeintel entry from {path} (command: {current})"
+
+
+def _report(results: list) -> tuple[bool, bool, list[str]]:
+    """Print one line per agent; return (removed_any, failed_any, legacy_paths)."""
+    removed_any = failed_any = False
+    legacy_paths: list[str] = []
+    for r in results:
+        agent, path, action = r["agent"], r["path"], r["action"]
+        if action == "removed":
+            print(f"v {agent}: removed from {path}")
+            removed_any = True
+        elif action == "absent":
+            print(f"~ {agent}: nothing registered at {path}")
+        else:
+            print(f"x {agent}: failed — {r['reason']}")
+            failed_any = True
+        if r.get("legacy"):
+            legacy_paths.append(r["legacy"])
+    return removed_any, failed_any, legacy_paths
+
+
+@never_raise("uninstall failed: {exc}", code=1)
+def run(args: Any) -> int:
+    from codeintel import installer as installer_mod
+
+    agents, untouched = _targets(args.agent, installer_mod)
+
+    if args.agent == "auto" and not agents:
+        # Exit 0, not 1. The requested end state — codeintel is not registered with anything —
+        # already holds, and reporting "failure" for a goal that is met would make this command
+        # unusable in a teardown script that runs whether or not install ever ran.
+        print("codeintel is not registered with any agent on this machine — nothing to remove.")
+        print(f"  (looked at: {', '.join(untouched)})")
+        return 0
+
+    if getattr(args, "dry_run", False):
+        for agent in agents:
+            print(_dry_run_line(installer_mod, agent))
+        print("\nDry run — nothing was changed.")
+        return 0
+
+    results = installer_mod.Installer().unregister_many(agents)
+    removed_any, failed_any, legacy_paths = _report(results)
+
+    for legacy in dict.fromkeys(legacy_paths):
+        print(f"\n! stale entry: {legacy} has an `mcpServers.codeintel` block that this host does "
+              f"NOT read (an older codeintel wrote it). It is inert, and safe to delete by hand.")
+
+    if args.agent == "auto" and untouched:
+        print(f"\n- untouched (nothing registered): {', '.join(untouched)}")
+
+    if removed_any:
+        # A host that is already running keeps the server it launched at session start, so the
+        # tools stay on screen until it is restarted — the mirror of the note `install` prints,
+        # and the same surprise in the other direction.
+        print("\nRestart your agent (or start a new session) — a running host does not reload "
+              "its MCP config, so codeintel's tools stay visible until it does.")
+
+    # Named, not done. The index is the expensive artifact and this command does not own it.
+    from codeintel.paths import codeintel_home
+    print(f"\nThe index is untouched: {codeintel_home()}")
+    print("  Remove it with `codeintel reset --all` if you are done with codeintel entirely.")
+
+    return 1 if failed_any else 0

{codecortex-0.23.3 → codecortex-0.24.0}/src/codeintel/doctor.py

@@ -280,7 +280,7 @@ def run_doctor(
     try:
         from codeintel.providers.graph import GraphProvider
         engines["graph"] = _probe_engine(
-            "graph", graph, GraphProvider, lambda p: p.probe(root), on_provider
+            "graph", graph, GraphProvider, lambda p: p.probe(root, deep=deep), on_provider
         )
     except Exception:
         engines["graph"] = {"engine": "graph", "status": "fail", "installed": False,
@@ -414,7 +414,10 @@ def render_doctor_text(report: dict) -> str:
     ready, total, healthy = summ.get("ready", "?"), summ.get("total", "?"), summ.get("healthy")
     count = c.bold(f"{ready} / {total}")
     count = c.red(count) if healthy is False else (c.green(count) if healthy else count)
-    tail = "" if report.get("deep") else c.dim("  (run with --deep to boot-check serena)")
+    # Names what --deep actually does. "boot-check serena" described one third of it and the
+    # least useful third: booting is the question this tool keeps learning not to stop at.
+    tail = "" if report.get("deep") else c.dim(
+        "  (run with --deep to ask each engine a real query)")
     out.append("")
     out.append(f"  {count} engines ready for this repo.{tail}")
     opt_down = [n for n in _ENGINES

codecortex-0.24.0/src/codeintel/filelock.py

@@ -0,0 +1,120 @@
+"""A cross-process advisory lock, used to stop two codeintel processes reindexing one repo at once.
+
+MEASURED 2026-09-18: two `codeintel index` passes started together on a 600-file repository each
+embedded all 600 chunks and each reported "Indexed 600 chunks". There is no natural serialisation —
+both read the same "what is new" answer from the database before either had written anything, so
+the whole embedding cost is paid twice. On a large repository that is ten minutes of CPU doubled,
+and the two passes contend for the same cores while doing it.
+
+WHY `flock` AND NOT A FLAG FILE. The obvious way to record "a reindex is running" is to write a
+file and delete it at the end. That is strictly WORSE than keeping the state in memory: a process
+that is killed never runs its cleanup, so the flag survives forever and the tool reports a reindex
+in progress that is not. A signal true of a FILE and false of the WORLD is the defect this
+repository catalogues in `tests/test_summary_integrity.py`, and buying one to save some CPU would
+be a poor trade.
+
+`flock` has no such failure mode, and that is the entire reason it is the primitive here: the lock
+belongs to an open file descriptor, and **the kernel releases it when the process dies**, however it
+dies. There is no cleanup to skip and no staleness to age out. The lock file itself is left behind,
+which is fine — it is an empty file whose existence means nothing; only the lock on it does.
+
+DEGRADES, NEVER BLOCKS THE WORK. `fcntl` is POSIX and this package claims `OS Independent`, so
+every entry point here has a fallback, and the two fallbacks are deliberately asymmetric:
+
+* failing to ACQUIRE degrades to "you hold it" — indexing still happens, exactly as it did before
+  this module existed. A locking bug must never be able to stop a repository being indexed.
+* failing to PROBE degrades to "nobody holds it" — we do not claim a reindex is running on evidence
+  we could not gather. That is the same rule the envelope's `gaps` follow one level up.
+"""
+from __future__ import annotations
+
+import contextlib
+import logging
+import os
+import pathlib
+import re
+from collections.abc import Iterator
+
+logger = logging.getLogger(__name__)
+
+try:                                    # POSIX only; Windows has no fcntl
+    import fcntl
+    _HAVE_FLOCK = True
+except Exception:                       # pragma: no cover - platform dependent
+    fcntl = None                        # type: ignore[assignment]
+    _HAVE_FLOCK = False
+
+
+def lock_path(canonical_root: str) -> pathlib.Path:
+    """The lock file for one repository root.
+
+    Slugged the same way the graph cache names its files — every run of non-alphanumeric characters
+    collapses to a single `-` — which also flattens away every path separator, so a crafted root can
+    never place the lock outside the lock directory.
+    """
+    from codeintel.paths import codeintel_home
+
+    slug = re.sub(r"[^A-Za-z0-9]+", "-", canonical_root).strip("-") or "root"
+    return codeintel_home() / "locks" / f"reindex-{slug}.lock"
+
+
+@contextlib.contextmanager
+def exclusive(path: pathlib.Path) -> Iterator[bool]:
+    """Hold an exclusive, non-blocking lock on *path*. Yields whether we got it.
+
+    `False` means another LIVE process holds it — not that one once did. Callers use that to skip
+    work somebody else is already doing, so the answer has to be about the world rather than about
+    a file that happens to exist.
+    """
+    if not _HAVE_FLOCK:
+        # No advisory locking available: behave exactly as the code did before this module.
+        yield True
+        return
+
+    fd = None
+    try:
+        path.parent.mkdir(parents=True, exist_ok=True)
+        fd = os.open(str(path), os.O_CREAT | os.O_RDWR, 0o644)
+    except Exception as exc:
+        # An unwritable home is not a reason to stop indexing.
+        logger.debug("reindex lock unavailable at %s: %s", path, exc)
+        if fd is not None:
+            with contextlib.suppress(Exception):
+                os.close(fd)
+        yield True
+        return
+
+    acquired = False
+    try:
+        try:
+            fcntl.flock(fd, fcntl.LOCK_EX | fcntl.LOCK_NB)
+            acquired = True
+        except OSError:
+            acquired = False            # somebody else is mid-pass
+        yield acquired
+    finally:
+        with contextlib.suppress(Exception):
+            if acquired:
+                fcntl.flock(fd, fcntl.LOCK_UN)
+            os.close(fd)                # closing would release it anyway; explicit is clearer
+
+
+def held_by_another_process(path: pathlib.Path) -> bool:
+    """True when some other live process holds this lock.
+
+    Implemented by trying to take it and letting go immediately, because that is the only question
+    the kernel will answer — "is this lock free right now". The momentary acquisition is harmless:
+    a real pass that loses this race retries on the next debounce window, and the alternative
+    (recording holders in a file) is the stale-state design this module exists to avoid.
+
+    Degrades to False. Reporting "a reindex is running" on evidence we could not gather would be a
+    claim about the world made from a failed syscall.
+    """
+    if not _HAVE_FLOCK or not path.exists():
+        return False
+    try:
+        with exclusive(path) as got:
+            return not got
+    except Exception as exc:
+        logger.debug("reindex lock probe failed for %s: %s", path, exc)
+        return False