execweave 0.8.9__tar.gz → 0.8.10__tar.gz

{execweave-0.8.9 → execweave-0.8.10}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.5
 Name: execweave
-Version: 0.8.9
+Version: 0.8.10
 Summary: Turn AI-agent runtime behavior into a local interactive execution graph.
 Project-URL: Homepage, https://github.com/Irish-kw/ExecWeave
 Project-URL: Repository, https://github.com/Irish-kw/ExecWeave
@@ -85,7 +85,7 @@ ExecWeave is a source-available, local-first observability project that turns AI
   <img src="docs/assets/codex.gif" alt="ExecWeave animated live demo" width="100%">
 </p>
 
-This README documents **v0.8.9**. It includes dagre route-point rendering and post-layout port recomputation, with provider graph/history/raw contracts verified across Ubuntu, macOS, and Windows.
+This README documents **v0.8.10**. This release fixes Windows Cursor launch resolution for arbitrary install roots and lets an ExecWeave-managed local `ollama serve` capture later loopback inference clients into the same semantic stream and dashboard.
 
 ## Why ExecWeave
 
@@ -126,7 +126,11 @@ execweave live --open -- python my_agent.py
 
 > **Approve the hook when prompted.** On the first provider-integrated run, the Agent/IDE may ask whether ExecWeave may enable its local hook integration. Choose **Allow / Yes**. Without approval, OS-runtime telemetry can still work, but provider-level tool, model, conversation, and supplied-content visibility may be reduced or unavailable.
 
-Google Antigravity currently uses the `agy` CLI command; ExecWeave also accepts `antigravity` as a friendly alias and resolves it to `agy`. For Cursor, `execweave live --open -- cursor` first tries a normal PATH launcher, then falls back to the standard desktop application binary on macOS and Windows.
+Google Antigravity currently uses the `agy` CLI command; ExecWeave also accepts `antigravity` as a friendly alias and resolves it to `agy`.
+
+For Cursor on Windows, bare `cursor` follows the PATH entry the user actually has and, when it is the normal `.../resources/app/bin/cursor.cmd` shim, derives the matching `Cursor.exe` from that same installation root. This works for custom drives and arbitrary install directories; explicit launcher paths are left untouched. macOS/Windows standard desktop paths remain a fallback when the PATH shim cannot identify the desktop binary.
+
+For a managed local Ollama server, `execweave live --open -- ollama serve` owns the public loopback endpoint and relays the real server through an internal loopback port. A later `ollama run`, Ollama SDK request, OpenAI-compatible client, or `curl` started from another terminal can therefore appear in the same ExecWeave semantic stream even though that client is not a child of the `ollama serve` process. Relay mode is deliberately limited to local `localhost` / `127.0.0.1` binds; external, wildcard, and IPv6 exposure is not rewritten.
 
 Build finalized run artifacts with:
 
@@ -170,9 +174,9 @@ These are presentation-layer changes. Raw graph evidence is unchanged, and Live,
 | Claude Code | Yes | native hooks + full-fidelity supplied hook content + exact subagent results when exposed |
 | OpenAI Codex | Yes | lifecycle hooks + validated rollout transcripts + agent-local task/message/final-response routing |
 | Google Antigravity / Antigravity CLI | Yes | passive native hooks + validated conversation/subagent routing where exposed |
-| Cursor | Yes | native hooks + exact subagent task/summary routing when exposed |
+| Cursor | Yes | native hooks + exact subagent task/summary routing when exposed; Windows bare-launch resolution follows the actual PATH installation |
 | OpenCode | Yes | project plugin + session/task routing + full-fidelity supplied plugin content |
-| Ollama | Yes | `execweave-model-runtime event/exchange/probe --runtime ollama` |
+| Ollama | Yes | managed local `ollama serve` loopback relay + `execweave-model-runtime event/exchange/probe --runtime ollama` |
 | llama.cpp | Yes | `execweave-model-runtime event/exchange/probe --runtime llamacpp` |
 | vLLM | Yes | `execweave-model-runtime event/exchange/probe --runtime vllm` |
 | LM Studio | Only when the local process is launched under ExecWeave | `execweave-model-runtime event/exchange/probe --runtime lmstudio` |

{execweave-0.8.9 → execweave-0.8.10}/README.de.md

@@ -32,7 +32,7 @@ ExecWeave ist ein source-available, local-first Observability-Projekt, das Aktiv
   <img src="docs/assets/codex.gif" alt="ExecWeave animated live demo" width="100%">
 </p>
 
-Dieses README dokumentiert **v0.8.9**.
+Dieses README dokumentiert **v0.8.10**.
 
 ## Warum ExecWeave
 

{execweave-0.8.9 → execweave-0.8.10}/README.fr.md

@@ -32,7 +32,7 @@ ExecWeave est un projet d’observabilité source-available et local-first qui t
   <img src="docs/assets/codex.gif" alt="ExecWeave animated live demo" width="100%">
 </p>
 
-Ce README décrit **v0.8.9**.
+Ce README décrit **v0.8.10**.
 
 ## Pourquoi ExecWeave
 

{execweave-0.8.9 → execweave-0.8.10}/README.ja.md

@@ -32,7 +32,7 @@ ExecWeave は source-available / local-first の observability プロジェク
   <img src="docs/assets/codex.gif" alt="ExecWeave animated live demo" width="100%">
 </p>
 
-この README は **v0.8.9** を説明します。
+この README は **v0.8.10** を説明します。
 
 ## ExecWeave を使う理由
 

{execweave-0.8.9 → execweave-0.8.10}/README.ko.md

@@ -32,7 +32,7 @@ ExecWeave는 source-available, local-first observability 프로젝트입니다.
   <img src="docs/assets/codex.gif" alt="ExecWeave animated live demo" width="100%">
 </p>
 
-이 README는 **v0.8.9**을 설명합니다.
+이 README는 **v0.8.10**을 설명합니다.
 
 ## 왜 ExecWeave인가
 

{execweave-0.8.9 → execweave-0.8.10}/README.md

@@ -32,7 +32,7 @@ ExecWeave is a source-available, local-first observability project that turns AI
   <img src="docs/assets/codex.gif" alt="ExecWeave animated live demo" width="100%">
 </p>
 
-This README documents **v0.8.9**. It includes dagre route-point rendering and post-layout port recomputation, with provider graph/history/raw contracts verified across Ubuntu, macOS, and Windows.
+This README documents **v0.8.10**. This release fixes Windows Cursor launch resolution for arbitrary install roots and lets an ExecWeave-managed local `ollama serve` capture later loopback inference clients into the same semantic stream and dashboard.
 
 ## Why ExecWeave
 
@@ -73,7 +73,11 @@ execweave live --open -- python my_agent.py
 
 > **Approve the hook when prompted.** On the first provider-integrated run, the Agent/IDE may ask whether ExecWeave may enable its local hook integration. Choose **Allow / Yes**. Without approval, OS-runtime telemetry can still work, but provider-level tool, model, conversation, and supplied-content visibility may be reduced or unavailable.
 
-Google Antigravity currently uses the `agy` CLI command; ExecWeave also accepts `antigravity` as a friendly alias and resolves it to `agy`. For Cursor, `execweave live --open -- cursor` first tries a normal PATH launcher, then falls back to the standard desktop application binary on macOS and Windows.
+Google Antigravity currently uses the `agy` CLI command; ExecWeave also accepts `antigravity` as a friendly alias and resolves it to `agy`.
+
+For Cursor on Windows, bare `cursor` follows the PATH entry the user actually has and, when it is the normal `.../resources/app/bin/cursor.cmd` shim, derives the matching `Cursor.exe` from that same installation root. This works for custom drives and arbitrary install directories; explicit launcher paths are left untouched. macOS/Windows standard desktop paths remain a fallback when the PATH shim cannot identify the desktop binary.
+
+For a managed local Ollama server, `execweave live --open -- ollama serve` owns the public loopback endpoint and relays the real server through an internal loopback port. A later `ollama run`, Ollama SDK request, OpenAI-compatible client, or `curl` started from another terminal can therefore appear in the same ExecWeave semantic stream even though that client is not a child of the `ollama serve` process. Relay mode is deliberately limited to local `localhost` / `127.0.0.1` binds; external, wildcard, and IPv6 exposure is not rewritten.
 
 Build finalized run artifacts with:
 
@@ -117,9 +121,9 @@ These are presentation-layer changes. Raw graph evidence is unchanged, and Live,
 | Claude Code | Yes | native hooks + full-fidelity supplied hook content + exact subagent results when exposed |
 | OpenAI Codex | Yes | lifecycle hooks + validated rollout transcripts + agent-local task/message/final-response routing |
 | Google Antigravity / Antigravity CLI | Yes | passive native hooks + validated conversation/subagent routing where exposed |
-| Cursor | Yes | native hooks + exact subagent task/summary routing when exposed |
+| Cursor | Yes | native hooks + exact subagent task/summary routing when exposed; Windows bare-launch resolution follows the actual PATH installation |
 | OpenCode | Yes | project plugin + session/task routing + full-fidelity supplied plugin content |
-| Ollama | Yes | `execweave-model-runtime event/exchange/probe --runtime ollama` |
+| Ollama | Yes | managed local `ollama serve` loopback relay + `execweave-model-runtime event/exchange/probe --runtime ollama` |
 | llama.cpp | Yes | `execweave-model-runtime event/exchange/probe --runtime llamacpp` |
 | vLLM | Yes | `execweave-model-runtime event/exchange/probe --runtime vllm` |
 | LM Studio | Only when the local process is launched under ExecWeave | `execweave-model-runtime event/exchange/probe --runtime lmstudio` |

{execweave-0.8.9 → execweave-0.8.10}/README.ru.md

@@ -32,7 +32,7 @@ ExecWeave — source-available, local-first проект наблюдаемос
   <img src="docs/assets/codex.gif" alt="ExecWeave animated live demo" width="100%">
 </p>
 
-Этот README описывает **v0.8.9**.
+Этот README описывает **v0.8.10**.
 
 ## Почему ExecWeave
 

{execweave-0.8.9 → execweave-0.8.10}/README.zh-CN.md

@@ -32,7 +32,7 @@ ExecWeave 是一个 source-available、local-first 的可观测性项目，把 A
   <img src="docs/assets/codex.gif" alt="ExecWeave animated live demo" width="100%">
 </p>
 
-本 README 对应 **v0.8.9**。
+本 README 对应 **v0.8.10**。
 
 ## 为什么是 ExecWeave
 

{execweave-0.8.9 → execweave-0.8.10}/README.zh-TW.md

@@ -32,7 +32,7 @@ ExecWeave 是一個 source-available、local-first 的可觀測性專案，會
   <img src="docs/assets/codex.gif" alt="ExecWeave animated live demo" width="100%">
 </p>
 
-本 README 對應 **v0.8.9**。
+本 README 對應 **v0.8.10**。
 
 ## 為什麼是 ExecWeave
 

{execweave-0.8.9 → execweave-0.8.10}/pyproject.toml

@@ -4,8 +4,8 @@ build-backend = "hatchling.build"
 
 [project]
 name = "execweave"
-# ExecWeave v0.8.9 release metadata
-version = "0.8.9"
+# ExecWeave v0.8.10 release metadata
+version = "0.8.10"
 description = "Turn AI-agent runtime behavior into a local interactive execution graph."
 readme = "README.md"
 requires-python = ">=3.10"

{execweave-0.8.9 → execweave-0.8.10}/src/execweave/__init__.py

@@ -1,3 +1,3 @@
 """ExecWeave runtime graph package."""
 
-__version__ = "0.8.9"
+__version__ = "0.8.10"

{execweave-0.8.9 → execweave-0.8.10}/src/execweave/auto_specialized.py

@@ -2,12 +2,13 @@ from __future__ import annotations
 
 import json
 import os
+import socket
 import threading
 import time
 from contextlib import contextmanager
 from dataclasses import dataclass
 from pathlib import Path
-from typing import Iterator
+from typing import Any, Iterator
 from urllib.error import HTTPError, URLError
 from urllib.parse import urlsplit, urlunsplit
 from urllib.request import Request, urlopen
@@ -22,6 +23,18 @@ from .model_runtime import (
 
 _SEMANTIC_ENV = "EXECWEAVE_SEMANTIC_SIDECAR"
 _OLLAMA_DEFAULT_ENDPOINT = "http://127.0.0.1:11434"
+_OLLAMA_INFERENCE_PATHS = frozenset(
+    {
+        "/api/chat",
+        "/api/generate",
+        "/api/embed",
+        "/api/embeddings",
+        "/v1/chat/completions",
+        "/v1/completions",
+        "/v1/embeddings",
+        "/v1/responses",
+    }
+)
 _PROBE_INTERVAL_SECONDS = 0.50
 _PROBE_TIMEOUT_SECONDS = 0.35
 _PROBE_STARTUP_GRACE_SECONDS = 0.10
@@ -132,6 +145,45 @@ def _ollama_endpoint_from_environment() -> str | None:
     return _local_endpoint(hostname, port)
 
 
+def _ollama_serve_relay_address() -> tuple[str, int] | None:
+    """Return a loopback endpoint ExecWeave can safely own for ``ollama serve``."""
+    raw = os.environ.get("OLLAMA_HOST", "").strip()
+    if not raw:
+        return ("127.0.0.1", 11434)
+    candidate = raw if "://" in raw else f"http://{raw}"
+    try:
+        parsed = urlsplit(candidate)
+        port = parsed.port or 11434
+    except ValueError:
+        return None
+    if parsed.scheme != "http" or parsed.username is not None or parsed.password is not None:
+        return None
+    if parsed.path not in {"", "/"} or parsed.query or parsed.fragment:
+        return None
+    hostname = parsed.hostname
+    # Do not silently narrow wildcard/external/IPv6 exposure to IPv4 loopback.
+    if hostname not in {"127.0.0.1", "localhost"} or not (1 <= port <= 65535):
+        return None
+    return ("127.0.0.1", port)
+
+
+def _allocate_loopback_port() -> int:
+    with socket.socket(socket.AF_INET, socket.SOCK_STREAM) as sock:
+        sock.bind(("127.0.0.1", 0))
+        return int(sock.getsockname()[1])
+
+
+def _record_ollama_inference_exchange(config: Any, **kwargs: Any) -> None:
+    """Record inference exchanges while still relaying every Ollama API route."""
+    method = str(kwargs.get("method", "")).upper()
+    request_path = str(kwargs.get("request_path", ""))
+    if method != "POST" or urlsplit(request_path).path not in _OLLAMA_INFERENCE_PATHS:
+        return
+    from .http_proxy import record_exchange_fail_open
+
+    record_exchange_fail_open(config, **kwargs)
+
+
 def _lmstudio_post_probe_spec(command: list[str]) -> _ProbeSpec | None:
     if not _is_lmstudio_server_start(command):
         return None
@@ -248,12 +300,7 @@ def _run_ollama_probe(
 
 
 def prepare_post_command_specialized_probe(command: list[str]) -> _ProbeSpec | None:
-    """Prepare an attribution-safe post-command probe for short-lived launch CLIs.
-
-    LM Studio's `lms server start` exits after starting a persistent server. ExecWeave
-    only prepares this probe when the user supplied an explicit local port and no
-    compatible API is already observable at that endpoint before launch.
-    """
+    """Prepare an attribution-safe post-command probe for short-lived launch CLIs."""
     spec = _lmstudio_post_probe_spec(command)
     if spec is None:
         return None
@@ -292,40 +339,85 @@ def run_post_command_specialized_probe(
                 time.sleep(_POST_PROBE_RETRY_SECONDS)
 
 
-
 @contextmanager
-def auto_specialized_launch(command: list[str]) -> Iterator[dict[str, str]]:
-    """Prepare child-only launch wiring for supported transparent local integrations."""
+def auto_specialized_launch(
+    command: list[str],
+    *,
+    server_relay: bool = False,
+) -> Iterator[dict[str, str]]:
+    """Prepare child launch wiring for supported transparent local integrations.
+
+    ``server_relay`` is deliberately opt-in so direct library callers preserve the
+    existing contract. RuntimeCollector enables it for a real managed server run.
+    """
     environment = dict(os.environ)
     configured_sidecar = os.environ.get(_SEMANTIC_ENV)
-    if not configured_sidecar or not _is_ollama_run(command):
-        yield environment
-        return
-    upstream = _ollama_endpoint_from_environment()
-    if upstream is None:
+    wants_run_relay = _is_ollama_run(command)
+    wants_serve_relay = server_relay and _is_ollama_serve(command)
+    if not configured_sidecar or not (wants_run_relay or wants_serve_relay):
         yield environment
         return
 
     from .http_proxy import ExecWeaveHTTPProxyServer, ProxyConfig
 
     sidecar = Path(configured_sidecar).expanduser().resolve()
+
+    if wants_run_relay:
+        upstream = _ollama_endpoint_from_environment()
+        if upstream is None:
+            yield environment
+            return
+        try:
+            server = ExecWeaveHTTPProxyServer(
+                ("127.0.0.1", 0),
+                ProxyConfig(upstream=upstream, sidecar=sidecar, mode="ollama"),
+                recorder=_record_ollama_inference_exchange,
+            )
+        except OSError:
+            yield environment
+            return
+        thread = threading.Thread(
+            target=server.serve_forever,
+            kwargs={"poll_interval": 0.05},
+            name="execweave-ollama-run-relay",
+            daemon=True,
+        )
+        thread.start()
+        host, port = server.server_address[:2]
+        environment["OLLAMA_HOST"] = f"http://{host}:{port}"
+        try:
+            yield environment
+        finally:
+            server.shutdown()
+            server.server_close()
+            thread.join(timeout=2.0)
+        return
+
+    listen_address = _ollama_serve_relay_address()
+    if listen_address is None:
+        yield environment
+        return
+    listen_host, listen_port = listen_address
+    internal_port = _allocate_loopback_port()
+    upstream = f"http://127.0.0.1:{internal_port}"
     try:
         server = ExecWeaveHTTPProxyServer(
-            ("127.0.0.1", 0),
+            (listen_host, listen_port),
             ProxyConfig(upstream=upstream, sidecar=sidecar, mode="ollama"),
+            recorder=_record_ollama_inference_exchange,
         )
     except OSError:
+        # Fail open: preserve the child command's normal bind/error behavior.
         yield environment
         return
     thread = threading.Thread(
         target=server.serve_forever,
         kwargs={"poll_interval": 0.05},
-        name="execweave-ollama-run-relay",
+        name="execweave-ollama-serve-relay",
         daemon=True,
     )
     thread.start()
-    host, port = server.server_address[:2]
-    environment["OLLAMA_HOST"] = f"http://{host}:{port}"
+    environment["OLLAMA_HOST"] = upstream
     try:
         yield environment
     finally:
@@ -336,12 +428,7 @@ def auto_specialized_launch(command: list[str]) -> Iterator[dict[str, str]]:
 
 @contextmanager
 def auto_specialized_probe(command: list[str]) -> Iterator[None]:
-    """Run supported local specialized probes without affecting command execution.
-
-    The probe activates only when a caller supplies ExecWeave's per-run semantic
-    sidecar environment variable. Auto-probing is limited to known local server
-    launch commands and loopback endpoints. All failures are fail-open.
-    """
+    """Run supported local specialized probes without affecting command execution."""
     configured_sidecar = os.environ.get(_SEMANTIC_ENV)
     spec = _probe_spec(command)
     if not configured_sidecar or spec is None:

{execweave-0.8.9 → execweave-0.8.10}/src/execweave/collector.py

@@ -168,7 +168,7 @@ class RuntimeCollector:
         return_code = 1
         post_command_probe = prepare_post_command_specialized_probe(command)
         try:
-            with auto_specialized_launch(command) as launch_environment:
+            with auto_specialized_launch(command, server_relay=True) as launch_environment:
                 process = subprocess.Popen(
                     launch_command,
                     cwd=str(self.watch_root),

{execweave-0.8.9 → execweave-0.8.10}/src/execweave/command.py

@@ -37,7 +37,35 @@ def _cursor_desktop_candidates() -> list[Path]:
     return []
 
 
-def _resolve_cursor_desktop() -> str | None:
+def _prefer_cursor_desktop_launcher() -> bool:
+    return os.name == "nt"
+
+
+def _cursor_desktop_from_launcher(launcher: str | None) -> str | None:
+    """Resolve Cursor.exe beside a VS Code-style PATH shim."""
+    if not launcher:
+        return None
+    path = Path(launcher).expanduser()
+    bin_dir = path.parent
+    app_dir = bin_dir.parent
+    resources_dir = app_dir.parent
+    install_root = resources_dir.parent
+    if (
+        bin_dir.name.lower() != "bin"
+        or app_dir.name.lower() != "app"
+        or resources_dir.name.lower() != "resources"
+    ):
+        return None
+    candidate = install_root / "Cursor.exe"
+    if candidate.is_file():
+        return str(candidate.resolve())
+    return None
+
+
+def _resolve_cursor_desktop(*, path_launcher: str | None = None) -> str | None:
+    derived = _cursor_desktop_from_launcher(path_launcher)
+    if derived is not None:
+        return derived
     for candidate in _cursor_desktop_candidates():
         if candidate.is_file():
             return str(candidate.resolve())
@@ -54,16 +82,21 @@ def _resolve_executable(executable: str, *, path: str | None = None) -> str:
             raise FileNotFoundError(f"command executable not found: {executable!r}")
         resolved = str(candidate.resolve())
     else:
-        resolved = shutil.which(executable, path=path)
-        if resolved is None and executable.lower() == "cursor":
-            resolved = _resolve_cursor_desktop()
+        path_launcher = shutil.which(executable, path=path)
+        if executable.lower() == "cursor" and _prefer_cursor_desktop_launcher():
+            resolved = _resolve_cursor_desktop(path_launcher=path_launcher) or path_launcher
+        else:
+            resolved = path_launcher
+            if resolved is None and executable.lower() == "cursor":
+                resolved = _resolve_cursor_desktop()
         if resolved is None and executable.lower() == "antigravity":
             resolved = shutil.which("agy", path=path)
         if resolved is None:
             detail = ""
             if executable.lower() == "cursor":
                 detail = (
-                    "; ExecWeave also checked the standard Cursor desktop-app install paths"
+                    "; ExecWeave also checked the Cursor desktop binary beside the PATH "
+                    "launcher and the standard desktop-app install paths"
                 )
             elif executable.lower() == "antigravity":
                 detail = "; the current Antigravity CLI executable is 'agy'"
@@ -83,9 +116,15 @@ def resolve_launch_command(command: list[str], *, path: str | None = None) -> li
     ExecWeave launches commands without ``shell=True``. Normal executables use PATH
     lookup. On Windows, PATHEXT keeps npm-style ``.cmd``/``.bat`` launchers working.
 
-    ``cursor`` additionally falls back to the standard macOS/Windows desktop app
-    binaries when no CLI launcher exists on PATH. ``antigravity`` is accepted as a
-    friendly alias for Google's current Antigravity CLI executable, ``agy``.
+    Bare ``cursor`` on Windows prefers the desktop application binary. ExecWeave
+    first derives ``Cursor.exe`` from a VS Code-style PATH shim such as
+    ``<install>/resources/app/bin/cursor.cmd`` and then checks standard desktop
+    install paths. This keeps the GUI process and provider hooks in the ExecWeave
+    launch environment instead of treating a short-lived CLI shim as the observed
+    application lifetime. Explicit launcher paths are never rewritten.
+
+    ``antigravity`` is accepted as a friendly alias for Google's current
+    Antigravity CLI executable, ``agy``.
 
     An explicitly supplied PowerShell ``.ps1`` file is launched through pwsh or
     Windows PowerShell because CreateProcess cannot execute a PowerShell script by

execweave-0.8.10/tests/test_cursor_ollama_session_lifecycle.py

@@ -0,0 +1,176 @@
+from __future__ import annotations
+
+import json
+import socket
+import threading
+from http.server import BaseHTTPRequestHandler, ThreadingHTTPServer
+from pathlib import Path
+from urllib.request import Request, urlopen
+
+import pytest
+
+import execweave.command as command_module
+from execweave.auto_specialized import auto_specialized_launch
+from execweave.command import resolve_launch_command
+
+
+def _free_loopback_port() -> int:
+    with socket.socket(socket.AF_INET, socket.SOCK_STREAM) as sock:
+        sock.bind(("127.0.0.1", 0))
+        return int(sock.getsockname()[1])
+
+
+@pytest.mark.parametrize(
+    "install_parts",
+    [
+        ("custom-cursor",),
+        ("Apps With Spaces", "Cursor Portable"),
+        ("任意位置", "Cursor"),
+    ],
+)
+def test_windows_cursor_policy_prefers_desktop_from_same_install(
+    monkeypatch: pytest.MonkeyPatch,
+    tmp_path: Path,
+    install_parts: tuple[str, ...],
+) -> None:
+    # The install root is deliberately arbitrary. The resolver must derive the
+    # desktop binary from the PATH shim's relative package layout, never from a
+    # drive letter, username, or fixed install directory.
+    install = tmp_path.joinpath(*install_parts)
+    bin_dir = install / "resources" / "app" / "bin"
+    bin_dir.mkdir(parents=True)
+    shim = bin_dir / "cursor.cmd"
+    shim.write_text("@echo off\r\n", encoding="utf-8")
+    desktop = install / "Cursor.exe"
+    desktop.write_bytes(b"MZ")
+    desktop.chmod(0o755)
+
+    monkeypatch.setattr(
+        command_module.shutil,
+        "which",
+        lambda executable, path=None: str(shim) if executable == "cursor" else None,
+    )
+    monkeypatch.setattr(command_module, "_cursor_desktop_candidates", lambda: [])
+    monkeypatch.setattr(command_module, "_prefer_cursor_desktop_launcher", lambda: True)
+
+    assert resolve_launch_command(["cursor"]) == [str(desktop.resolve())]
+
+
+def test_explicit_cursor_launcher_path_is_not_rewritten(tmp_path: Path) -> None:
+    install = tmp_path / "custom-cursor"
+    bin_dir = install / "resources" / "app" / "bin"
+    bin_dir.mkdir(parents=True)
+    shim = bin_dir / "cursor.cmd"
+    shim.write_text("#!/bin/sh\nexit 0\n", encoding="utf-8")
+    shim.chmod(0o755)
+    desktop = install / "Cursor.exe"
+    desktop.write_bytes(b"MZ")
+
+    assert resolve_launch_command([str(shim), "--version"])[0] == str(shim.resolve())
+
+
+class _IndependentOllamaHandler(BaseHTTPRequestHandler):
+    def log_message(self, format: str, *args: object) -> None:
+        return
+
+    def do_POST(self) -> None:
+        if self.path != "/api/chat":
+            self.send_response(404)
+            self.end_headers()
+            return
+        length = int(self.headers.get("Content-Length", "0"))
+        request = json.loads(self.rfile.read(length).decode("utf-8"))
+        payload = {
+            "model": request.get("model", "tiny"),
+            "message": {"role": "assistant", "content": "serve relay answer"},
+            "done": True,
+        }
+        body = json.dumps(payload).encode("utf-8")
+        self.send_response(200)
+        self.send_header("Content-Type", "application/json")
+        self.send_header("Content-Length", str(len(body)))
+        self.end_headers()
+        self.wfile.write(body)
+
+
+def test_ollama_serve_relay_captures_client_outside_process_tree(
+    monkeypatch: pytest.MonkeyPatch,
+    tmp_path: Path,
+) -> None:
+    public_port = _free_loopback_port()
+    public_endpoint = f"http://127.0.0.1:{public_port}"
+    sidecar = tmp_path / "events.semantic.jsonl"
+    monkeypatch.setenv("EXECWEAVE_SEMANTIC_SIDECAR", str(sidecar))
+    monkeypatch.setenv("OLLAMA_HOST", public_endpoint)
+
+    upstream: ThreadingHTTPServer | None = None
+    thread: threading.Thread | None = None
+    try:
+        with auto_specialized_launch(
+            ["ollama", "serve"],
+            server_relay=True,
+        ) as environment:
+            assert environment["OLLAMA_HOST"] != public_endpoint
+            upstream_port = int(environment["OLLAMA_HOST"].rsplit(":", 1)[1])
+            upstream = ThreadingHTTPServer(
+                ("127.0.0.1", upstream_port),
+                _IndependentOllamaHandler,
+            )
+            thread = threading.Thread(target=upstream.serve_forever, daemon=True)
+            thread.start()
+
+            # This request deliberately uses the original/public endpoint rather
+            # than the child environment returned above. It models `ollama run`,
+            # an SDK, or curl started from another terminal after `ollama serve`.
+            request = Request(
+                public_endpoint + "/api/chat",
+                data=json.dumps(
+                    {
+                        "model": "tiny",
+                        "messages": [{"role": "user", "content": "outside client"}],
+                    }
+                ).encode("utf-8"),
+                headers={"Content-Type": "application/json"},
+                method="POST",
+            )
+            with urlopen(request, timeout=5) as response:
+                result = json.loads(response.read().decode("utf-8"))
+            assert result["message"]["content"] == "serve relay answer"
+
+        rows = [
+            json.loads(line)
+            for line in sidecar.read_text(encoding="utf-8").splitlines()
+            if line.strip()
+        ]
+        kinds = {
+            (row.get("target", {}).get("attributes") or {}).get("content_kind")
+            for row in rows
+            if isinstance(row, dict)
+        }
+        assert "model_runtime.ollama.request_messages" in kinds
+        assert "model_runtime.ollama.assistant_messages" in kinds
+        assert any(
+            row.get("attributes", {}).get("transport_relay_observed") is True
+            for row in rows
+            if isinstance(row, dict)
+        )
+    finally:
+        if upstream is not None:
+            upstream.shutdown()
+            upstream.server_close()
+        if thread is not None:
+            thread.join(timeout=2)
+
+
+def test_ollama_serve_relay_does_not_change_remote_bind(
+    monkeypatch: pytest.MonkeyPatch,
+    tmp_path: Path,
+) -> None:
+    monkeypatch.setenv("EXECWEAVE_SEMANTIC_SIDECAR", str(tmp_path / "semantic.jsonl"))
+    monkeypatch.setenv("OLLAMA_HOST", "http://192.0.2.10:11434")
+
+    with auto_specialized_launch(
+        ["ollama", "serve"],
+        server_relay=True,
+    ) as environment:
+        assert environment["OLLAMA_HOST"] == "http://192.0.2.10:11434"

{execweave-0.8.9 → execweave-0.8.10}/tests/test_v069_dashboard_release.py

@@ -103,10 +103,10 @@ def test_release_version_and_noncommercial_license_metadata_are_078() -> None:
     stage-integrity node-ID floor refuses renames; the assertions below are the
     current release.
     """
-    assert __version__ == "0.8.9"
+    assert __version__ == "0.8.10"
     pyproject = Path("pyproject.toml").read_text(encoding="utf-8")
-    assert 'version = "0.8.9"' in pyproject
-    assert "ExecWeave v0.8.9 release metadata" in pyproject
+    assert 'version = "0.8.10"' in pyproject
+    assert "ExecWeave v0.8.10 release metadata" in pyproject
     assert "License :: Other/Proprietary License" in pyproject
     assert "License :: OSI Approved :: MIT License" not in pyproject