test: refresh oversized split plan

The three public webhook emitters in chat_helpers and webhook_routes
schedule deliveries via asyncio.create_task(webhook_manager.fire(...)),
which bypasses WebhookManager._bg_tasks. asyncio only holds a weak
reference to the outer task, so the GC can collect it mid-delivery and
the webhook is silently dropped.

Route all three through webhook_manager.fire_and_forget() so the task
is tracked by _spawn_tracked() and the manager owns the full lifecycle.

Adds an AST-level guard test that scans routes/ for direct
asyncio.create_task wrapping webhook_manager.fire(...) to prevent
regressions.

app.py

@@ -331,8 +331,8 @@ if AUTH_ENABLED:
                         request.state.current_user = "internal-tool"
                     request.state.api_token = False
                     return await call_next(request)
-            except Exception:
-                pass
+            except Exception as _e:
+                logger.warning("Internal tool auth header check failed", exc_info=_e)
             # Allow DIRECT localhost requests (internal service calls from
             # heartbeats etc.). Tunnel/proxy-forwarded requests are excluded by
             # _is_trusted_loopback so LOCALHOST_BYPASS can't be abused over a
@@ -385,11 +385,10 @@ if AUTH_ENABLED:
                                     _db.close()
                             try:
                                 await _asyncio.to_thread(_do)
-                            except Exception:
-                                pass
+                            except Exception as _e:
+                                logger.debug("Failed to update token last_used_at", exc_info=_e)
                         _asyncio.create_task(_touch_last_used(matched_id))
                         # Keep bearer-token callers out of normal cookie/user
-                        # routes. API-aware routes can read api_token_owner.
                         request.state.current_user = "api"
                         request.state.api_token = True
                         request.state.api_token_id = matched_id
@@ -464,8 +463,8 @@ async def serve_generated_image(filename: str, request: Request):
                 _db.close()
     except HTTPException:
         raise
-    except Exception:
-        pass
+    except Exception as _e:
+        logger.warning("Image ownership verification failed for %r", filename, exc_info=_e)
     ext = filename.rsplit('.', 1)[-1].lower()
     mime = {
         "png": "image/png", "jpg": "image/jpeg", "jpeg": "image/jpeg",

companion/routes.py

@@ -5,8 +5,9 @@ offers and pair to it, without duplicating any LLM logic.
 
 Auth is enforced globally by AuthMiddleware (app.py), so reaching a handler here
 means the caller is authenticated by either a cookie session or a Bearer `ody_`
-API token. The read endpoints (ping/info/models) accept either; the pairing
-endpoints are admin-cookie only.
+API token. Ping/info accept either credential type, models requires a chat-
+scoped API token for bearer callers, and the pairing endpoints are admin-cookie
+only.
 
 Pairing CSRF posture: minting happens ONLY on POST. The session cookie is
 SameSite=Lax (routes/auth_routes.py), which a browser does not send on a
@@ -18,7 +19,7 @@ on a GET would be unsafe (Lax cookies ride top-level GET navigations), so GET
 
 import html
 
-from fastapi import APIRouter, Request
+from fastapi import APIRouter, HTTPException, Request
 from fastapi.responses import HTMLResponse
 
 from core.middleware import require_admin
@@ -52,6 +53,18 @@ def owner_can_see(row_owner, owner) -> bool:
     return row_owner is None or row_owner == owner
 
 
+def require_models_scope(request: Request) -> None:
+    """Require the companion chat scope for bearer-token model inventory."""
+    if not getattr(request.state, "api_token", False):
+        return
+    scopes = getattr(request.state, "api_token_scopes", None) or []
+    if isinstance(scopes, str):
+        scopes = [scope.strip() for scope in scopes.split(",")]
+    scope_set = {str(scope).strip() for scope in scopes if str(scope).strip()}
+    if _pairing.COMPANION_SCOPE not in scope_set:
+        raise HTTPException(403, "API token requires chat scope")
+
+
 def mint_pairing_token(owner: str, invalidate=None) -> tuple[str, str]:
     """Mint a pairing token AND invalidate the auth middleware's in-memory token
     cache, so the new token is accepted on the very next request without a server
@@ -103,6 +116,7 @@ def setup_companion_routes() -> APIRouter:
         rows -- the same rule as owner_filter. Read-only; never returns api_key
         material.
         """
+        require_models_scope(request)
         import json as _json
 
         from core.database import SessionLocal, ModelEndpoint

core/database.py

@@ -2,12 +2,15 @@ import os
 import logging
 import sqlite3
 from datetime import datetime, timezone
+from pathlib import Path
 from sqlalchemy import event, create_engine, Column, String, Text, Boolean, DateTime, Integer, ForeignKey, JSON, Index, func, text
 from sqlalchemy.engine import Engine
 from sqlalchemy.types import TypeDecorator
 from sqlalchemy.ext.declarative import declarative_base, declared_attr
 from sqlalchemy.orm import relationship, sessionmaker, backref
 
+from src.runtime_paths import get_app_root
+
 logger = logging.getLogger(__name__)
 
 # Create base class for declarative models
@@ -29,9 +32,26 @@ class TimestampMixin:
     def updated_at(cls):
         return Column(DateTime, default=utcnow_naive, onupdate=utcnow_naive, nullable=False)
 
-# Get database URL from environment, default to SQLite in DATA_DIR
+# Ensure the writable data directory exists before SQLite connects.
 from src.constants import DATA_DIR, AUTH_FILE, MEMORY_FILE, USER_PREFS_FILE, SETTINGS_FILE
-DATABASE_URL = os.getenv("DATABASE_URL", f"sqlite:///{DATA_DIR}/app.db")
+Path(DATA_DIR).mkdir(parents=True, exist_ok=True)
+
+
+def _default_database_url() -> str:
+    return f"sqlite:///{Path(DATA_DIR) / 'app.db'}"
+
+
+def _normalize_sqlite_url(url: str) -> str:
+    if not url.startswith("sqlite:///"):
+        return url
+    db_path = url.replace("sqlite:///", "", 1)
+    if db_path == ":memory:" or os.path.isabs(db_path):
+        return url
+    return f"sqlite:///{(Path(get_app_root()) / db_path).resolve().as_posix()}"
+
+
+# Get database URL from environment, default to SQLite in DATA_DIR
+DATABASE_URL = _normalize_sqlite_url(os.getenv("DATABASE_URL", _default_database_url()))
 
 # Create engine
 engine = create_engine(

routes/api_token_routes.py

@@ -160,6 +160,8 @@ def setup_api_token_routes() -> APIRouter:
             payload = await request.json()
         except Exception:
             payload = {}
+        if not isinstance(payload, dict):
+            payload = {}
         with get_db_session() as db:
             token = db.query(ApiToken).filter(ApiToken.id == token_id).first()
             if not token:

routes/chat_helpers.py

@@ -14,7 +14,7 @@ from core.database import Session as DBSession, ModelEndpoint
 from src.llm_core import normalize_model_id
 from src.endpoint_resolver import normalize_base
 from src.context_compactor import maybe_compact, trim_for_context
-from src.auth_helpers import get_current_user
+from src.auth_helpers import effective_user
 from src.prompt_security import untrusted_context_message
 from routes.prefs_routes import _load_for_user as load_prefs_for_user
 
@@ -78,7 +78,7 @@ def _enforce_chat_privileges(request, sess) -> None:
     which means unrestricted allowed_models / zero cap -> no-op for them.
     """
     try:
-        user = get_current_user(request)
+        user = effective_user(request)
     except Exception:
         user = None
     if not user:
@@ -346,11 +346,11 @@ def add_user_message(sess, chat_handler, preprocessed: PreprocessedMessage, inco
 def fire_message_event(request, webhook_manager, session_id: str, sess, message: str, compare_mode: bool = False):
     """Fire webhook and event_bus events for a new user message."""
     if webhook_manager and not compare_mode:
-        asyncio.create_task(webhook_manager.fire("chat.message", {
+        webhook_manager.fire_and_forget("chat.message", {
             "session_id": session_id, "model": sess.model, "message": message[:2000],
-        }))
+        })
     from src.event_bus import fire_event
-    user = get_current_user(request)
+    user = effective_user(request)
     fire_event("message_sent", user)
 
 
@@ -577,7 +577,7 @@ async def build_chat_context(
         fire_message_event(request, webhook_manager, session_id, sess, message, compare_mode)
 
     # Resolve user prefs
-    user = get_current_user(request)
+    user = effective_user(request)
     uprefs = load_prefs_for_user(user)
 
     # Memory enabled?
@@ -1120,10 +1120,10 @@ def run_post_response_tasks(
 
     # Webhook
     if webhook_manager and not compare_mode:
-        asyncio.create_task(webhook_manager.fire("chat.completed", {
+        webhook_manager.fire_and_forget("chat.completed", {
             "session_id": session_id, "model": sess.model,
             "user_message": message, "response": full_response[:2000],
-        }))
+        })
 
     # Auto-name
     if needs_auto_name(sess.name):

routes/chat_routes.py

@@ -23,7 +23,7 @@ from src.endpoint_resolver import normalize_base as _normalize_base, build_chat_
 from src.session_search import search_session_messages
 from src.prompt_security import untrusted_context_message
 from core.exceptions import SessionNotFoundError
-from src.auth_helpers import get_current_user
+from src.auth_helpers import effective_user, get_current_user
 from routes.session_routes import _verify_session_owner
 from routes.document_helpers import _owner_session_filter
 from core.database import SessionLocal, get_session_mode, set_session_mode
@@ -126,7 +126,8 @@ def _clear_orphaned_session_endpoint(sess, owner: str | None = None) -> bool:
         sess.model = ""
         sess.headers = {}
         return True
-    except Exception:
+    except Exception as e:
+        logger.warning("Failed to clear orphaned session endpoint", exc_info=e)
         db.rollback()
         return False
     finally:
@@ -144,7 +145,8 @@ def _endpoint_cache_contains_model(endpoint, model: str) -> bool:
         return True
     try:
         models = json.loads(raw) if isinstance(raw, str) else raw
-    except Exception:
+    except Exception as e:
+        logger.warning("Failed to parse cached models list, treating as containing model", exc_info=e)
         return True
     if not isinstance(models, list) or not models:
         return True
@@ -236,7 +238,8 @@ def _recover_empty_session_model(sess, session_id: str, owner: str | None = None
                 is_chatgpt_subscription = False
         try:
             cached = json.loads(ep.cached_models) if isinstance(ep.cached_models, str) else (ep.cached_models or [])
-        except Exception:
+        except Exception as e:
+            logger.warning("Failed to parse cached_models for endpoint %r", getattr(ep, "id", "?"), exc_info=e)
             cached = []
         if not cached:
             visible = []
@@ -360,7 +363,7 @@ def setup_chat_routes(
             sess = session_manager.get_session(session)
         except KeyError:
             raise HTTPException(404, f"Session '{session}' not found")
-        owner = get_current_user(request)
+        owner = effective_user(request)
         if _clear_orphaned_session_endpoint(sess, owner=owner):
             raise HTTPException(400, "Selected model endpoint was removed. Pick another model in Settings.")
 
@@ -600,7 +603,7 @@ def setup_chat_routes(
             # but BEFORE loading. Prevents cross-user session hijack.
             _verify_session_owner(request, session)
             sess = session_manager.get_session(session)
-            owner = get_current_user(request)
+            owner = effective_user(request)
             if _clear_orphaned_session_endpoint(sess, owner=owner):
                 raise HTTPException(400, "Selected model endpoint was removed. Pick another model in Settings.")
             # Issue #587: picker shows a model from the endpoint cache but
@@ -631,7 +634,7 @@ def setup_chat_routes(
         _enforce_chat_privileges(request, sess)
 
         # Ensure session has auth headers
-        resolve_session_auth(sess, session, owner=get_current_user(request))
+        resolve_session_auth(sess, session, owner=effective_user(request))
 
         # Check for research_pending BEFORE mode persist overwrites it
         do_research = str(use_research).lower() == "true"
@@ -646,8 +649,8 @@ def setup_chat_routes(
         elif attachments:
             try:
                 att_ids = [str(x) for x in json.loads(attachments)]
-            except Exception:
-                pass
+            except Exception as e:
+                logger.warning("Failed to parse attachments JSON, ignoring attachments", exc_info=e)
 
         no_memory = str(form_data.get("no_memory", "")).lower() == "true"
         pre_context_tool_policy = build_effective_tool_policy(
@@ -1482,7 +1485,7 @@ def setup_chat_routes(
         if not q or not q.strip():
             return []
 
-        _user = get_current_user(request)
+        _user = effective_user(request)
         return [
             result.to_dict()
             for result in search_session_messages(

routes/cookbook_helpers.py

@@ -505,6 +505,8 @@ def _cached_model_scan_script(model_dirs: list[str] | None = None, add_hf_cache:
         "    if u.startswith('KB'): return int(n * 1024)",
         "    return int(n)",
         "def scan_ollama():",
+        "    if any(m.get('is_ollama') for m in models): return",
+        "    if os.name == 'nt' and not os.environ.get('ODYSSEUS_ALLOW_OLLAMA_CLI_SCAN'): return",
         "    if not shutil.which('ollama'): return",
         "    try:",
         "        p = subprocess.run(['ollama', 'list'], stdout=subprocess.PIPE, stderr=subprocess.DEVNULL, text=True, timeout=6)",
@@ -535,8 +537,8 @@ def _cached_model_scan_script(model_dirs: list[str] | None = None, add_hf_cache:
         "            models.append({'repo_id':name,'size_bytes':size_bytes,'nb_files':1,'has_incomplete':False,'path':'ollama','backend':'ollama','is_ollama':True})",
         "        return",
         "for _hf_cache in hf_cache_paths(): scan_hf(_hf_cache)",
-        "scan_ollama()",
         "scan_ollama_api()",
+        "scan_ollama()",
     ]
     for model_dir in model_dirs or []:
         lines.append(f"scan_dir(os.path.expanduser({model_dir!r}))")

routes/document_routes.py

@@ -503,7 +503,8 @@ def setup_document_routes(session_manager, upload_handler=None) -> APIRouter:
         user = get_current_user(request)
         try:
             data = await request.json()
-        except Exception:
+        except Exception as e:
+            logger.warning("Failed to parse export request body, defaulting to empty", exc_info=e)
             data = {}
         ids = data.get("ids") or []
         if not ids:
@@ -645,8 +646,8 @@ def setup_document_routes(session_manager, upload_handler=None) -> APIRouter:
                     try:
                         from src.agent_tools.document_tools import clear_active_document
                         clear_active_document(doc_id)
-                    except Exception:
-                        pass
+                    except Exception as e:
+                        logger.warning("Failed to clear active document %r on detach", doc_id, exc_info=e)
             db.commit()
             db.refresh(doc)
             return _doc_to_dict(doc)

routes/email_routes.py

@@ -79,15 +79,16 @@ def _email_tag_owner_aliases(account_id: str | None, owner: str = "") -> list[st
                         cfg.get("smtp_user") or "",
                         cfg.get("from_address") or "",
                     ])
-                except Exception:
+                except Exception as _e:
+                    logger.warning("Failed to resolve email account alias", exc_info=_e)
                     resolved_account_id = None
             row = db.get(_EA, resolved_account_id) if resolved_account_id else None
             if row:
                 aliases.extend([row.owner or "", row.imap_user or "", row.from_address or ""])
         finally:
             db.close()
-    except Exception:
-        pass
+    except Exception as _e:
+        logger.warning("Failed to load email aliases", exc_info=_e)
     out = []
     for a in aliases:
         a = (a or "").strip()

routes/embedding_routes.py

@@ -9,6 +9,7 @@ from pathlib import Path
 from fastapi import APIRouter, HTTPException, Form, Depends
 from core.constants import EMBEDDING_ENDPOINT_FILE, FASTEMBED_CACHE_DIR
 from core.middleware import require_admin
+from src.runtime_paths import get_app_root
 
 logger = logging.getLogger(__name__)
 

routes/session_routes.py

@@ -11,7 +11,7 @@ from core.session_manager import SessionManager
 from core.models import ChatMessage
 from src.request_models import SessionResponse
 from core.database import Session as DbSession, SessionLocal, Document, GalleryImage, utcnow_naive
-from src.auth_helpers import get_current_user, effective_user, _auth_disabled, owner_filter
+from src.auth_helpers import effective_user, _auth_disabled, owner_filter
 from src.session_actions import is_session_recently_active
 
 
@@ -328,7 +328,7 @@ def setup_session_routes(session_manager: SessionManager, config: dict, webhook_
         endpoint_id: str = Form(""),
     ):
         skip_val = str(skip_validation).lower() == "true"
-        user = get_current_user(request)
+        user = effective_user(request)
         endpoint_api_key = ""
         endpoint_base_url = ""
         _reject_raw_endpoint_url_for_non_admin(request, user, endpoint_id, endpoint_url)
@@ -477,7 +477,7 @@ def setup_session_routes(session_manager: SessionManager, config: dict, webhook_
                 db.close()
         # Switch model/endpoint mid-session
         if model is not None and endpoint_url is not None:
-            user = get_current_user(request)
+            user = effective_user(request)
             _reject_raw_endpoint_url_for_non_admin(request, user, endpoint_id, endpoint_url)
             endpoint_api_key = ""
             endpoint_base_url = ""

routes/upload_routes.py

@@ -7,7 +7,7 @@ from fastapi import APIRouter, Request, File, UploadFile, HTTPException
 from typing import List
 import logging
 from core.middleware import require_admin
-from src.auth_helpers import get_current_user
+from src.auth_helpers import effective_user
 from src.upload_handler import count_recent_uploads
 
 logger = logging.getLogger(__name__)
@@ -78,7 +78,7 @@ def setup_upload_routes(upload_handler):
         
         for u in files:
             try:
-                meta = upload_handler.save_upload(u, client_ip, owner=get_current_user(request))
+                meta = upload_handler.save_upload(u, client_ip, owner=effective_user(request))
                 out.append({
                     "id": meta["id"],
                     "name": meta["name"],
@@ -138,7 +138,7 @@ def setup_upload_routes(upload_handler):
                 original_name = info.get("name", file_id)
         auth_mgr = getattr(request.app.state, "auth_manager", None)
         auth_configured = bool(auth_mgr and auth_mgr.is_configured)
-        current_user = get_current_user(request)
+        current_user = effective_user(request)
         file_owner = info.get("owner") if info else None
         if auth_configured:
             if not current_user:
@@ -204,7 +204,7 @@ def setup_upload_routes(upload_handler):
         info = _load_upload_info(file_id)
         auth_mgr = getattr(request.app.state, "auth_manager", None)
         auth_configured = bool(auth_mgr and auth_mgr.is_configured)
-        current_user = get_current_user(request)
+        current_user = effective_user(request)
         file_owner = info.get("owner") if info else None
         if auth_configured:
             if not current_user:
@@ -247,7 +247,7 @@ def setup_upload_routes(upload_handler):
             raise HTTPException(404, "File not found")
         auth_mgr = getattr(request.app.state, "auth_manager", None)
         auth_configured = bool(auth_mgr and auth_mgr.is_configured)
-        current_user = get_current_user(request)
+        current_user = effective_user(request)
         file_owner = info.get("owner")
         if auth_configured:
             if not current_user:

routes/webhook_routes.py

@@ -1,6 +1,5 @@
 """Webhook, API Token, and sync chat routes."""
 
-import asyncio
 import uuid
 import logging
 from typing import Optional
@@ -385,10 +384,10 @@ def setup_webhook_routes(
         sess.add_message(ChatMessage("assistant", reply))
         session_manager.save_sessions()
 
-        asyncio.create_task(webhook_manager.fire("chat.completed", {
+        webhook_manager.fire_and_forget("chat.completed", {
             "session_id": session_id, "model": sess.model,
             "user_message": message[:2000], "response": reply[:2000],
-        }))
+        })
 
         return {"response": reply, "session_id": session_id, "model": sess.model}
 

services/hwfit/fit.py

@@ -19,6 +19,10 @@ GPU_BANDWIDTH = {
     "6950 xt": 576, "6900 xt": 512, "6800 xt": 512, "6800": 512, "6700 xt": 384, "6600 xt": 256, "6600": 224,
     "mi300x": 5300, "mi300": 5300, "mi250x": 3277, "mi250": 3277, "mi210": 1638, "mi100": 1229,
     "9070 xt": 624, "9070": 488, "9060 xt": 322, "9060": 322,
+    # NVIDIA GB10 Grace-Blackwell superchip (DGX Spark). Unified LPDDR5X memory,
+    # not Apple Silicon, so it lives in the generic GPU table — the Apple-only
+    # lookup never matches it (its name carries no "apple").
+    "gb10": 273,
 }
 
 # Pre-sort keys by length descending for correct substring matching

services/search/content.py

@@ -15,6 +15,8 @@ from urllib.parse import urljoin, urlparse
 import httpx
 from bs4 import BeautifulSoup
 
+from src.constants import WEB_FETCH_SOFT_MAX_BYTES, WEB_FETCH_HARD_MAX_BYTES
+
 from .analytics import RateLimitError, error_logger
 from .cache import (
     CONTENT_CACHE_DIR,
@@ -89,18 +91,128 @@ def _public_http_url(url: str) -> bool:
         return False
 
 
-def _get_public_url(url: str, headers: dict, timeout: int, max_redirects: int = 5) -> httpx.Response:
+class BodyTooLargeError(Exception):
+    """The server declared a body larger than the hard fetch ceiling."""
+
+    def __init__(self, url: str, declared_bytes: int):
+        self.url = url
+        self.declared_bytes = declared_bytes
+        super().__init__(
+            f"response body is {declared_bytes:,} bytes, over the "
+            f"{WEB_FETCH_HARD_MAX_BYTES:,}-byte hard cap"
+        )
+
+
+class _CappedFetch:
+    """Result of a size-capped streaming GET.
+
+    Carries just what fetch_webpage_content needs from an httpx.Response,
+    plus the cap bookkeeping: the (possibly truncated) body, whether the
+    cap cut it short, and the size the server declared via Content-Length
+    (wire bytes; None when absent).
+    """
+
+    __slots__ = ("status_code", "headers", "content", "truncated",
+                 "declared_bytes", "encoding", "url")
+
+    def __init__(self, status_code, headers, content, truncated,
+                 declared_bytes, encoding, url):
+        self.status_code = status_code
+        self.headers = headers
+        self.content = content
+        self.truncated = truncated
+        self.declared_bytes = declared_bytes
+        self.encoding = encoding
+        self.url = url
+
+    @property
+    def text(self) -> str:
+        return self.content.decode(self.encoding or "utf-8", errors="replace")
+
+    def raise_for_status(self):
+        if self.status_code >= 400:
+            request = httpx.Request("GET", self.url)
+            raise httpx.HTTPStatusError(
+                f"HTTP {self.status_code} for {self.url}",
+                request=request,
+                response=httpx.Response(self.status_code, request=request),
+            )
+
+
+def _get_public_url(url: str, headers: dict, timeout: int, max_redirects: int = 5,
+                    max_bytes: int = None) -> "_CappedFetch":
+    """Capped streaming GET with SSRF-guarded manual redirects.
+
+    The body is streamed and buffering stops at ``max_bytes`` (default: the
+    soft cap), so an oversized resource cannot be pulled into memory or the
+    content cache in full. When Content-Length already declares a body over
+    the hard ceiling, the fetch is refused before any body bytes are read.
+    """
+    cap = min(max_bytes or WEB_FETCH_SOFT_MAX_BYTES, WEB_FETCH_HARD_MAX_BYTES)
     current = url
     for _ in range(max_redirects + 1):
         if not _public_http_url(current):
             raise httpx.RequestError("Blocked private/internal URL", request=httpx.Request("GET", current))
-        response = httpx.get(current, headers=headers, timeout=timeout, follow_redirects=False)
-        if response.status_code not in (301, 302, 303, 307, 308):
-            return response
-        location = response.headers.get("location")
-        if not location:
-            return response
-        current = urljoin(str(response.url), location)
+        # Force identity transfer-encoding. With gzip/deflate the wire bytes
+        # (and Content-Length) can be a small fraction of the decoded body, so
+        # a tiny compressed response could pass the hard-cap preflight and then
+        # expand past the ceiling in a single decoded chunk before the streamed
+        # cap below can slice it. Identity makes Content-Length the true body
+        # size and keeps each streamed chunk bounded by the network read.
+        req_headers = dict(headers or {})
+        req_headers["Accept-Encoding"] = "identity"
+        with httpx.stream("GET", current, headers=req_headers, timeout=timeout,
+                          follow_redirects=False) as response:
+            if response.status_code in (301, 302, 303, 307, 308):
+                location = response.headers.get("location")
+                if not location:
+                    return _CappedFetch(response.status_code, response.headers, b"",
+                                        False, None, response.encoding, str(response.url))
+                current = urljoin(str(response.url), location)
+                continue
+
+            # A server can ignore the identity request and still return a
+            # compressed body; httpx.iter_bytes would then decode it, and a tiny
+            # gzip can balloon into one decoded chunk far past the cap before we
+            # slice. Refuse a compressed Content-Encoding so the streamed cap
+            # stays a real memory bound (Content-Length is the compressed wire
+            # length here, so the preflight and size metadata are unreliable too).
+            enc = (response.headers.get("content-encoding") or "").strip().lower()
+            if enc and enc != "identity":
+                raise httpx.RequestError(
+                    f"Refusing compressed response (Content-Encoding: {enc}) after "
+                    "requesting identity: cannot bound decoded body size",
+                    request=httpx.Request("GET", current),
+                )
+
+            declared = None
+            raw_len = response.headers.get("content-length")
+            if raw_len and raw_len.isdigit():
+                declared = int(raw_len)
+            # Refuse before buffering anything when the server already tells
+            # us the body exceeds the absolute ceiling (Content-Length is wire
+            # bytes; the decompressed body can only be larger).
+            if declared is not None and declared > WEB_FETCH_HARD_MAX_BYTES:
+                raise BodyTooLargeError(current, declared)
+
+            chunks = []
+            read = 0
+            truncated = False
+            # We requested identity above, so iter_bytes yields the raw body in
+            # network-read-sized chunks (no decompression expansion); the cap
+            # therefore bounds what we actually buffer.
+            for chunk in response.iter_bytes():
+                read += len(chunk)
+                if read > cap:
+                    keep = cap - (read - len(chunk))
+                    if keep > 0:
+                        chunks.append(chunk[:keep])
+                    truncated = True
+                    break
+                chunks.append(chunk)
+            return _CappedFetch(response.status_code, response.headers,
+                                b"".join(chunks), truncated, declared,
+                                response.encoding, str(response.url))
     raise httpx.RequestError("Too many redirects", request=httpx.Request("GET", current))
 
 # PDF extraction (optional dependency)
@@ -222,9 +334,19 @@ def _empty_result(url: str, error: str = "") -> dict:
 # ----------------------------------------------------------------------
 # Main content fetcher
 # ----------------------------------------------------------------------
-def fetch_webpage_content(url: str, timeout: int = 5, retry_attempt: int = 0) -> dict:
-    """Fetch and extract meaningful content from a webpage with caching."""
-    cache_key = generate_cache_key(url)
+def fetch_webpage_content(url: str, timeout: int = 5, retry_attempt: int = 0,
+                          max_bytes: int = None) -> dict:
+    """Fetch and extract meaningful content from a webpage with caching.
+
+    ``max_bytes`` raises the download budget per call (clamped to the hard
+    cap); the default is the soft cap. When the body is cut short the result
+    carries ``truncated``/``fetched_bytes``/``total_bytes`` so callers can
+    tell the model the content is partial (#3812).
+    """
+    effective_cap = min(max_bytes or WEB_FETCH_SOFT_MAX_BYTES, WEB_FETCH_HARD_MAX_BYTES)
+    # The cap is part of the cache identity: a truncated soft-cap fetch must
+    # not be served to a later full-budget request for the same URL.
+    cache_key = generate_cache_key(f"{url}#cap={effective_cap}")
     cache_file = CONTENT_CACHE_DIR / f"{cache_key}.cache"
 
     # Check cache
@@ -250,15 +372,21 @@ def fetch_webpage_content(url: str, timeout: int = 5, retry_attempt: int = 0) ->
             "User-Agent": "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/91.0.4472.124 Safari/537.36",
             "Accept": "text/html,application/xhtml+xml,application/xml;q=0.9,*/*;q=0.8",
             "Accept-Language": "en-US,en;q=0.5",
-            "Accept-Encoding": "gzip, deflate",
+            # identity so the streamed size cap in _get_public_url stays honest
+            # (a compressed body can decode to far more than Content-Length).
+            "Accept-Encoding": "identity",
             "Connection": "keep-alive",
         }
-        response = _get_public_url(url, headers=headers, timeout=timeout)
+        response = _get_public_url(url, headers=headers, timeout=timeout,
+                                   max_bytes=effective_cap)
 
         if response.status_code == 429:
             raise RateLimitError(f"Rate limit hit for {url} (attempt {retry_attempt})")
 
         response.raise_for_status()
+    except BodyTooLargeError as e:
+        error_logger.warning(f"Refused oversized body for {url}: {e}")
+        return _empty_result(url, f"TooLarge: {e}")
     except httpx.HTTPStatusError as e:
         error_logger.warning(f"HTTP {e.response.status_code} fetching {url}: {e}")
         return _empty_result(url, f"HTTP {e.response.status_code}: {e}")
@@ -269,9 +397,27 @@ def fetch_webpage_content(url: str, timeout: int = 5, retry_attempt: int = 0) ->
         error_logger.error(str(e))
         return _empty_result(url, str(e))
 
+    # Size bookkeeping shared by every content branch below. getattr keeps
+    # plain httpx.Response stand-ins (tests) working without the cap fields.
+    _size_fields = {
+        "truncated": getattr(response, "truncated", False),
+        "fetched_bytes": len(response.content),
+        "total_bytes": getattr(response, "declared_bytes", None),
+    }
+
     # PDF handling
     content_type = response.headers.get("Content-Type", "").lower()
     if "application/pdf" in content_type or url.lower().endswith(".pdf"):
+        if _size_fields["truncated"]:
+            # A PDF cut mid-stream is not parseable; unlike text there is no
+            # useful partial result, so report the budget problem instead.
+            _declared = _size_fields["total_bytes"]
+            return _empty_result(
+                url,
+                f"TooLarge: PDF exceeds the {effective_cap:,}-byte fetch budget"
+                + (f" (size {_declared:,} bytes)" if _declared else "")
+                + "; retry with a larger budget if it fits under the hard cap",
+            )
         if pdf_extract_text is None:
             logger.error("pdfminer.six is not installed; cannot extract PDF text.")
             pdf_text = ""
@@ -295,6 +441,7 @@ def fetch_webpage_content(url: str, timeout: int = 5, retry_attempt: int = 0) ->
             "js_message": "",
             "success": bool(pdf_text),
             "error": "" if pdf_text else "Failed to extract PDF text",
+            **_size_fields,
         }
         _cache_result(cache_file, cache_key, result, url)
         return result
@@ -329,6 +476,7 @@ def fetch_webpage_content(url: str, timeout: int = 5, retry_attempt: int = 0) ->
             "js_message": "",
             "success": bool(text_body),
             "error": "" if text_body else "Empty response body",
+            **_size_fields,
         }
         _cache_result(cache_file, cache_key, result, url)
         return result
@@ -391,6 +539,7 @@ def fetch_webpage_content(url: str, timeout: int = 5, retry_attempt: int = 0) ->
         "js_message": js_message,
         "success": True,
         "error": "",
+        **_size_fields,
     }
     _cache_result(cache_file, cache_key, result, url)
     return result

services/search/providers.py

@@ -9,14 +9,12 @@ from urllib.parse import urljoin, urlparse, parse_qs
 import httpx
 from bs4 import BeautifulSoup
 
-from src.constants import SEARXNG_INSTANCE
+from src.constants import SEARXNG_INSTANCE, REQUEST_TIMEOUT
 from .analytics import RateLimitError, error_logger
 from .query import build_enhanced_query
 
 logger = logging.getLogger(__name__)
 
-REQUEST_TIMEOUT = 20
-
 # Provider registry — maps setting value to (label, needs_key, needs_url)
 PROVIDER_INFO = {
     "searxng":  ("SearXNG",           False, True),

src/agent_loop.py

@@ -524,7 +524,7 @@ def get_builtin_overrides() -> dict:
         ov = get_setting("builtin_tool_overrides", {})
         return ov if isinstance(ov, dict) else {}
     except Exception as e:
-        logger.warning('Failed to load builtin tool overrides: %s', e)
+        logger.warning("Failed to load builtin tool overrides, using defaults", exc_info=e)
         return {}
 
 
@@ -929,8 +929,8 @@ def _build_system_prompt(
     try:
         from src.user_time import current_datetime_context_message
         _datetime_message = current_datetime_context_message()
-    except Exception:
-        pass
+    except Exception as e:
+        logger.warning("Failed to build datetime context message", exc_info=e)
 
     # Document context is kept as a SEPARATE message (not merged into the tool
     # prompt) so the context trimmer doesn't destroy it when truncating the
@@ -973,8 +973,8 @@ def _build_system_prompt(
             try:
                 from src.pdf_form_doc import find_source_upload_id
                 _is_form_backed = bool(find_source_upload_id(active_document.current_content or ""))
-            except Exception:
-                pass
+            except Exception as e:
+                logger.warning("Failed to detect if document is form-backed, assuming plain", exc_info=e)
 
             if _is_form_backed:
                 doc_ctx = (

src/agent_tools/web_tools.py

@@ -57,13 +57,23 @@ class WebSearchTool:
 class WebFetchTool:
     async def execute(self, content: str, ctx: dict) -> dict:
         from src.search.content import fetch_webpage_content
+        from src.constants import WEB_FETCH_HARD_MAX_BYTES
         raw = content.strip()
         url = ""
+        max_bytes = None
         if raw.startswith("{"):
             try:
                 parsed = json.loads(raw)
                 if isinstance(parsed, dict):
                     url = str(parsed.get("url") or "").strip()
+                    # Download-budget override (#3812): "full": true raises the
+                    # budget to the hard cap; an explicit max_bytes is clamped
+                    # to the hard cap downstream. Default stays the soft cap.
+                    if parsed.get("full") is True:
+                        max_bytes = WEB_FETCH_HARD_MAX_BYTES
+                    mb = parsed.get("max_bytes")
+                    if isinstance(mb, int) and mb > 0:
+                        max_bytes = mb
             except json.JSONDecodeError:
                 url = ""
         if not url:
@@ -78,7 +88,7 @@ class WebFetchTool:
         loop = asyncio.get_running_loop()
         try:
             result = await asyncio.wait_for(
-                loop.run_in_executor(None, lambda: fetch_webpage_content(url, timeout=10)),
+                loop.run_in_executor(None, lambda: fetch_webpage_content(url, timeout=10, max_bytes=max_bytes)),
                 timeout=30,
             )
         except asyncio.TimeoutError:
@@ -94,8 +104,28 @@ class WebFetchTool:
                 return {"error": f"web_fetch: {url}: {err}", "exit_code": 1}
             return {"error": f"web_fetch: {url}: no readable text content (not HTML, or the page needs JS/login)", "exit_code": 1}
 
+        # Tell the model when the download budget cut the body short and how
+        # to get the rest, instead of silently presenting a partial page as
+        # the whole thing.
+        size_note = ""
+        if result.get("truncated"):
+            fetched = result.get("fetched_bytes") or 0
+            total = result.get("total_bytes")
+            total_txt = f" of {total:,} bytes" if total else ""
+            size_note = (
+                f"[partial content: download stopped at {fetched:,} bytes{total_txt}. "
+                f'Re-call with {{"url": "{url}", "full": true}} to fetch up to '
+                f"{WEB_FETCH_HARD_MAX_BYTES:,} bytes.]\n\n"
+            )
+
+        # The notice must lead the output so the MAX_OUTPUT_CHARS trim below can
+        # never drop it. The title is untrusted, uncapped page content, so a
+        # giant title ahead of the notice could push it out of range; keep the
+        # notice first and cap the title as a second guard.
+        if len(title) > 300:
+            title = title[:300] + "..."
         header = (f"# {title}\n" if title else "") + f"Source: {url}\n\n"
-        output = header + text
+        output = size_note + header + text
         if len(output) > MAX_OUTPUT_CHARS:
             output = output[:MAX_OUTPUT_CHARS] + "\n\n[...truncated]"
         return {"output": output, "exit_code": 0}

src/builtin_mcp.py

@@ -14,6 +14,7 @@ import subprocess
 import sys
 
 from core.platform_compat import IS_WINDOWS, which_tool
+from src.runtime_paths import get_app_root
 
 logger = logging.getLogger(__name__)
 
@@ -81,7 +82,7 @@ _BUILTIN_NPX_SERVERS = {
         "name": "Built-in: Browser",
         "command": "npx",
         "args": ["-y", "@playwright/mcp@latest", "--headless", "--caps", "vision"],
-    },
+    }
 }
 
 # Global flag to disable MCP if there are compatibility issues
@@ -94,7 +95,7 @@ async def register_builtin_servers(mcp_manager):
         logger.info("Built-in MCP servers disabled via ODYSSEUS_DISABLE_MCP")
         return
 
-    base_dir = os.path.dirname(os.path.dirname(os.path.abspath(__file__)))
+    base_dir = get_app_root()
     python = sys.executable
 
     async def _connect_python_server(server_id: str, script_path: str, name: str):

src/config.py

@@ -5,6 +5,7 @@ from pydantic_settings import BaseSettings, SettingsConfigDict
 from pydantic import Field, field_validator
 
 from src.constants import DATA_DIR as _DATA_DIR_CONST
+from src.runtime_paths import get_app_root
 
 # Cross-platform OS flag, exposed here so callers can `from src.config import
 # IS_WINDOWS`. Defined locally (a trivial `os.name == "nt"`) rather than imported
@@ -19,7 +20,7 @@ IS_WINDOWS = os.name == "nt"
 class DataConfig(BaseSettings):
     """Configuration for data storage and file handling."""
     # Base directory
-    base_dir: Path = Field(default=Path(__file__).parent.parent, description="Base directory for the application")
+    base_dir: Path = Field(default=Path(get_app_root()), description="Base directory for the application")
     
     # Data paths
     data_dir: Path = Field(default=Path(_DATA_DIR_CONST), description="Main data directory")
@@ -138,7 +139,7 @@ class AppConfig(BaseSettings):
         if isinstance(v, dict) and "base_dir" in v:
             base_dir = v["base_dir"]
         else:
-            base_dir = Path(__file__).parent.parent
+            base_dir = Path(get_app_root())
         
         # Convert string paths to Path objects relative to base_dir
         data_dir = Path(_DATA_DIR_CONST)

src/constants.py

@@ -2,12 +2,14 @@
 """Application-wide constants and configuration values."""
 import os
 
+from src.runtime_paths import get_app_root, get_default_data_dir
+
 APP_VERSION = "1.0.0"
 
 # Base paths
-BASE_DIR = os.path.dirname(os.path.dirname(os.path.abspath(__file__))) + "/"
+BASE_DIR = os.path.join(get_app_root(), "")
 STATIC_DIR = os.path.join(BASE_DIR, "static")
-DATA_DIR = os.getenv("ODYSSEUS_DATA_DIR", os.path.join(BASE_DIR, "data"))
+DATA_DIR = os.getenv("ODYSSEUS_DATA_DIR", get_default_data_dir())
 
 # Data file paths
 # Single source of truth: every persisted file/dir lives under DATA_DIR, which
@@ -63,6 +65,14 @@ MAX_OUTPUT_CHARS = 10_000       # cap for bash/python/web_search/web_fetch outpu
 MAX_READ_CHARS = 20_000         # cap for read_file / document preview
 MAX_DIFF_LINES = 400            # cap for edit_file unified-diff display
 
+# web_fetch response-size policy (#3812). MAX_OUTPUT_CHARS above only trims
+# what the agent SEES; these caps bound what the server downloads, parses,
+# and writes to the content cache. The soft cap is the default download
+# budget; the agent can raise it per call (full/max_bytes) but never past
+# the hard cap, so a model can't decide to pull a multi-GB file.
+WEB_FETCH_SOFT_MAX_BYTES = 2_000_000    # default download budget (2 MB)
+WEB_FETCH_HARD_MAX_BYTES = 20_000_000   # absolute ceiling, even with override (20 MB)
+
 # API Configuration
 MAX_CONTEXT_MESSAGES = 90
 REQUEST_TIMEOUT = 20

src/embeddings.py

@@ -31,6 +31,8 @@ import numpy as np
 import httpx
 from typing import List, Optional
 
+from src.runtime_paths import get_app_root
+
 logger = logging.getLogger(__name__)
 
 _DEFAULT_MODEL = "all-minilm:l6-v2"

src/llm_core.py

@@ -283,7 +283,8 @@ def _is_ollama_native_url(url: str) -> bool:
     """Return True for native Ollama API URLs, including Ollama Cloud."""
     try:
         parsed = urlparse(url or "")
-    except Exception:
+    except Exception as e:
+        logger.warning("Failed to parse URL for Ollama detection", exc_info=e)
         return False
     host = parsed.hostname or ""
     path = (parsed.path or "").rstrip("/")
@@ -1345,8 +1346,8 @@ def list_model_ids(
                 r = httpx.get(root + "/api/tags", timeout=timeout)
                 r.raise_for_status()
                 return [m.get("name") or m.get("model") for m in (r.json().get("models") or []) if m.get("name") or m.get("model")]
-        except Exception:
-            pass
+        except Exception as e:
+            logger.warning("Failed to fetch model list from configured endpoint", exc_info=e)
         return []
 
 def normalize_model_id(

src/mcp_manager.py

@@ -11,6 +11,8 @@ import os
 import re
 from typing import Any, Dict, List, Optional, Set, Tuple
 
+from src.runtime_paths import get_app_root
+
 logger = logging.getLogger(__name__)
 
 def _format_mcp_connection_error(name: str, command: str = "", args: Optional[List[str]] = None, error: Exception = None) -> str:
@@ -508,7 +510,7 @@ class McpManager:
             return False
 
         script_rel, name = _BUILTIN_SERVERS[server_id]
-        base_dir = os.path.dirname(os.path.dirname(os.path.abspath(__file__)))
+        base_dir = get_app_root()
         script_path = os.path.join(base_dir, script_rel)
 
         # Clean up old connection

src/rag_singleton.py

@@ -7,6 +7,7 @@ import time
 from pathlib import Path
 
 from src.constants import RAG_DIR
+from src.runtime_paths import get_app_root
 
 logger = logging.getLogger(__name__)
 

src/runtime_paths.py

@@ -0,0 +1,30 @@
+"""Helpers for resolving runtime paths in source and frozen builds."""
+
+import os
+import sys
+
+
+def get_app_root() -> str:
+    """Return the app root directory.
+
+    In normal source runs, this is the repository root. In a frozen Windows
+    build, it is the bundle content root (PyInstaller's internal directory)
+    so bundled runtime folders like `static/`, `scripts/`, and `data/` stay
+    together with the executable payload.
+    """
+    if getattr(sys, "frozen", False):
+        return getattr(sys, "_MEIPASS", os.path.dirname(os.path.abspath(sys.executable)))
+    return os.path.dirname(os.path.dirname(os.path.abspath(__file__)))
+
+
+def get_default_data_dir() -> str:
+    """Return the default path to the data directory.
+
+    In normal runs, this is a 'data' subdirectory under the app root.
+    In frozen builds, it is a persistent user directory (~/.odysseus/data)
+    to prevent SQLite databases and other persistent files from being
+    written to the ephemeral, temporary extraction bundle directory.
+    """
+    if getattr(sys, "frozen", False):
+        return os.path.join(os.path.expanduser("~"), ".odysseus", "data")
+    return os.path.join(get_app_root(), "data")

src/tool_implementations.py

@@ -3797,7 +3797,7 @@ async def do_resolve_contact(content: str, owner: Optional[str] = None) -> Dict:
     if not name:
         return {"error": "name is required", "exit_code": 1}
 
-    contacts = {}  # email -> {name, source}
+    contacts = {}  # email_or_phone -> {name, source, phone?}
 
     # 1. CardDAV (Radicale) — structured contacts. Call in-process: a
     # server-side httpx GET to /api/contacts/search carries no session
@@ -3812,10 +3812,18 @@ async def do_resolve_contact(content: str, owner: Optional[str] = None) -> Dict:
             match = q in hay_name or any(q in (e or "").lower() for e in c.get("emails", []))
             if not match:
                 continue
+            has_email = False
             for email in (c.get("emails") or []):
                 email = (email or "").strip().lower()
                 if email and "@" in email:
                     contacts[email] = {"name": c.get("name") or email, "source": "contacts"}
+                    has_email = True
+            # Fall back to phone numbers when the contact has no email address
+            if not has_email:
+                for phone in (c.get("phones") or []):
+                    phone = (phone or "").strip()
+                    if phone:
+                        contacts[phone] = {"name": c.get("name") or phone, "source": "contacts", "phone": phone}
     except Exception:
         pass
 
@@ -3835,8 +3843,11 @@ async def do_resolve_contact(content: str, owner: Optional[str] = None) -> Dict:
         return {"output": f"No contacts found matching '{name}'.", "exit_code": 0}
 
     lines = [f"Contacts matching '{name}':"]
-    for email, info in contacts.items():
-        lines.append(f"- {info['name']} <{email}> ({info['source']})")
+    for key, info in contacts.items():
+        if info.get("phone"):
+            lines.append(f"- {info['name']} — phone: {info['phone']} ({info['source']})")
+        else:
+            lines.append(f"- {info['name']} <{key}> ({info['source']})")
     return {"output": "\n".join(lines), "exit_code": 0}
 
 

src/tool_schemas.py

@@ -68,11 +68,12 @@ FUNCTION_TOOL_SCHEMAS = [
         "type": "function",
         "function": {
             "name": "web_fetch",
-            "description": "Fetch and read the text content of a specific URL the user names (e.g. 'check example.com', 'what's on this page <url>'). Use when you already have a concrete URL/domain. NOT for open-ended searches (use web_search) or 'research X' jobs (use trigger_research).",
+            "description": "Fetch and read the text content of a specific URL the user names (e.g. 'check example.com', 'what's on this page <url>'). Use when you already have a concrete URL/domain. NOT for open-ended searches (use web_search) or 'research X' jobs (use trigger_research). Downloads are size-budgeted; a '[partial content: ...]' notice in the result means the body was cut short and you can re-call with full=true for the rest.",
             "parameters": {
                 "type": "object",
                 "properties": {
-                    "url": {"type": "string", "description": "The URL or domain to fetch (http/https; a bare domain like example.com is fine)"}
+                    "url": {"type": "string", "description": "The URL or domain to fetch (http/https; a bare domain like example.com is fine)"},
+                    "full": {"type": "boolean", "description": "Raise the download budget to the hard cap for large pages/files. Use only after a result reported partial content."}
                 },
                 "required": ["url"]
             }
@@ -1008,7 +1009,7 @@ FUNCTION_TOOL_SCHEMAS = [
         "type": "function",
         "function": {
             "name": "resolve_contact",
-            "description": "Look up a contact's email address by name. Searches CardDAV address book and sent email history. Use when the user says 'message [name]' or 'email [name]' without an email address.",
+            "description": "Look up a contact by name. Searches CardDAV address book and sent email history. Returns email addresses (when available) or phone numbers. Use when the user says 'message [name]', 'email [name]', or asks for someone's contact details.",
             "parameters": {
                 "type": "object",
                 "properties": {

tests/OVERSIZED_TEST_SPLIT_PLAN.md

@@ -0,0 +1,326 @@
+# Oversized Test File Split Plan
+
+## Purpose
+
+This document plans future oversized test-file splits using current repo data.
+It does not move files, rewrite assertions, extract helpers, or change CI.
+
+## Roadmap context
+
+- Issue: #3983
+- Parent tracker: #2523
+- Follows #3973 / #3982, the report-only order-sensitivity diagnostics slice.
+
+## Methodology
+
+Metrics were generated from the current test tree using:
+
+- physical line counts for every recursive `test_*.py` file under `tests/`;
+- AST counts for `test_*` functions and `Test*` classes;
+- one `pytest --collect-only -q tests` run to count collected items per file;
+- current taxonomy classification from `tests._taxonomy.classify_test_path`; and
+- static setup-signal scans for route/API, DB/session, import-state, security, filesystem, subprocess/script, async/threading, and UI/static indicators.
+
+Static signals are not proof of risk. They are review prompts.
+Future split PRs must still inspect each file manually before editing.
+
+## Current summary
+
+- test files scanned: 583
+- collected pytest items counted: 3586
+- large-file threshold: 300 lines
+- large-collected threshold: 20 collected items
+
+Area distribution:
+
+| Value | Files |
+|---|---:|
+| cli | 28 |
+| helpers | 1 |
+| js | 39 |
+| routes | 23 |
+| security | 77 |
+| services | 144 |
+| uncategorized | 234 |
+| unit | 37 |
+
+Sub-area distribution:
+
+| Value | Files |
+|---|---:|
+| api | 6 |
+| atomic | 3 |
+| auth | 9 |
+| calendar | 10 |
+| cli | 28 |
+| confinement | 7 |
+| cookbook | 13 |
+| document | 11 |
+| email | 12 |
+| embedding | 3 |
+| gallery | 5 |
+| history | 3 |
+| js | 39 |
+| llm | 16 |
+| mcp | 8 |
+| memory | 15 |
+| nondict | 7 |
+| nonstring | 22 |
+| owner | 14 |
+| owner_scope | 23 |
+| parse | 4 |
+| provider | 6 |
+| research | 16 |
+| route | 6 |
+| routes | 9 |
+| scheduler | 3 |
+| scope | 5 |
+| security | 9 |
+| session | 16 |
+| ssrf | 3 |
+| webhook | 3 |
+| xss | 5 |
+
+Values below 2 files: 244 values covering 244 files.
+
+## Top files by collected pytest items
+
+| File | Lines | Collected tests | Test defs | Test classes | Area | Sub-area | Signals |
+|---|---:|---:|---:|---:|---|---|---|
+| `tests/test_model_routes.py` | 1778 | 139 | 116 | 10 | routes | routes | route/api, db/session, import-state, async/threading |
+| `tests/test_security_regressions.py` | 1224 | 92 | 68 | 0 | security | security | route/api, db/session, import-state, security, filesystem, async/threading, ui/static |
+| `tests/test_provider_classification.py` | 188 | 67 | 21 | 4 | services | provider | - |
+| `tests/test_cookbook_helpers.py` | 912 | 65 | 65 | 0 | services | cookbook | route/api, filesystem, subprocess/script, async/threading, ui/static |
+| `tests/test_shell_routes.py` | 481 | 63 | 48 | 8 | routes | routes | route/api, import-state, filesystem |
+| `tests/test_pr_blocker_audit.py` | 964 | 58 | 58 | 0 | uncategorized | pr_blocker_audit | import-state, security, filesystem |
+| `tests/test_provider_endpoints.py` | 241 | 58 | 18 | 1 | services | provider | subprocess/script |
+| `tests/test_agent_loop.py` | 469 | 52 | 52 | 5 | uncategorized | agent_loop | db/session, import-state |
+| `tests/test_service_health.py` | 472 | 47 | 42 | 0 | uncategorized | service_health | async/threading |
+| `tests/test_run_focus.py` | 399 | 47 | 44 | 0 | uncategorized | run_focus | security, filesystem, subprocess/script, ui/static |
+| `tests/test_llm_core_temperature.py` | 196 | 41 | 17 | 0 | services | llm | - |
+| `tests/test_endpoint_probing.py` | 411 | 34 | 30 | 6 | uncategorized | endpoint_probing | route/api, db/session, import-state |
+| `tests/test_llm_core_anthropic_temp_omit.py` | 94 | 32 | 6 | 0 | services | llm | db/session |
+| `tests/test_chat_helpers.py` | 264 | 31 | 18 | 0 | uncategorized | chat_helpers | route/api |
+| `tests/test_provider_detection.py` | 148 | 31 | 31 | 5 | services | provider | - |
+| `tests/test_model_context.py` | 251 | 30 | 30 | 4 | uncategorized | model_context | db/session, import-state |
+| `tests/test_endpoint_resolver.py` | 148 | 30 | 30 | 6 | uncategorized | endpoint_resolver | - |
+| `tests/test_embedding_lanes.py` | 1104 | 29 | 29 | 0 | services | embedding | filesystem |
+| `tests/test_upload_limits_centralized.py` | 110 | 29 | 5 | 0 | uncategorized | upload_limits_centralized | import-state, filesystem |
+| `tests/test_email_oauth.py` | 580 | 28 | 25 | 0 | services | email | route/api, db/session, security, async/threading |
+| `tests/test_review_regressions.py` | 930 | 26 | 26 | 0 | uncategorized | review_regressions | route/api, db/session, import-state, filesystem, async/threading |
+| `tests/test_rename_user_owner_sync.py` | 686 | 26 | 26 | 0 | security | owner | route/api, db/session, import-state, filesystem, async/threading |
+| `tests/test_helpers_import_state.py` | 426 | 26 | 26 | 0 | helpers | helpers | route/api, db/session, import-state |
+| `tests/test_taxonomy.py` | 145 | 26 | 16 | 0 | uncategorized | taxonomy | security, ui/static |
+| `tests/test_tool_path_confinement.py` | 282 | 24 | 24 | 0 | security | confinement | import-state, filesystem, async/threading |
+| `tests/test_copilot.py` | 170 | 23 | 16 | 0 | uncategorized | copilot | - |
+| `tests/test_research_utils.py` | 97 | 23 | 23 | 2 | services | research | - |
+| `tests/test_api_chat_security.py` | 401 | 22 | 8 | 0 | security | security | route/api, db/session, import-state, filesystem, async/threading |
+| `tests/test_tool_support_heuristic.py` | 166 | 22 | 22 | 3 | uncategorized | tool_support_heuristic | - |
+| `tests/test_platform_compat.py` | 318 | 21 | 21 | 0 | uncategorized | platform_compat | import-state, filesystem, subprocess/script |
+
+## Top files by physical line count
+
+| File | Lines | Collected tests | Test defs | Test classes | Area | Sub-area | Signals |
+|---|---:|---:|---:|---:|---|---|---|
+| `tests/test_model_routes.py` | 1778 | 139 | 116 | 10 | routes | routes | route/api, db/session, import-state, async/threading |
+| `tests/test_security_regressions.py` | 1224 | 92 | 68 | 0 | security | security | route/api, db/session, import-state, security, filesystem, async/threading, ui/static |
+| `tests/test_embedding_lanes.py` | 1104 | 29 | 29 | 0 | services | embedding | filesystem |
+| `tests/test_pr_blocker_audit.py` | 964 | 58 | 58 | 0 | uncategorized | pr_blocker_audit | import-state, security, filesystem |
+| `tests/test_review_regressions.py` | 930 | 26 | 26 | 0 | uncategorized | review_regressions | route/api, db/session, import-state, filesystem, async/threading |
+| `tests/test_cookbook_helpers.py` | 912 | 65 | 65 | 0 | services | cookbook | route/api, filesystem, subprocess/script, async/threading, ui/static |
+| `tests/test_rename_user_owner_sync.py` | 686 | 26 | 26 | 0 | security | owner | route/api, db/session, import-state, filesystem, async/threading |
+| `tests/test_email_oauth.py` | 580 | 28 | 25 | 0 | services | email | route/api, db/session, security, async/threading |
+| `tests/test_api_token_routes.py` | 578 | 17 | 17 | 0 | routes | api_routes | route/api, db/session, import-state, async/threading |
+| `tests/test_shell_routes.py` | 481 | 63 | 48 | 8 | routes | routes | route/api, import-state, filesystem |
+| `tests/test_email_owner_scope.py` | 474 | 9 | 9 | 0 | security | owner_scope | route/api, db/session, filesystem, async/threading |
+| `tests/test_service_health.py` | 472 | 47 | 42 | 0 | uncategorized | service_health | async/threading |
+| `tests/test_agent_loop.py` | 469 | 52 | 52 | 5 | uncategorized | agent_loop | db/session, import-state |
+| `tests/test_kv_cache_invalidation_2927.py` | 463 | 8 | 8 | 0 | uncategorized | kv_cache_invalidation_2927 | route/api, db/session, import-state, async/threading |
+| `tests/test_helpers_import_state.py` | 426 | 26 | 26 | 0 | helpers | helpers | route/api, db/session, import-state |
+| `tests/test_endpoint_owner_scope_followup.py` | 414 | 11 | 11 | 0 | security | owner_scope | route/api, db/session, filesystem |
+| `tests/test_endpoint_probing.py` | 411 | 34 | 30 | 6 | uncategorized | endpoint_probing | route/api, db/session, import-state |
+| `tests/test_imap_leak_fixes.py` | 404 | 15 | 15 | 0 | uncategorized | imap_leak_fixes | route/api, db/session, security, filesystem |
+| `tests/test_companion_readonly.py` | 402 | 17 | 17 | 0 | uncategorized | companion_readonly | db/session, import-state |
+| `tests/test_api_chat_security.py` | 401 | 22 | 8 | 0 | security | security | route/api, db/session, import-state, filesystem, async/threading |
+| `tests/test_upload_handler_atomicity.py` | 401 | 9 | 9 | 0 | uncategorized | upload_handler_atomicity | filesystem, async/threading |
+| `tests/test_run_focus.py` | 399 | 47 | 44 | 0 | uncategorized | run_focus | security, filesystem, subprocess/script, ui/static |
+| `tests/test_auth_regressions.py` | 375 | 15 | 15 | 0 | security | auth | route/api, db/session, import-state, async/threading |
+| `tests/test_calendar_owner_scope.py` | 345 | 7 | 7 | 0 | security | owner_scope | route/api, db/session, import-state, filesystem, async/threading, ui/static |
+| `tests/test_null_owner_gates.py` | 342 | 20 | 20 | 0 | security | owner | route/api, db/session, import-state |
+| `tests/test_agent_migration_manifest.py` | 340 | 15 | 15 | 0 | uncategorized | agent_migration_manifest | import-state, filesystem |
+| `tests/test_calendar_recurrence.py` | 338 | 19 | 19 | 0 | services | calendar | - |
+| `tests/test_tool_policy.py` | 330 | 13 | 13 | 0 | uncategorized | tool_policy | import-state, async/threading |
+| `tests/test_workspace_confine.py` | 328 | 18 | 18 | 0 | uncategorized | workspace_confine | route/api, filesystem, subprocess/script, async/threading |
+| `tests/test_diffusion_server_security.py` | 325 | 14 | 14 | 0 | security | security | route/api, import-state, security, filesystem, async/threading, ui/static |
+
+## Split planning candidates
+
+This section is generated from metrics, not from manual judgement.
+Files are included when they meet at least one threshold:
+
+- at least 300 physical lines; or
+- at least 20 collected pytest items.
+
+These are planning candidates only. A later split PR still needs a focused manual review of each file before moving tests.
+
+| File | Why included | Setup/risk signals | Suggested handling |
+|---|---|---|---|
+| `tests/test_model_routes.py` | 1778 lines, 139 collected tests | route/api, db/session, import-state, async/threading | Defer mechanical split until setup/risk boundaries are mapped. |
+| `tests/test_security_regressions.py` | 1224 lines, 92 collected tests | route/api, db/session, import-state, security, filesystem, async/threading, ui/static | Defer mechanical split until setup/risk boundaries are mapped. |
+| `tests/test_provider_classification.py` | 67 collected tests | No obvious setup signals from static scan. | Good first manual-review candidate if test themes are cohesive. |
+| `tests/test_cookbook_helpers.py` | 912 lines, 65 collected tests | route/api, filesystem, subprocess/script, async/threading, ui/static | Defer mechanical split until setup/risk boundaries are mapped. |
+| `tests/test_shell_routes.py` | 481 lines, 63 collected tests | route/api, import-state, filesystem | Defer mechanical split until setup/risk boundaries are mapped. |
+| `tests/test_pr_blocker_audit.py` | 964 lines, 58 collected tests | import-state, security, filesystem | Defer mechanical split until setup/risk boundaries are mapped. |
+| `tests/test_provider_endpoints.py` | 58 collected tests | subprocess/script | Good first manual-review candidate if test themes are cohesive. |
+| `tests/test_agent_loop.py` | 469 lines, 52 collected tests | db/session, import-state | Defer mechanical split until setup/risk boundaries are mapped. |
+| `tests/test_service_health.py` | 472 lines, 47 collected tests | async/threading | Good first manual-review candidate if test themes are cohesive. |
+| `tests/test_run_focus.py` | 399 lines, 47 collected tests | security, filesystem, subprocess/script, ui/static | Defer mechanical split until setup/risk boundaries are mapped. |
+| `tests/test_llm_core_temperature.py` | 41 collected tests | No obvious setup signals from static scan. | Good first manual-review candidate if test themes are cohesive. |
+| `tests/test_endpoint_probing.py` | 411 lines, 34 collected tests | route/api, db/session, import-state | Defer mechanical split until setup/risk boundaries are mapped. |
+| `tests/test_llm_core_anthropic_temp_omit.py` | 32 collected tests | db/session | Defer mechanical split until setup/risk boundaries are mapped. |
+| `tests/test_chat_helpers.py` | 31 collected tests | route/api | Defer mechanical split until setup/risk boundaries are mapped. |
+| `tests/test_provider_detection.py` | 31 collected tests | No obvious setup signals from static scan. | Good first manual-review candidate if test themes are cohesive. |
+| `tests/test_model_context.py` | 30 collected tests | db/session, import-state | Defer mechanical split until setup/risk boundaries are mapped. |
+| `tests/test_endpoint_resolver.py` | 30 collected tests | No obvious setup signals from static scan. | Good first manual-review candidate if test themes are cohesive. |
+| `tests/test_embedding_lanes.py` | 1104 lines, 29 collected tests | filesystem | Good first manual-review candidate if test themes are cohesive. |
+| `tests/test_upload_limits_centralized.py` | 29 collected tests | import-state, filesystem | Defer mechanical split until setup/risk boundaries are mapped. |
+| `tests/test_email_oauth.py` | 580 lines, 28 collected tests | route/api, db/session, security, async/threading | Defer mechanical split until setup/risk boundaries are mapped. |
+| `tests/test_review_regressions.py` | 930 lines, 26 collected tests | route/api, db/session, import-state, filesystem, async/threading | Defer mechanical split until setup/risk boundaries are mapped. |
+| `tests/test_rename_user_owner_sync.py` | 686 lines, 26 collected tests | route/api, db/session, import-state, filesystem, async/threading | Defer mechanical split until setup/risk boundaries are mapped. |
+| `tests/test_helpers_import_state.py` | 426 lines, 26 collected tests | route/api, db/session, import-state | Defer mechanical split until setup/risk boundaries are mapped. |
+| `tests/test_taxonomy.py` | 26 collected tests | security, ui/static | Defer mechanical split until setup/risk boundaries are mapped. |
+| `tests/test_tool_path_confinement.py` | 24 collected tests | import-state, filesystem, async/threading | Defer mechanical split until setup/risk boundaries are mapped. |
+| `tests/test_copilot.py` | 23 collected tests | No obvious setup signals from static scan. | Good first manual-review candidate if test themes are cohesive. |
+| `tests/test_research_utils.py` | 23 collected tests | No obvious setup signals from static scan. | Good first manual-review candidate if test themes are cohesive. |
+| `tests/test_api_chat_security.py` | 401 lines, 22 collected tests | route/api, db/session, import-state, filesystem, async/threading | Defer mechanical split until setup/risk boundaries are mapped. |
+| `tests/test_tool_support_heuristic.py` | 22 collected tests | No obvious setup signals from static scan. | Good first manual-review candidate if test themes are cohesive. |
+| `tests/test_platform_compat.py` | 318 lines, 21 collected tests | import-state, filesystem, subprocess/script | Defer mechanical split until setup/risk boundaries are mapped. |
+| `tests/test_context_compactor.py` | 21 collected tests | db/session, import-state, async/threading | Defer mechanical split until setup/risk boundaries are mapped. |
+| `tests/test_prompt_security.py` | 21 collected tests | No obvious setup signals from static scan. | Good first manual-review candidate if test themes are cohesive. |
+| `tests/test_null_owner_gates.py` | 342 lines, 20 collected tests | route/api, db/session, import-state | Defer mechanical split until setup/risk boundaries are mapped. |
+| `tests/test_youtube_handler_consolidation.py` | 20 collected tests | route/api, import-state | Defer mechanical split until setup/risk boundaries are mapped. |
+| `tests/test_calendar_recurrence.py` | 338 lines | No obvious setup signals from static scan. | Plan split boundaries before editing. |
+| `tests/test_workspace_confine.py` | 328 lines | route/api, filesystem, subprocess/script, async/threading | Defer mechanical split until setup/risk boundaries are mapped. |
+| `tests/test_api_token_routes.py` | 578 lines | route/api, db/session, import-state, async/threading | Defer mechanical split until setup/risk boundaries are mapped. |
+| `tests/test_companion_readonly.py` | 402 lines | db/session, import-state | Defer mechanical split until setup/risk boundaries are mapped. |
+| `tests/test_set_admin.py` | 317 lines | route/api, import-state, filesystem, async/threading | Defer mechanical split until setup/risk boundaries are mapped. |
+| `tests/test_imap_leak_fixes.py` | 404 lines | route/api, db/session, security, filesystem | Defer mechanical split until setup/risk boundaries are mapped. |
+| `tests/test_auth_regressions.py` | 375 lines | route/api, db/session, import-state, async/threading | Defer mechanical split until setup/risk boundaries are mapped. |
+| `tests/test_agent_migration_manifest.py` | 340 lines | import-state, filesystem | Defer mechanical split until setup/risk boundaries are mapped. |
+| `tests/test_diffusion_server_security.py` | 325 lines | route/api, import-state, security, filesystem, async/threading, ui/static | Defer mechanical split until setup/risk boundaries are mapped. |
+| `tests/test_tool_policy.py` | 330 lines | import-state, async/threading | Defer mechanical split until setup/risk boundaries are mapped. |
+| `tests/test_endpoint_owner_scope_followup.py` | 414 lines | route/api, db/session, filesystem | Defer mechanical split until setup/risk boundaries are mapped. |
+| `tests/test_upload_routes_owner_scope.py` | 315 lines | route/api, filesystem, async/threading | Defer mechanical split until setup/risk boundaries are mapped. |
+| `tests/test_email_owner_scope.py` | 474 lines | route/api, db/session, filesystem, async/threading | Defer mechanical split until setup/risk boundaries are mapped. |
+| `tests/test_upload_handler_atomicity.py` | 401 lines | filesystem, async/threading | Plan split boundaries before editing. |
+| `tests/test_kv_cache_invalidation_2927.py` | 463 lines | route/api, db/session, import-state, async/threading | Defer mechanical split until setup/risk boundaries are mapped. |
+| `tests/test_calendar_owner_scope.py` | 345 lines | route/api, db/session, import-state, filesystem, async/threading, ui/static | Defer mechanical split until setup/risk boundaries are mapped. |
+| `tests/test_skills_manager_owner_isolation.py` | 306 lines | import-state, filesystem | Defer mechanical split until setup/risk boundaries are mapped. |
+
+## Taxonomy coverage gaps among split candidates
+
+`uncategorized` is a current taxonomy area, not a builder failure.
+This plan does not reclassify tests because taxonomy changes should be reviewed separately from oversized-file split planning.
+
+Before using any of these files as a split target, first decide whether the taxonomy should be refined in a separate focused issue/PR.
+
+| File | Lines | Collected tests | Sub-area | Signals | Suggested follow-up |
+|---|---:|---:|---|---|---|
+| `tests/test_pr_blocker_audit.py` | 964 | 58 | pr_blocker_audit | import-state, security, filesystem | Review taxonomy and setup/risk boundaries before any split. |
+| `tests/test_agent_loop.py` | 469 | 52 | agent_loop | db/session, import-state | Review taxonomy and setup/risk boundaries before any split. |
+| `tests/test_service_health.py` | 472 | 47 | service_health | async/threading | Review taxonomy mapping before using as a split target. |
+| `tests/test_run_focus.py` | 399 | 47 | run_focus | security, filesystem, subprocess/script, ui/static | Review taxonomy and setup/risk boundaries before any split. |
+| `tests/test_endpoint_probing.py` | 411 | 34 | endpoint_probing | route/api, db/session, import-state | Review taxonomy and setup/risk boundaries before any split. |
+| `tests/test_chat_helpers.py` | 264 | 31 | chat_helpers | route/api | Review taxonomy and setup/risk boundaries before any split. |
+| `tests/test_model_context.py` | 251 | 30 | model_context | db/session, import-state | Review taxonomy and setup/risk boundaries before any split. |
+| `tests/test_endpoint_resolver.py` | 148 | 30 | endpoint_resolver | - | Review taxonomy mapping before using as a split target. |
+| `tests/test_upload_limits_centralized.py` | 110 | 29 | upload_limits_centralized | import-state, filesystem | Review taxonomy and setup/risk boundaries before any split. |
+| `tests/test_review_regressions.py` | 930 | 26 | review_regressions | route/api, db/session, import-state, filesystem, async/threading | Review taxonomy and setup/risk boundaries before any split. |
+| `tests/test_taxonomy.py` | 145 | 26 | taxonomy | security, ui/static | Review taxonomy and setup/risk boundaries before any split. |
+| `tests/test_copilot.py` | 170 | 23 | copilot | - | Review taxonomy mapping before using as a split target. |
+| `tests/test_tool_support_heuristic.py` | 166 | 22 | tool_support_heuristic | - | Review taxonomy mapping before using as a split target. |
+| `tests/test_platform_compat.py` | 318 | 21 | platform_compat | import-state, filesystem, subprocess/script | Review taxonomy and setup/risk boundaries before any split. |
+| `tests/test_context_compactor.py` | 233 | 21 | context_compactor | db/session, import-state, async/threading | Review taxonomy and setup/risk boundaries before any split. |
+| `tests/test_youtube_handler_consolidation.py` | 104 | 20 | youtube_handler_consolidation | route/api, import-state | Review taxonomy and setup/risk boundaries before any split. |
+| `tests/test_workspace_confine.py` | 328 | 18 | workspace_confine | route/api, filesystem, subprocess/script, async/threading | Review taxonomy and setup/risk boundaries before any split. |
+| `tests/test_companion_readonly.py` | 402 | 17 | companion_readonly | db/session, import-state | Review taxonomy and setup/risk boundaries before any split. |
+| `tests/test_set_admin.py` | 317 | 17 | set_admin | route/api, import-state, filesystem, async/threading | Review taxonomy and setup/risk boundaries before any split. |
+| `tests/test_imap_leak_fixes.py` | 404 | 15 | imap_leak_fixes | route/api, db/session, security, filesystem | Review taxonomy and setup/risk boundaries before any split. |
+| `tests/test_agent_migration_manifest.py` | 340 | 15 | agent_migration_manifest | import-state, filesystem | Review taxonomy and setup/risk boundaries before any split. |
+| `tests/test_tool_policy.py` | 330 | 13 | tool_policy | import-state, async/threading | Review taxonomy and setup/risk boundaries before any split. |
+| `tests/test_upload_handler_atomicity.py` | 401 | 9 | upload_handler_atomicity | filesystem, async/threading | Review taxonomy mapping before using as a split target. |
+| `tests/test_kv_cache_invalidation_2927.py` | 463 | 8 | kv_cache_invalidation_2927 | route/api, db/session, import-state, async/threading | Review taxonomy and setup/risk boundaries before any split. |
+
+## Suggested first manual-review candidates
+
+These are not automatic split approvals. They are categorized candidates with enough size/collection value and no route/API, DB/session, import-state, or security signal from the static scan.
+
+Files still in the `uncategorized` taxonomy area are listed separately below so taxonomy review does not get mixed into the first split decision.
+
+| File | Lines | Collected tests | Area | Sub-area | Signals | Why this is a candidate |
+|---|---:|---:|---|---|---|---|
+| `tests/test_provider_classification.py` | 188 | 67 | services | provider | - | 67 collected tests |
+| `tests/test_provider_endpoints.py` | 241 | 58 | services | provider | subprocess/script | 58 collected tests |
+| `tests/test_llm_core_temperature.py` | 196 | 41 | services | llm | - | 41 collected tests |
+| `tests/test_provider_detection.py` | 148 | 31 | services | provider | - | 31 collected tests |
+| `tests/test_embedding_lanes.py` | 1104 | 29 | services | embedding | filesystem | 1104 lines, 29 collected tests |
+| `tests/test_research_utils.py` | 97 | 23 | services | research | - | 23 collected tests |
+| `tests/test_prompt_security.py` | 203 | 21 | security | security | - | 21 collected tests |
+| `tests/test_calendar_recurrence.py` | 338 | 19 | services | calendar | - | 338 lines |
+
+## High-risk candidates to defer first
+
+These files may still be split later, but not as the first implementation slice without a separate manual boundary review.
+
+| File | Lines | Collected tests | High-risk signals |
+|---|---:|---:|---|
+| `tests/test_model_routes.py` | 1778 | 139 | db/session, import-state, route/api |
+| `tests/test_security_regressions.py` | 1224 | 92 | db/session, import-state, route/api, security |
+| `tests/test_cookbook_helpers.py` | 912 | 65 | route/api |
+| `tests/test_shell_routes.py` | 481 | 63 | import-state, route/api |
+| `tests/test_pr_blocker_audit.py` | 964 | 58 | import-state, security |
+| `tests/test_agent_loop.py` | 469 | 52 | db/session, import-state |
+| `tests/test_run_focus.py` | 399 | 47 | security |
+| `tests/test_endpoint_probing.py` | 411 | 34 | db/session, import-state, route/api |
+| `tests/test_llm_core_anthropic_temp_omit.py` | 94 | 32 | db/session |
+| `tests/test_chat_helpers.py` | 264 | 31 | route/api |
+| `tests/test_model_context.py` | 251 | 30 | db/session, import-state |
+| `tests/test_upload_limits_centralized.py` | 110 | 29 | import-state |
+| `tests/test_email_oauth.py` | 580 | 28 | db/session, route/api, security |
+| `tests/test_review_regressions.py` | 930 | 26 | db/session, import-state, route/api |
+| `tests/test_rename_user_owner_sync.py` | 686 | 26 | db/session, import-state, route/api |
+
+## Rules for future split PRs
+
+- One file or one coherent file-family per PR.
+- No assertion rewrites mixed with file moves.
+- No helper extraction mixed with file moves.
+- No production code changes.
+- No CI workflow changes.
+- Preserve existing markers and taxonomy unless the split issue explicitly says otherwise.
+- Validate the original file's collected tests before and after the split.
+- Validate any neighboring taxonomy/focused-runner behavior if paths change.
+- Treat files with route/API, DB/session, import-state, or security signals as higher-risk until manually reviewed.
+
+## Suggested next step
+
+Use this plan to choose the first actual oversized-file split issue.
+The first split should prefer a file with high review value and low setup risk.
+Do not start a split PR from this planning issue alone if the file's boundaries are still ambiguous.
+
+## Reproduction command
+
+This document was generated with:
+
+```bash
+.venv/bin/python tests/tools/build_oversized_test_split_plan.py
+```
+
+## Freshness check
+
+After editing the builder or rebasing the branch, regenerate the plan and confirm no unexpected plan drift:
+
+```bash
+.venv/bin/python tests/tools/build_oversized_test_split_plan.py
+git diff --exit-code -- tests/OVERSIZED_TEST_SPLIT_PLAN.md
+```

tests/test_api_token_routes.py

@@ -502,3 +502,77 @@ def test_delete_token_owner_check_skipped_when_auth_disabled(monkeypatch, token_
     resp = delete_token(request=req, token_id="tok123")
     assert resp == {"status": "deleted"}
     fake_session.delete.assert_called_once_with(fake_token)
+
+
+# ---------------------------------------------------------------------------
+# 7. PATCH /api/tokens/{id} — non-object JSON bodies must not 500
+# ---------------------------------------------------------------------------
+
+
+def test_update_token_with_array_body_does_not_500(monkeypatch, token_routes_mod):
+    """PATCH body of [] must be normalised to {} and not raise."""
+    monkeypatch.setenv("AUTH_ENABLED", "true")
+    mod = token_routes_mod
+
+    token = SimpleNamespace(
+        id="tok123", name="original", owner="alice",
+        token_prefix="ody_orig", scopes="email:read", is_active=True,
+    )
+    fake_session = MagicMock()
+    fake_session.query.return_value.filter.return_value.first.return_value = token
+    monkeypatch.setattr(mod, "get_db_session", lambda: _db_ctx(fake_session))
+
+    invalidator = MagicMock()
+    req = _patch_request(invalidator, [])
+    update_token = _get_handler(mod, "PATCH", "/tokens/{token_id}")
+    resp = asyncio.run(update_token(request=req, token_id="tok123"))
+
+    # Name and scopes must be unchanged — payload was normalised to {}
+    assert token.name == "original"
+    assert token.scopes == "email:read"
+    assert resp["name"] == "original"
+
+
+def test_update_token_with_null_body_does_not_500(monkeypatch, token_routes_mod):
+    """PATCH body of null must be normalised to {} and not raise."""
+    monkeypatch.setenv("AUTH_ENABLED", "true")
+    mod = token_routes_mod
+
+    token = SimpleNamespace(
+        id="tok123", name="original", owner="alice",
+        token_prefix="ody_orig", scopes="chat", is_active=True,
+    )
+    fake_session = MagicMock()
+    fake_session.query.return_value.filter.return_value.first.return_value = token
+    monkeypatch.setattr(mod, "get_db_session", lambda: _db_ctx(fake_session))
+
+    invalidator = MagicMock()
+    req = _patch_request(invalidator, None)
+    update_token = _get_handler(mod, "PATCH", "/tokens/{token_id}")
+    resp = asyncio.run(update_token(request=req, token_id="tok123"))
+
+    assert token.name == "original"
+    assert token.scopes == "chat"
+
+
+def test_update_token_normal_object_still_works(monkeypatch, token_routes_mod):
+    """Normal dict payload continues to update fields as before."""
+    monkeypatch.setenv("AUTH_ENABLED", "true")
+    mod = token_routes_mod
+
+    token = SimpleNamespace(
+        id="tok123", name="original", owner="alice",
+        token_prefix="ody_orig", scopes="email:read", is_active=True,
+    )
+    fake_session = MagicMock()
+    fake_session.query.return_value.filter.return_value.first.return_value = token
+    monkeypatch.setattr(mod, "get_db_session", lambda: _db_ctx(fake_session))
+
+    invalidator = MagicMock()
+    req = _patch_request(invalidator, {"name": "updated"})
+    update_token = _get_handler(mod, "PATCH", "/tokens/{token_id}")
+    resp = asyncio.run(update_token(request=req, token_id="tok123"))
+
+    assert token.name == "updated"
+    assert resp["name"] == "updated"
+    invalidator.assert_called_once()

tests/test_chat_helpers.py

@@ -30,7 +30,7 @@ class _Session:
 
 
 def test_allowed_models_legacy_empty_list_remains_unrestricted(monkeypatch):
-    monkeypatch.setattr("routes.chat_helpers.get_current_user", lambda request: "alice")
+    monkeypatch.setattr("routes.chat_helpers.effective_user", lambda request: "alice")
 
     _enforce_chat_privileges(
         _Request({"allowed_models": [], "max_messages_per_day": 0}),
@@ -39,7 +39,7 @@ def test_allowed_models_legacy_empty_list_remains_unrestricted(monkeypatch):
 
 
 def test_allowed_models_explicit_empty_restricted_list_blocks_all_models(monkeypatch):
-    monkeypatch.setattr("routes.chat_helpers.get_current_user", lambda request: "alice")
+    monkeypatch.setattr("routes.chat_helpers.effective_user", lambda request: "alice")
 
     with pytest.raises(HTTPException) as exc:
         _enforce_chat_privileges(
@@ -56,7 +56,7 @@ def test_allowed_models_explicit_empty_restricted_list_blocks_all_models(monkeyp
 
 
 def test_allowed_models_nonempty_list_still_restricts_without_new_flag(monkeypatch):
-    monkeypatch.setattr("routes.chat_helpers.get_current_user", lambda request: "alice")
+    monkeypatch.setattr("routes.chat_helpers.effective_user", lambda request: "alice")
 
     _enforce_chat_privileges(
         _Request({"allowed_models": ["provider/model-a"], "max_messages_per_day": 0}),
@@ -70,7 +70,7 @@ def test_allowed_models_nonempty_list_still_restricts_without_new_flag(monkeypat
 
 
 def test_no_restriction_allows_any_model(monkeypatch):
-    monkeypatch.setattr("routes.chat_helpers.get_current_user", lambda request: "alice")
+    monkeypatch.setattr("routes.chat_helpers.effective_user", lambda request: "alice")
 
     privs = {"allowed_models": [], "block_all_models": False, "max_messages_per_day": 0}
     _enforce_chat_privileges(_Request(privs), _Session("provider/model-a"))
@@ -78,7 +78,7 @@ def test_no_restriction_allows_any_model(monkeypatch):
 
 
 def test_specific_allowlist_blocks_models_outside_it(monkeypatch):
-    monkeypatch.setattr("routes.chat_helpers.get_current_user", lambda request: "alice")
+    monkeypatch.setattr("routes.chat_helpers.effective_user", lambda request: "alice")
 
     privs = {
         "allowed_models": ["gpt-4"],
@@ -92,7 +92,7 @@ def test_specific_allowlist_blocks_models_outside_it(monkeypatch):
 
 
 def test_block_all_models_blocks_regardless_of_allowed_models_contents(monkeypatch):
-    monkeypatch.setattr("routes.chat_helpers.get_current_user", lambda request: "alice")
+    monkeypatch.setattr("routes.chat_helpers.effective_user", lambda request: "alice")
 
     # Even if allowed_models contains entries, block_all_models wins.
     privs = {
@@ -111,7 +111,7 @@ def test_block_all_models_blocks_regardless_of_allowed_models_contents(monkeypat
 def test_admin_user_is_never_blocked(monkeypatch):
     from core.auth import ADMIN_PRIVILEGES
 
-    monkeypatch.setattr("routes.chat_helpers.get_current_user", lambda request: "admin")
+    monkeypatch.setattr("routes.chat_helpers.effective_user", lambda request: "admin")
 
     class _AdminAuthManager:
         def get_privileges(self, username):

tests/test_companion_readonly.py

@@ -13,6 +13,9 @@ import json
 from types import SimpleNamespace
 from unittest.mock import MagicMock
 
+import pytest
+from fastapi import HTTPException
+
 sys.path.insert(0, os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
 
 # core.database instantiates SQLAlchemy declarative classes at import time, which
@@ -225,12 +228,34 @@ def test_models_route_scopes_api_token_to_token_owner(monkeypatch):
     endpoints = _call_models_route(
         monkeypatch,
         rows,
-        _request(api_token=True, api_token_owner="alice", current_user="api"),
+        _request(
+            api_token=True,
+            api_token_owner="alice",
+            api_token_scopes=["chat"],
+            current_user="api",
+        ),
     )
 
     assert _endpoint_names(endpoints) == ["alice-endpoint", "shared-endpoint"]
 
 
+def test_models_route_rejects_api_token_without_chat_scope(monkeypatch):
+    monkeypatch.setattr(companion_routes, "get_current_user", lambda request: "api")
+
+    with pytest.raises(HTTPException) as exc:
+        _models_route()(
+            _request(
+                api_token=True,
+                api_token_owner="alice",
+                api_token_scopes=["todos:read"],
+                current_user="api",
+            )
+        )
+
+    assert exc.value.status_code == 403
+    assert "chat scope" in exc.value.detail
+
+
 def test_models_route_unresolved_owner_returns_only_shared_rows(monkeypatch):
     rows = [
         _ep(1, "alice-endpoint", "alice"),
@@ -242,7 +267,12 @@ def test_models_route_unresolved_owner_returns_only_shared_rows(monkeypatch):
     endpoints = _call_models_route(
         monkeypatch,
         rows,
-        _request(api_token=True, api_token_owner=None, current_user="api"),
+        _request(
+            api_token=True,
+            api_token_owner=None,
+            api_token_scopes=["chat"],
+            current_user="api",
+        ),
     )
 
     assert _endpoint_names(endpoints) == ["shared-endpoint"]

tests/test_cookbook_helpers.py

@@ -786,6 +786,50 @@ def test_cached_model_scan_reports_plain_dir_gguf(tmp_path):
     assert ggufs[3]["quant"] == "BF16"
 
 
+def test_cached_model_scan_uses_ollama_api_before_cli_and_windows_opt_in():
+    script = _cached_model_scan_script()
+
+    assert "scan_ollama_api()\nscan_ollama()" in script
+    assert "if any(m.get('is_ollama') for m in models): return" in script
+    assert "os.name == 'nt'" in script
+    assert "ODYSSEUS_ALLOW_OLLAMA_CLI_SCAN" in script
+
+
+@pytest.mark.skipif(os.name != "nt", reason="Windows Ollama CLI startup guard")
+def test_cached_model_scan_does_not_launch_ollama_cli_on_windows(tmp_path):
+    """Official Ollama for Windows can auto-start the tray/server on `ollama list`.
+    The read-only cache scanner must not invoke that CLI unless explicitly opted in.
+    """
+    marker = tmp_path / "ollama-called.txt"
+    fake_ollama = tmp_path / "ollama.cmd"
+    fake_ollama.write_text(
+        "@echo off\r\n"
+        f'echo called>"{marker}"\r\n'
+        "echo NAME ID SIZE MODIFIED\r\n"
+        "echo local-model:latest abc 1 GB now\r\n",
+        encoding="utf-8",
+    )
+
+    empty_home = tmp_path / "home"
+    empty_home.mkdir()
+    scan_py = tmp_path / "scan_cache.py"
+    scan_py.write_text(_cached_model_scan_script(), encoding="utf-8")
+    env = dict(os.environ)
+    env["PATH"] = str(tmp_path) + os.pathsep + env.get("PATH", "")
+    env["HOME"] = str(empty_home)
+    env.pop("ODYSSEUS_ALLOW_OLLAMA_CLI_SCAN", None)
+    proc = subprocess.run(
+        [sys.executable, str(scan_py)],
+        check=True,
+        capture_output=True,
+        text=True,
+        env=env,
+    )
+
+    assert marker.exists() is False
+    assert all(m.get("backend") != "ollama" for m in json.loads(proc.stdout))
+
+
 def test_cached_model_scan_uses_huggingface_cache_env(tmp_path):
     """Docker recreates can leave the persisted HF cache outside HOME.
     The Serve scanner should honor the cache env path instead of only ~/.cache.

tests/test_kv_cache_invalidation_2927.py

@@ -79,7 +79,7 @@ def _build_context_harness(monkeypatch, chat_helpers, history):
     monkeypatch.setattr(chat_helpers, "extract_preset", fake_extract_preset)
     monkeypatch.setattr(chat_helpers, "add_user_message", fake_add_user_message)
     monkeypatch.setattr(chat_helpers, "load_prefs_for_user", lambda user: {})
-    monkeypatch.setattr(chat_helpers, "get_current_user", lambda request: "tester")
+    monkeypatch.setattr(chat_helpers, "effective_user", lambda request: "tester")
     monkeypatch.setattr(chat_helpers, "normalize_model_id", lambda endpoint_url, model, **kwargs: None)
     monkeypatch.setattr(chat_helpers, "maybe_compact", fake_maybe_compact)
     monkeypatch.setattr(chat_helpers, "trim_for_context", lambda messages, context_length: messages)

tests/test_review_regressions.py

@@ -385,7 +385,7 @@ async def test_build_chat_context_incognito_does_not_duplicate_current_user_mess
     monkeypatch.setattr(chat_helpers, "extract_preset", fake_extract_preset)
     monkeypatch.setattr(chat_helpers, "add_user_message", fake_add_user_message)
     monkeypatch.setattr(chat_helpers, "load_prefs_for_user", lambda user: {})
-    monkeypatch.setattr(chat_helpers, "get_current_user", lambda request: "tester")
+    monkeypatch.setattr(chat_helpers, "effective_user", lambda request: "tester")
     monkeypatch.setattr(chat_helpers, "normalize_model_id", lambda endpoint_url, model, **kwargs: None)
     monkeypatch.setattr(chat_helpers, "maybe_compact", fake_maybe_compact)
     monkeypatch.setattr(chat_helpers, "trim_for_context", lambda messages, context_length: messages)

tests/test_runtime_paths.py

@@ -0,0 +1,50 @@
+import os
+import sys
+from unittest import mock
+import pytest
+from src.runtime_paths import get_app_root, get_default_data_dir
+
+
+def test_get_app_root_normal_run():
+    """Verify that get_app_root returns the repository root parent of src/ when not frozen."""
+    with mock.patch.object(sys, "frozen", False, create=True):
+        app_root = get_app_root()
+        # Verify it is a valid directory path and matches expected parent structure
+        assert os.path.isdir(app_root)
+        assert os.path.exists(os.path.join(app_root, "src"))
+
+
+def test_get_app_root_frozen_with_meipass():
+    """Verify that get_app_root returns the sys._MEIPASS directory when frozen by PyInstaller."""
+    mock_meipass = os.path.abspath("mock_meipass_dir")
+    with mock.patch.object(sys, "frozen", True, create=True), \
+         mock.patch.object(sys, "_MEIPASS", mock_meipass, create=True):
+        app_root = get_app_root()
+        assert app_root == mock_meipass
+
+
+def test_get_app_root_frozen_without_meipass():
+    """Verify that get_app_root falls back to the sys.executable parent directory when frozen but _MEIPASS is absent."""
+    mock_exe_path = os.path.join(os.path.abspath("mock_exe_dir"), "Odysseus.exe")
+    with mock.patch.object(sys, "frozen", True, create=True), \
+         mock.patch.object(sys, "executable", mock_exe_path, create=True):
+        # Remove sys._MEIPASS if it exists in the test process environment
+        if hasattr(sys, "_MEIPASS"):
+            delattr(sys, "_MEIPASS")
+        app_root = get_app_root()
+        assert app_root == os.path.abspath("mock_exe_dir")
+
+
+def test_get_default_data_dir_normal():
+    """Verify that get_default_data_dir resolves to get_app_root() / 'data' when not frozen."""
+    with mock.patch.object(sys, "frozen", False, create=True):
+        res = get_default_data_dir()
+        assert res == os.path.join(get_app_root(), "data")
+
+
+def test_get_default_data_dir_frozen():
+    """Verify that get_default_data_dir resolves to a persistent user path under ~ when frozen."""
+    with mock.patch.object(sys, "frozen", True, create=True):
+        res = get_default_data_dir()
+        expected = os.path.join(os.path.expanduser("~"), ".odysseus", "data")
+        assert res == expected

tests/test_search_content_extraction_parity.py

@@ -58,7 +58,7 @@ def test_content_fetcher_extracts_og_image_and_body_fallback(module, tmp_path, m
 
     monkeypatch.setattr(module, "CONTENT_CACHE_DIR", tmp_path)
     module.content_cache_index.clear()
-    monkeypatch.setattr(module, "_get_public_url", lambda url, headers, timeout: _FakeResponse(html))
+    monkeypatch.setattr(module, "_get_public_url", lambda url, headers, timeout, **kwargs: _FakeResponse(html))
 
     result = module.fetch_webpage_content("https://example.com/parity-test")
 
@@ -82,7 +82,7 @@ def test_fetch_webpage_content_returns_empty_result_on_http_status_error(status_
     monkeypatch.setattr(
         service_content,
         "_get_public_url",
-        lambda url, headers, timeout: _FakeErrorResponse(status_code),
+        lambda url, headers, timeout, **kwargs: _FakeErrorResponse(status_code),
     )
 
     result = service_content.fetch_webpage_content(f"https://example.com/status-{status_code}")
@@ -119,7 +119,7 @@ def test_fetch_webpage_content_429_takes_distinct_rate_limit_path(tmp_path, monk
     monkeypatch.setattr(
         service_content,
         "_get_public_url",
-        lambda url, headers, timeout: _FakeRateLimitResponse(),
+        lambda url, headers, timeout, **kwargs: _FakeRateLimitResponse(),
     )
 
     result = service_content.fetch_webpage_content("https://example.com/rate-limited")

tests/test_security_regressions.py

@@ -904,7 +904,13 @@ def test_web_fetch_guard_blocks_redirect_into_private(monkeypatch):
         url = "http://public.example/start"
         headers = {"location": "http://169.254.169.254/latest/meta-data/"}
 
-    monkeypatch.setattr(httpx, "get", lambda url, **kwargs: _Resp())
+    from contextlib import contextmanager
+
+    @contextmanager
+    def _fake_stream(method, url, **kwargs):
+        yield _Resp()
+
+    monkeypatch.setattr(httpx, "stream", _fake_stream)
 
     with _pytest.raises(httpx.RequestError) as exc:
         content._get_public_url("http://public.example/start", headers={}, timeout=5)

tests/test_session_endpoint_owner_scope.py

@@ -52,6 +52,6 @@ def test_chat_endpoint_recovery_paths_are_owner_scoped():
     assert "def _clear_orphaned_session_endpoint(sess, owner:" in chat_routes
     assert "def _recover_empty_session_model(sess, session_id: str, owner:" in chat_routes
     assert "q = owner_filter(q, ModelEndpoint, owner)" in chat_routes
-    assert "resolve_session_auth(sess, session, owner=get_current_user(request))" in chat_routes
+    assert "resolve_session_auth(sess, session, owner=effective_user(request))" in chat_routes
     assert "def resolve_session_auth(sess, session_id: str, owner:" in chat_helpers
     assert "update_q = update_q.filter(DBSession.owner == owner)" in chat_helpers

tests/test_web_fetch_plaintext.py

@@ -35,7 +35,7 @@ def _patch_fetch(monkeypatch, text, content_type):
     monkeypatch.setattr(
         content_mod,
         "_get_public_url",
-        lambda url, headers=None, timeout=5: _FakeResponse(text, content_type),
+        lambda url, headers=None, timeout=5, **kwargs: _FakeResponse(text, content_type),
     )
 
 

tests/test_web_fetch_size_caps.py

@@ -0,0 +1,206 @@
+"""web_fetch download budgets (#3812).
+
+MAX_OUTPUT_CHARS only trims what the agent sees; these caps bound what the
+server downloads, parses, and caches. Soft cap by default with a truncation
+notice, per-call override clamped to the hard cap, and a pre-buffer refusal
+when Content-Length already exceeds the hard ceiling.
+"""
+import json
+from contextlib import contextmanager
+
+import pytest
+
+from src.constants import WEB_FETCH_SOFT_MAX_BYTES, WEB_FETCH_HARD_MAX_BYTES
+from services.search import content as content_mod
+
+
+class _FakeStream:
+    """Stands in for the httpx.stream(...) context manager."""
+
+    def __init__(self, body: bytes, content_type="text/plain", content_length=None,
+                 status_code=200, chunk=8192):
+        self._body = body
+        self._chunk = chunk
+        self.status_code = status_code
+        self.encoding = "utf-8"
+        self.url = "https://example.com/x"
+        self.headers = {"Content-Type": content_type}
+        if content_length is not None:
+            self.headers["content-length"] = str(content_length)
+        self.body_reads = 0
+
+    def iter_bytes(self):
+        for i in range(0, len(self._body), self._chunk):
+            self.body_reads += 1
+            yield self._body[i:i + self._chunk]
+
+
+@pytest.fixture
+def no_cache(monkeypatch, tmp_path):
+    monkeypatch.setattr(content_mod, "CONTENT_CACHE_DIR", tmp_path)
+    monkeypatch.setattr(content_mod, "_cache_result", lambda *a, **k: None)
+    monkeypatch.setattr(content_mod, "_public_http_url", lambda u: True)
+
+
+def _patch_stream(monkeypatch, fake):
+    @contextmanager
+    def fake_stream(method, url, **kwargs):
+        yield fake
+    monkeypatch.setattr(content_mod.httpx, "stream", fake_stream)
+    return fake
+
+
+def test_body_under_cap_is_untouched(monkeypatch, no_cache):
+    _patch_stream(monkeypatch, _FakeStream(b"hello world"))
+    r = content_mod.fetch_webpage_content("https://example.com/a.txt")
+    assert r["success"] is True
+    assert r["content"] == "hello world"
+    assert r["truncated"] is False
+    assert r["fetched_bytes"] == len(b"hello world")
+
+
+def test_body_over_soft_cap_truncates_with_flags(monkeypatch, no_cache):
+    body = b"x" * (WEB_FETCH_SOFT_MAX_BYTES + 50_000)
+    _patch_stream(monkeypatch, _FakeStream(body, content_length=len(body)))
+    r = content_mod.fetch_webpage_content("https://example.com/big.txt")
+    assert r["truncated"] is True
+    assert r["fetched_bytes"] == WEB_FETCH_SOFT_MAX_BYTES
+    assert r["total_bytes"] == len(body)
+    assert len(r["content"]) == WEB_FETCH_SOFT_MAX_BYTES
+
+
+def test_max_bytes_override_raises_budget(monkeypatch, no_cache):
+    body = b"y" * (WEB_FETCH_SOFT_MAX_BYTES + 50_000)
+    _patch_stream(monkeypatch, _FakeStream(body))
+    r = content_mod.fetch_webpage_content(
+        "https://example.com/big.txt", max_bytes=len(body) + 1
+    )
+    assert r["truncated"] is False
+    assert r["fetched_bytes"] == len(body)
+
+
+def test_override_is_clamped_to_hard_cap(monkeypatch, no_cache):
+    # Ask for more than the ceiling; the effective budget must be the ceiling.
+    fake = _patch_stream(monkeypatch, _FakeStream(b"z" * 10, chunk=4))
+    r = content_mod.fetch_webpage_content(
+        "https://example.com/a.txt", max_bytes=WEB_FETCH_HARD_MAX_BYTES * 10
+    )
+    assert r["success"] is True
+    # The clamp itself: effective cap recorded in the cache key path is the
+    # hard cap, and a declared body over the ceiling is refused regardless.
+    big = _FakeStream(b"", content_length=WEB_FETCH_HARD_MAX_BYTES + 1)
+    _patch_stream(monkeypatch, big)
+    r = content_mod.fetch_webpage_content(
+        "https://example.com/huge.bin", max_bytes=WEB_FETCH_HARD_MAX_BYTES * 10
+    )
+    assert r["success"] is False
+    assert "TooLarge" in r["error"]
+    assert big.body_reads == 0  # refused before buffering
+
+
+def test_declared_over_hard_cap_refused_before_buffering(monkeypatch, no_cache):
+    fake = _FakeStream(b"irrelevant", content_length=WEB_FETCH_HARD_MAX_BYTES + 1)
+    _patch_stream(monkeypatch, fake)
+    r = content_mod.fetch_webpage_content("https://example.com/huge.iso")
+    assert r["success"] is False
+    assert "TooLarge" in r["error"]
+    assert fake.body_reads == 0
+
+
+def test_truncated_pdf_is_an_error_not_garbage(monkeypatch, no_cache):
+    body = b"%PDF-1.4 " + b"p" * (WEB_FETCH_SOFT_MAX_BYTES + 10)
+    _patch_stream(monkeypatch, _FakeStream(body, content_type="application/pdf"))
+    r = content_mod.fetch_webpage_content("https://example.com/big.pdf")
+    assert r["success"] is False
+    assert "TooLarge" in r["error"]
+
+
+def test_fetch_requests_identity_encoding(monkeypatch, no_cache):
+    # Compressed responses can decode to far more than Content-Length, so the
+    # streamed cap and the hard-cap preflight are only honest when we refuse
+    # transfer compression. Pin that the fetch advertises identity, not gzip.
+    seen = {}
+
+    @contextmanager
+    def fake_stream(method, url, **kwargs):
+        seen["headers"] = kwargs.get("headers") or {}
+        yield _FakeStream(b"hello")
+    monkeypatch.setattr(content_mod.httpx, "stream", fake_stream)
+
+    content_mod.fetch_webpage_content("https://example.com/a.txt")
+    assert seen["headers"].get("Accept-Encoding") == "identity"
+
+
+def test_rejects_compressed_response_that_ignored_identity(monkeypatch, no_cache):
+    # We request Accept-Encoding: identity, but a server can ignore it and send
+    # gzip anyway. httpx would decode it, so a tiny compressed body could balloon
+    # past the cap in one decoded chunk. Refuse before reading the body.
+    fake = _FakeStream(b"x" * 5000, content_length=40)
+    fake.headers["content-encoding"] = "gzip"
+    _patch_stream(monkeypatch, fake)
+    r = content_mod.fetch_webpage_content("https://example.com/a.txt")
+    assert r["success"] is False
+    assert "Content-Encoding" in r["error"] or "compressed" in r["error"]
+    assert fake.body_reads == 0  # refused before decoding any body
+
+
+def test_oversized_title_does_not_hide_partial_notice(monkeypatch):
+    # The partial-content notice is the PR's core contract; an untrusted,
+    # oversized page title must not push it past MAX_OUTPUT_CHARS.
+    import asyncio
+    from src.agent_tools.web_tools import WebFetchTool
+    from src.constants import MAX_OUTPUT_CHARS
+
+    def fake_fetch(url, timeout=10, max_bytes=None):
+        return {
+            "content": "partial body",
+            "title": "T" * (MAX_OUTPUT_CHARS + 5_000),
+            "error": "",
+            "truncated": True,
+            "fetched_bytes": WEB_FETCH_SOFT_MAX_BYTES,
+            "total_bytes": 9_000_000,
+        }
+
+    import src.search.content as alias_mod
+    monkeypatch.setattr(alias_mod, "fetch_webpage_content", fake_fetch)
+
+    out = asyncio.run(WebFetchTool().execute(
+        json.dumps({"url": "https://example.com/big.txt"}), ctx={}
+    ))
+    assert out["exit_code"] == 0
+    assert out["output"].startswith("[partial content:")
+    assert '"full": true' in out["output"]
+
+
+def test_tool_layer_emits_partial_notice_and_parses_full(monkeypatch):
+    import asyncio
+    from src.agent_tools.web_tools import WebFetchTool
+
+    calls = {}
+
+    def fake_fetch(url, timeout=10, max_bytes=None):
+        calls["max_bytes"] = max_bytes
+        return {
+            "content": "partial body",
+            "title": "Big File",
+            "error": "",
+            "truncated": True,
+            "fetched_bytes": WEB_FETCH_SOFT_MAX_BYTES,
+            "total_bytes": 5_000_000,
+        }
+
+    import src.search.content as alias_mod
+    monkeypatch.setattr(alias_mod, "fetch_webpage_content", fake_fetch)
+
+    out = asyncio.run(WebFetchTool().execute(
+        json.dumps({"url": "https://example.com/big.txt"}), ctx={}
+    ))
+    assert out["exit_code"] == 0
+    assert "[partial content:" in out["output"]
+    assert '"full": true' in out["output"]
+    assert calls["max_bytes"] is None
+
+    asyncio.run(WebFetchTool().execute(
+        json.dumps({"url": "https://example.com/big.txt", "full": True}), ctx={}
+    ))
+    assert calls["max_bytes"] == WEB_FETCH_HARD_MAX_BYTES

tests/test_webhook_emitters_use_manager.py

@@ -0,0 +1,54 @@
+"""Guard: every public webhook emitter goes through the manager.
+
+Public emitters in `routes/` must schedule their fire through
+`webhook_manager.fire_and_forget(...)` (or `_spawn_tracked`). A bare
+`asyncio.create_task(webhook_manager.fire(...))` escapes
+`WebhookManager._bg_tasks`, so asyncio only holds a weak reference to the
+delivery task and the GC can collect it before it sends — silently dropping
+the webhook. Catching this with a scan stops a regression from sneaking
+back in via a copy-paste.
+"""
+import ast
+from pathlib import Path
+
+ROUTES_DIR = Path(__file__).resolve().parent.parent / "routes"
+
+
+def _untracked_fire_calls(tree: ast.AST) -> list[tuple[int, str]]:
+    """Return (lineno, snippet) for any asyncio.create_task(webhook_manager.fire(...))."""
+    hits: list[tuple[int, str]] = []
+    for node in ast.walk(tree):
+        if not isinstance(node, ast.Call):
+            continue
+        func = node.func
+        if not (isinstance(func, ast.Attribute) and func.attr == "create_task"):
+            continue
+        if not (isinstance(func.value, ast.Name) and func.value.id == "asyncio"):
+            continue
+        if not node.args:
+            continue
+        inner = node.args[0]
+        if not isinstance(inner, ast.Call):
+            continue
+        inner_func = inner.func
+        if (
+            isinstance(inner_func, ast.Attribute)
+            and inner_func.attr == "fire"
+            and isinstance(inner_func.value, ast.Name)
+            and inner_func.value.id == "webhook_manager"
+        ):
+            hits.append((node.lineno, ast.unparse(node)))
+    return hits
+
+
+def test_no_untracked_webhook_fire_in_routes():
+    offenders: list[str] = []
+    for path in ROUTES_DIR.rglob("*.py"):
+        tree = ast.parse(path.read_text(), filename=str(path))
+        for lineno, snippet in _untracked_fire_calls(tree):
+            offenders.append(f"{path.relative_to(ROUTES_DIR.parent)}:{lineno}: {snippet}")
+    assert not offenders, (
+        "Public webhook emitters must use webhook_manager.fire_and_forget(...) "
+        "so the delivery task is tracked in WebhookManager._bg_tasks. Found "
+        "untracked emitter(s):\n  " + "\n  ".join(offenders)
+    )

tests/tools/build_oversized_test_split_plan.py

@@ -0,0 +1,574 @@
+#!/usr/bin/env python3
+"""Build the oversized test-file split plan for issue #3983.
+
+The output is a planning document only. It does not move tests, rewrite
+assertions, extract helpers, or change CI.
+"""
+from __future__ import annotations
+
+import ast
+import json
+import os
+import re
+import subprocess
+import sys
+from collections import Counter
+from dataclasses import dataclass
+from pathlib import Path
+
+ROOT = Path(__file__).resolve().parents[2]
+TESTS_DIR = ROOT / "tests"
+OUTPUT = TESTS_DIR / "OVERSIZED_TEST_SPLIT_PLAN.md"
+RAW_OUTPUT = Path("/tmp/oversized-test-file-metrics.json")
+
+LARGE_LINE_THRESHOLD = 300
+LARGE_NODE_THRESHOLD = 20
+TOP_LIMIT = 30
+
+HIGH_RISK_SIGNALS = {"route/api", "db/session", "import-state", "security"}
+
+
+@dataclass(frozen=True)
+class FileMetric:
+    path: str
+    lines: int
+    nonblank: int
+    test_defs: int
+    test_classes: int
+    collected: int
+    area: str
+    sub_area: str
+    signals: tuple[str, ...]
+
+
+def read_text(path: Path) -> str:
+    return path.read_text(encoding="utf-8", errors="replace")
+
+
+def count_ast_tests(text: str) -> tuple[int, int]:
+    tree = ast.parse(text)
+    test_defs = 0
+    test_classes = 0
+
+    for node in ast.walk(tree):
+        if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):
+            if node.name.startswith("test_"):
+                test_defs += 1
+        elif isinstance(node, ast.ClassDef):
+            if node.name.startswith("Test"):
+                test_classes += 1
+
+    return test_defs, test_classes
+
+
+def load_taxonomy_classifier():
+    sys.path.insert(0, str(ROOT))
+    from tests._taxonomy import classify_test_path
+
+    return classify_test_path
+
+
+def classify(path: Path, classify_test_path) -> tuple[str, str]:
+    rel_path = Path(path.relative_to(ROOT).as_posix())
+
+    try:
+        result = classify_test_path(rel_path)
+    except Exception:
+        return "unknown", "unknown"
+
+    return getattr(result, "area", "unknown"), getattr(result, "sub_area", "unknown")
+
+
+def collect_node_counts() -> Counter[str]:
+    cmd = [
+        sys.executable,
+        "-m",
+        "pytest",
+        "--collect-only",
+        "-q",
+        "tests",
+    ]
+    env = dict(os.environ)
+    env["PY_COLORS"] = "0"
+
+    result = subprocess.run(
+        cmd,
+        cwd=ROOT,
+        env=env,
+        text=True,
+        capture_output=True,
+    )
+
+    if result.returncode != 0:
+        print(result.stdout)
+        print(result.stderr, file=sys.stderr)
+        raise SystemExit(result.returncode)
+
+    counts: Counter[str] = Counter()
+    for line in result.stdout.splitlines():
+        line = line.strip()
+        if "::" not in line:
+            continue
+        if not line.startswith("tests/"):
+            continue
+        file_path = line.split("::", 1)[0]
+        counts[file_path] += 1
+
+    return counts
+
+
+def detect_signals(text: str, path: str) -> tuple[str, ...]:
+    signal_patterns = {
+        "route/api": [
+            r"\bTestClient\b",
+            r"\bapp\.",
+            r"\broutes\.",
+            r"\bfrom routes\b",
+            r"\bimport routes\b",
+        ],
+        "db/session": [
+            r"\bSessionLocal\b",
+            r"\bsqlite\b",
+            r"\bDATABASE_URL\b",
+            r"\bcore\.database\b",
+            r"\bdb\.query\b",
+            r"\bcommit\(",
+        ],
+        "import-state": [
+            r"\bsys\.modules\b",
+            r"\bimportlib\b",
+            r"\bclear_module\b",
+            r"\bpreserve_import_state\b",
+            r"\bmonkeypatch\.setitem\b",
+        ],
+        "security": [
+            r"\bsecurity\b",
+            r"\bssrf\b",
+            r"\bpath traversal\b",
+            r"\bcsrf\b",
+            r"\bpermission\b",
+        ],
+        "filesystem": [
+            r"\btmp_path\b",
+            r"\bTemporaryDirectory\b",
+            r"\bPath\(",
+            r"\bmkdir\b",
+            r"\bwrite_text\b",
+            r"\bread_text\b",
+        ],
+        "subprocess/script": [
+            r"\bsubprocess\b",
+            r"\brunpy\b",
+            r"\bload_script\b",
+            r"\bsys\.argv\b",
+        ],
+        "async/threading": [
+            r"\basyncio\b",
+            r"\bthreading\b",
+            r"\bconcurrent\.futures\b",
+            r"\bThreadPoolExecutor\b",
+        ],
+        "ui/static": [
+            r"\bstatic/",
+            r"\bjsdom\b",
+            r"\bnode\b",
+            r"\.js\b",
+        ],
+    }
+
+    signals = []
+    for name, patterns in signal_patterns.items():
+        if any(re.search(pattern, text, flags=re.IGNORECASE) for pattern in patterns):
+            signals.append(name)
+
+    if path.startswith("tests/cli/"):
+        signals.append("cli-directory")
+
+    return tuple(signals)
+
+
+def metric_for(path: Path, node_counts: Counter[str], classify_test_path) -> FileMetric:
+    rel = path.relative_to(ROOT).as_posix()
+    text = read_text(path)
+    lines = len(text.splitlines())
+    nonblank = sum(1 for line in text.splitlines() if line.strip())
+    test_defs, test_classes = count_ast_tests(text)
+    area, sub_area = classify(path, classify_test_path)
+
+    return FileMetric(
+        path=rel,
+        lines=lines,
+        nonblank=nonblank,
+        test_defs=test_defs,
+        test_classes=test_classes,
+        collected=node_counts.get(rel, 0),
+        area=area,
+        sub_area=sub_area,
+        signals=detect_signals(text, rel),
+    )
+
+
+def test_files() -> list[Path]:
+    return sorted(TESTS_DIR.rglob("test_*.py"))
+
+
+def as_metric_row(metric: FileMetric) -> str:
+    signals = ", ".join(metric.signals) if metric.signals else "-"
+    return (
+        f"| `{metric.path}` | {metric.lines} | {metric.collected} | "
+        f"{metric.test_defs} | {metric.test_classes} | "
+        f"{metric.area} | {metric.sub_area} | {signals} |"
+    )
+
+
+def metric_table(title: str, metrics: list[FileMetric]) -> list[str]:
+    lines = [
+        f"## {title}",
+        "",
+        "| File | Lines | Collected tests | Test defs | Test classes | Area | Sub-area | Signals |",
+        "|---|---:|---:|---:|---:|---|---|---|",
+    ]
+    lines.extend(as_metric_row(metric) for metric in metrics)
+    lines.append("")
+    return lines
+
+
+def candidate_metrics(metrics: list[FileMetric]) -> list[FileMetric]:
+    return [
+        metric
+        for metric in metrics
+        if metric.lines >= LARGE_LINE_THRESHOLD
+        or metric.collected >= LARGE_NODE_THRESHOLD
+    ]
+
+
+def include_reasons(metric: FileMetric) -> str:
+    reasons = []
+    if metric.lines >= LARGE_LINE_THRESHOLD:
+        reasons.append(f"{metric.lines} lines")
+    if metric.collected >= LARGE_NODE_THRESHOLD:
+        reasons.append(f"{metric.collected} collected tests")
+    return ", ".join(reasons)
+
+
+def risk_notes(metric: FileMetric) -> str:
+    if not metric.signals:
+        return "No obvious setup signals from static scan."
+    return ", ".join(metric.signals)
+
+
+def suggested_handling(metric: FileMetric) -> str:
+    if HIGH_RISK_SIGNALS.intersection(metric.signals):
+        return "Defer mechanical split until setup/risk boundaries are mapped."
+    if metric.collected >= LARGE_NODE_THRESHOLD:
+        return "Good first manual-review candidate if test themes are cohesive."
+    return "Plan split boundaries before editing."
+
+
+def candidate_section(metrics: list[FileMetric]) -> list[str]:
+    lines = [
+        "## Split planning candidates",
+        "",
+        "This section is generated from metrics, not from manual judgement.",
+        "Files are included when they meet at least one threshold:",
+        "",
+        f"- at least {LARGE_LINE_THRESHOLD} physical lines; or",
+        f"- at least {LARGE_NODE_THRESHOLD} collected pytest items.",
+        "",
+        "These are planning candidates only. A later split PR still needs a focused manual review of each file before moving tests.",
+        "",
+        "| File | Why included | Setup/risk signals | Suggested handling |",
+        "|---|---|---|---|",
+    ]
+
+    for metric in metrics:
+        lines.append(
+            f"| `{metric.path}` | {include_reasons(metric)} | "
+            f"{risk_notes(metric)} | {suggested_handling(metric)} |"
+        )
+
+    lines.append("")
+    return lines
+
+
+def first_manual_review_section(metrics: list[FileMetric]) -> list[str]:
+    low_risk = [
+        metric
+        for metric in metrics
+        if metric.area != "uncategorized"
+        and not HIGH_RISK_SIGNALS.intersection(metric.signals)
+    ]
+    low_risk = sorted(low_risk, key=lambda m: (m.collected, m.lines), reverse=True)
+
+    lines = [
+        "## Suggested first manual-review candidates",
+        "",
+        "These are not automatic split approvals. They are categorized candidates with enough size/collection value and no route/API, DB/session, import-state, or security signal from the static scan.",
+        "",
+        "Files still in the `uncategorized` taxonomy area are listed separately below so taxonomy review does not get mixed into the first split decision.",
+        "",
+        "| File | Lines | Collected tests | Area | Sub-area | Signals | Why this is a candidate |",
+        "|---|---:|---:|---|---|---|---|",
+    ]
+
+    if not low_risk:
+        lines.append("| _None_ | - | - | - | - | - | - |")
+
+    for metric in low_risk[:10]:
+        signals = ", ".join(metric.signals) if metric.signals else "-"
+        lines.append(
+            f"| `{metric.path}` | {metric.lines} | {metric.collected} | "
+            f"{metric.area} | {metric.sub_area} | {signals} | {include_reasons(metric)} |"
+        )
+
+    lines.append("")
+    return lines
+
+
+def taxonomy_gap_section(metrics: list[FileMetric]) -> list[str]:
+    uncategorized = [
+        metric
+        for metric in metrics
+        if metric.area == "uncategorized"
+    ]
+    uncategorized = sorted(
+        uncategorized,
+        key=lambda m: (m.collected, m.lines),
+        reverse=True,
+    )
+
+    lines = [
+        "## Taxonomy coverage gaps among split candidates",
+        "",
+        "`uncategorized` is a current taxonomy area, not a builder failure.",
+        "This plan does not reclassify tests because taxonomy changes should be reviewed separately from oversized-file split planning.",
+        "",
+        "Before using any of these files as a split target, first decide whether the taxonomy should be refined in a separate focused issue/PR.",
+        "",
+        "| File | Lines | Collected tests | Sub-area | Signals | Suggested follow-up |",
+        "|---|---:|---:|---|---|---|",
+    ]
+
+    if not uncategorized:
+        lines.append("| _None_ | - | - | - | - | - |")
+
+    for metric in uncategorized:
+        signals = ", ".join(metric.signals) if metric.signals else "-"
+        follow_up = "Review taxonomy mapping before using as a split target."
+        if HIGH_RISK_SIGNALS.intersection(metric.signals):
+            follow_up = "Review taxonomy and setup/risk boundaries before any split."
+        lines.append(
+            f"| `{metric.path}` | {metric.lines} | {metric.collected} | "
+            f"{metric.sub_area} | {signals} | {follow_up} |"
+        )
+
+    lines.append("")
+    return lines
+
+
+def deferred_section(metrics: list[FileMetric]) -> list[str]:
+    deferred = [
+        metric
+        for metric in metrics
+        if HIGH_RISK_SIGNALS.intersection(metric.signals)
+    ]
+    deferred = sorted(deferred, key=lambda m: (m.collected, m.lines), reverse=True)
+
+    lines = [
+        "## High-risk candidates to defer first",
+        "",
+        "These files may still be split later, but not as the first implementation slice without a separate manual boundary review.",
+        "",
+        "| File | Lines | Collected tests | High-risk signals |",
+        "|---|---:|---:|---|",
+    ]
+
+    for metric in deferred[:15]:
+        signals = ", ".join(sorted(HIGH_RISK_SIGNALS.intersection(metric.signals)))
+        lines.append(
+            f"| `{metric.path}` | {metric.lines} | {metric.collected} | {signals} |"
+        )
+
+    lines.append("")
+    return lines
+
+
+def write_distribution(
+    lines: list[str],
+    title: str,
+    values: Counter[str],
+    *,
+    min_count: int = 1,
+) -> None:
+    displayed = [
+        (value, count)
+        for value, count in sorted(values.items())
+        if count >= min_count
+    ]
+    omitted_values = sum(1 for count in values.values() if count < min_count)
+    omitted_files = sum(count for count in values.values() if count < min_count)
+
+    lines.extend([
+        f"{title}:",
+        "",
+        "| Value | Files |",
+        "|---|---:|",
+    ])
+    for value, count in displayed:
+        lines.append(f"| {value} | {count} |")
+
+    if omitted_values:
+        lines.extend([
+            "",
+            f"Values below {min_count} files: {omitted_values} values covering {omitted_files} files.",
+        ])
+
+    lines.append("")
+
+
+def write_report(metrics: list[FileMetric], node_count_total: int) -> None:
+    by_lines = sorted(metrics, key=lambda m: (m.lines, m.collected), reverse=True)
+    by_collected = sorted(metrics, key=lambda m: (m.collected, m.lines), reverse=True)
+    candidates = sorted(
+        candidate_metrics(metrics),
+        key=lambda m: (m.collected, m.lines),
+        reverse=True,
+    )
+
+    areas = Counter(metric.area for metric in metrics)
+    sub_areas = Counter(metric.sub_area for metric in metrics)
+
+    lines = [
+        "# Oversized Test File Split Plan",
+        "",
+        "## Purpose",
+        "",
+        "This document plans future oversized test-file splits using current repo data.",
+        "It does not move files, rewrite assertions, extract helpers, or change CI.",
+        "",
+        "## Roadmap context",
+        "",
+        "- Issue: #3983",
+        "- Parent tracker: #2523",
+        "- Follows #3973 / #3982, the report-only order-sensitivity diagnostics slice.",
+        "",
+        "## Methodology",
+        "",
+        "Metrics were generated from the current test tree using:",
+        "",
+        "- physical line counts for every recursive `test_*.py` file under `tests/`;",
+        "- AST counts for `test_*` functions and `Test*` classes;",
+        "- one `pytest --collect-only -q tests` run to count collected items per file;",
+        "- current taxonomy classification from `tests._taxonomy.classify_test_path`; and",
+        "- static setup-signal scans for route/API, DB/session, import-state, security, filesystem, subprocess/script, async/threading, and UI/static indicators.",
+        "",
+        "Static signals are not proof of risk. They are review prompts.",
+        "Future split PRs must still inspect each file manually before editing.",
+        "",
+        "## Current summary",
+        "",
+        f"- test files scanned: {len(metrics)}",
+        f"- collected pytest items counted: {node_count_total}",
+        f"- large-file threshold: {LARGE_LINE_THRESHOLD} lines",
+        f"- large-collected threshold: {LARGE_NODE_THRESHOLD} collected items",
+        "",
+    ]
+
+    write_distribution(lines, "Area distribution", areas)
+    write_distribution(lines, "Sub-area distribution", sub_areas, min_count=2)
+
+    lines.extend(metric_table("Top files by collected pytest items", by_collected[:TOP_LIMIT]))
+    lines.extend(metric_table("Top files by physical line count", by_lines[:TOP_LIMIT]))
+    lines.extend(candidate_section(candidates))
+    lines.extend(taxonomy_gap_section(candidates))
+    lines.extend(first_manual_review_section(candidates))
+    lines.extend(deferred_section(candidates))
+
+    lines.extend([
+        "## Rules for future split PRs",
+        "",
+        "- One file or one coherent file-family per PR.",
+        "- No assertion rewrites mixed with file moves.",
+        "- No helper extraction mixed with file moves.",
+        "- No production code changes.",
+        "- No CI workflow changes.",
+        "- Preserve existing markers and taxonomy unless the split issue explicitly says otherwise.",
+        "- Validate the original file's collected tests before and after the split.",
+        "- Validate any neighboring taxonomy/focused-runner behavior if paths change.",
+        "- Treat files with route/API, DB/session, import-state, or security signals as higher-risk until manually reviewed.",
+        "",
+        "## Suggested next step",
+        "",
+        "Use this plan to choose the first actual oversized-file split issue.",
+        "The first split should prefer a file with high review value and low setup risk.",
+        "Do not start a split PR from this planning issue alone if the file's boundaries are still ambiguous.",
+        "",
+        "## Reproduction command",
+        "",
+        "This document was generated with:",
+        "",
+        "```bash",
+        ".venv/bin/python tests/tools/build_oversized_test_split_plan.py",
+        "```",
+        "",
+        "## Freshness check",
+        "",
+        "After editing the builder or rebasing the branch, regenerate the plan and confirm no unexpected plan drift:",
+        "",
+        "```bash",
+        ".venv/bin/python tests/tools/build_oversized_test_split_plan.py",
+        "git diff --exit-code -- tests/OVERSIZED_TEST_SPLIT_PLAN.md",
+        "```",
+        "",
+    ])
+
+    OUTPUT.write_text("\n".join(lines), encoding="utf-8")
+
+
+def write_raw(metrics: list[FileMetric]) -> None:
+    raw = [
+        {
+            "area": metric.area,
+            "collected": metric.collected,
+            "lines": metric.lines,
+            "nonblank": metric.nonblank,
+            "path": metric.path,
+            "signals": list(metric.signals),
+            "sub_area": metric.sub_area,
+            "test_classes": metric.test_classes,
+            "test_defs": metric.test_defs,
+        }
+        for metric in metrics
+    ]
+    RAW_OUTPUT.write_text(json.dumps(raw, indent=2, sort_keys=True), encoding="utf-8")
+
+
+def assert_taxonomy_worked(metrics: list[FileMetric]) -> None:
+    if not metrics:
+        raise SystemExit("ERROR: no test files were scanned")
+
+    unknown = sum(1 for metric in metrics if metric.area == "unknown")
+    if unknown == len(metrics):
+        raise SystemExit("ERROR: taxonomy classification returned unknown for every file")
+
+
+def main() -> int:
+    if not TESTS_DIR.exists():
+        print("ERROR: tests/ directory not found", file=sys.stderr)
+        return 1
+
+    classify_test_path = load_taxonomy_classifier()
+    node_counts = collect_node_counts()
+    metrics = [metric_for(path, node_counts, classify_test_path) for path in test_files()]
+
+    assert_taxonomy_worked(metrics)
+    write_report(metrics, sum(node_counts.values()))
+    write_raw(metrics)
+
+    print(f"Wrote {OUTPUT.relative_to(ROOT)}")
+    print(f"Wrote {RAW_OUTPUT}")
+    return 0
+
+
+if __name__ == "__main__":
+    raise SystemExit(main())