feat(forge): per-user forge connections — keyring, host-keyed resolution, project pin (#2778)
CI & Build / Python lint (push) Successful in 4s
CI & Build / Plugin hooks (push) Successful in 8s
CI & Build / TypeScript typecheck (push) Successful in 41s
CI & Build / integration (push) Successful in 37s
CI & Build / Python tests (push) Successful in 1m4s
CI & Build / Build & push image (push) Successful in 40s

A forge token is a user's credential, not an instance's. The single
admin-settings config is replaced by per-user keyring rows (one per forge
host), and every server-side forge read runs on the PROJECT OWNER's keyring:

- forge_connections table + projects.forge_connection_id pin (migration 0078,
  which also carries the existing admin config into the first admin's row and
  deletes the old setting keys — no legacy dual-read)
- get_forge() replaced by get_forges(owner_id, project_id) -> ForgeSelector;
  resolve(repo) picks the connection whose host serves the repo. A pinned
  project uses ONLY its pinned connection; a stale pin (ownership moved) is
  ignored, never honored across users
- env FORGE_* config survives as an implicit entry for admin owners only;
  a stored row for the same host beats it
- consumers threaded: pull-time freshness (owner of the note), coverage
  (owner of the project), coverage routes' configured flag
- routes: /api/settings/forge-connections CRUD + per-connection test
  (own-rows only, tokens never returned); /api/admin/forge shrinks to
  /api/admin/forge-webhook (secret only); PUT /api/projects/<id>/forge pins,
  owner-or-admin asking, owner's connections only
- UI: Git Forges card moves to Settings -> Integrations as a connection
  list; webhook secret stays in the admin Config tab; owner-only forge
  select on the project coverage card
- backups exclude forge_connections (credentials, api_keys precedent) and
  the pin, so restores fall back to keyring resolution

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
This commit is contained in:
2026-08-19 11:23:22 -04:00
co-authored by Claude Fable 5
parent 7a5e2b18d9
commit 1faf8f3ece
19 changed files with 1252 additions and 310 deletions
+6 -5
View File
@@ -60,11 +60,12 @@ class Config:
# the MCP layer doesn't proxy web search (Claude has its own).
SEARXNG_URL: str = os.environ.get("SEARXNG_URL", "")
# Git forge integration (#2689) — optional read access to the operator's
# forge so snippet bodies can be fetched/verified server-side. Normally
# configured in Settings → Config (stored as admin settings); these env
# fallbacks exist so a deployment can keep the token in a Docker secret
# instead of the database. DB value wins when both are set.
# Git forge integration (#2689) — optional read access to a git forge so
# snippet bodies can be fetched/verified server-side. Connections are
# per-user keyring rows (#2778, Settings → Git forges); these env values
# survive as an implicit keyring entry for ADMIN users' projects only, so
# a deployment can keep the operator's token in a Docker secret instead
# of the database. A stored row for the same host wins over the env entry.
FORGE_KIND: str = os.environ.get("FORGE_KIND", "")
FORGE_BASE_URL: str = os.environ.get("FORGE_BASE_URL", "").rstrip("/")
FORGE_TOKEN: str = _read_secret("FORGE_TOKEN", "FORGE_TOKEN_FILE", "")
+1
View File
@@ -43,5 +43,6 @@ from scribe.models.rulebook import ( # noqa: E402, F401
Rulebook, RulebookTopic, Rule, project_rulebook_subscriptions,
)
from scribe.models.repo_binding import RepoBinding # noqa: E402, F401
from scribe.models.forge_connection import ForgeConnection # noqa: E402, F401
from scribe.models.system import System, RecordSystem # noqa: E402, F401
from scribe.models.design_system import DesignSystem, DesignToken # noqa: E402, F401
+45
View File
@@ -0,0 +1,45 @@
from sqlalchemy import ForeignKey, Integer, Text, UniqueConstraint
from sqlalchemy.orm import Mapped, mapped_column
from scribe.models import Base
from scribe.models.base import TimestampMixin
class ForgeConnection(Base, TimestampMixin):
"""One user's read-only credential for one git forge host (#2778).
The keyring model: a user owns a set of connections and every server-side
forge read for a project runs on the PROJECT OWNER's set, resolved by the
repo's host. One row per (user, host) — the repo's host picks the
connection deterministically, so there is no "default forge" pointer to
maintain or tie-break.
`host` is derived from `base_url` at write time and stored because it is
the lookup key; the service layer keeps the two in step. The token is a
secret: to_dict never includes it, and no route may return it.
"""
__tablename__ = "forge_connections"
__table_args__ = (
UniqueConstraint("user_id", "host", name="uq_forge_connections_user_host"),
)
id: Mapped[int] = mapped_column(primary_key=True)
user_id: Mapped[int] = mapped_column(
Integer, ForeignKey("users.id", ondelete="CASCADE"), nullable=False
)
kind: Mapped[str] = mapped_column(Text, nullable=False)
base_url: Mapped[str] = mapped_column(Text, nullable=False)
host: Mapped[str] = mapped_column(Text, nullable=False)
token: Mapped[str] = mapped_column(Text, nullable=False)
def to_dict(self) -> dict:
return {
"id": self.id,
"user_id": self.user_id,
"kind": self.kind,
"base_url": self.base_url,
"host": self.host,
"created_at": self.created_at.isoformat(),
"updated_at": self.updated_at.isoformat(),
}
+10
View File
@@ -27,6 +27,15 @@ class Project(Base, TimestampMixin, SoftDeleteMixin):
design_system_id: Mapped[int | None] = mapped_column(
BigInteger, ForeignKey("design_systems.id", ondelete="SET NULL"), nullable=True
)
# The per-project forge pin (#2778). NULL is the ordinary state: forge
# reads resolve against the owner's keyring by repo host. When set, the
# project's forge reads use ONLY this connection — an explicit, auditable
# choice, constrained by the service layer to a connection the project
# OWNER holds (never a collaborator's token).
forge_connection_id: Mapped[int | None] = mapped_column(
BigInteger, ForeignKey("forge_connections.id", ondelete="SET NULL"),
nullable=True,
)
def to_dict(self) -> dict:
return {
@@ -38,6 +47,7 @@ class Project(Base, TimestampMixin, SoftDeleteMixin):
"status": self.status,
"color": self.color,
"design_system_id": self.design_system_id,
"forge_connection_id": self.forge_connection_id,
"created_at": self.created_at.isoformat(),
"updated_at": self.updated_at.isoformat(),
}
+16 -62
View File
@@ -19,15 +19,6 @@ from scribe.services.backup import (
restore_full_backup,
)
from scribe.services.email import SMTP_SETTING_KEYS, get_base_url, get_smtp_config, is_smtp_configured, send_test_email
from scribe.services.forge import (
FORGE_BASE_URL_KEY,
FORGE_KIND_KEY,
FORGE_KINDS,
FORGE_TOKEN_KEY,
ForgeError,
forge_config,
get_forge,
)
from scribe.services.logging import get_logs, get_log_stats, log_audit
from scribe.services.notifications import send_invitation_email
from scribe.services.settings import (
@@ -169,85 +160,48 @@ async def test_smtp():
_TOKEN_MASK = "********"
@admin_bp.route("/forge", methods=["GET"])
# The forge CONFIG moved to per-user keyring rows (#2778, Settings → Git
# forges); what stays admin is the webhook secret, because the push endpoint
# is one URL per instance and authenticates deliveries, not users.
@admin_bp.route("/forge-webhook", methods=["GET"])
@admin_required
async def get_forge_settings():
async def get_forge_webhook_settings():
from scribe.config import Config
from scribe.routes.webhooks import FORGE_WEBHOOK_SECRET_KEY
cfg = await forge_config()
webhook_secret = (
await get_admin_setting(FORGE_WEBHOOK_SECRET_KEY, "")
or Config.FORGE_WEBHOOK_SECRET
)
return jsonify({
"kind": cfg["kind"],
"base_url": cfg["base_url"],
# Secrets never leave the server — the smtp_password convention:
# masked when set, empty when not.
"token": _TOKEN_MASK if cfg["token"] else "",
"webhook_secret": _TOKEN_MASK if webhook_secret else "",
"configured": bool(await get_forge()),
"kinds": list(FORGE_KINDS),
})
@admin_bp.route("/forge", methods=["PUT"])
@admin_bp.route("/forge-webhook", methods=["PUT"])
@admin_required
async def update_forge_settings():
data = await request.get_json() or {}
uid = get_current_user_id()
kind = str(data.get("kind", "")).strip().lower()
if kind and kind not in FORGE_KINDS:
return jsonify({"error": f"Unknown forge kind {kind!r}"}), 400
base_url = str(data.get("base_url", "")).strip().rstrip("/")
if base_url and not base_url.startswith(("http://", "https://")):
return jsonify({"error": "Forge base URL must use http or https"}), 400
await set_admin_setting(FORGE_KIND_KEY, kind)
await set_admin_setting(FORGE_BASE_URL_KEY, base_url)
token = data.get("token")
# The mask coming back means "unchanged" — the form round-trips what GET
# showed it, and storing the mask would silently break the integration.
if token is not None and token != _TOKEN_MASK:
await set_admin_setting(FORGE_TOKEN_KEY, str(token))
async def update_forge_webhook_settings():
from scribe.routes.webhooks import FORGE_WEBHOOK_SECRET_KEY
data = await request.get_json() or {}
uid = get_current_user_id()
webhook_secret = data.get("webhook_secret")
# The mask coming back means "unchanged" — the form round-trips what GET
# showed it, and storing the mask would silently break the integration.
if webhook_secret is not None and webhook_secret != _TOKEN_MASK:
await set_admin_setting(FORGE_WEBHOOK_SECRET_KEY, str(webhook_secret))
# The token is deliberately absent from the audit detail.
# The secret is deliberately absent from the audit detail.
await log_audit(
"forge_config", user_id=uid, username=g.user.username,
ip_address=request.remote_addr,
details={"kind": kind, "base_url": base_url},
"forge_webhook_config", user_id=uid, username=g.user.username,
ip_address=request.remote_addr, details={},
)
return jsonify({"status": "ok"})
@admin_bp.route("/forge/test", methods=["POST"])
@admin_required
async def test_forge():
"""Probe the SAVED forge config: reachability and token acceptance in one
press, so a misconfiguration is visible now rather than as silent
fallbacks later (#2663's lesson, applied to integrations)."""
uid = get_current_user_id()
forge = await get_forge()
if forge is None:
return jsonify({"error": "Forge is not configured — save kind, base URL and token first"}), 400
try:
result = await forge.check()
except ForgeError as e:
return jsonify({"error": str(e)}), 502
await log_audit(
"forge_test", user_id=uid, username=g.user.username,
ip_address=request.remote_addr,
details={"ok": True, "username": result.get("username", "")},
)
return jsonify(result)
@admin_bp.route("/logs", methods=["GET"])
@admin_required
async def list_logs():
+47 -9
View File
@@ -1,7 +1,7 @@
"""Project management routes."""
import logging
from quart import Blueprint, jsonify, request
from quart import Blueprint, g, jsonify, request
from scribe.auth import login_required, get_current_user_id
from scribe.routes.utils import not_found, parse_pagination
@@ -130,7 +130,7 @@ async def get_coverage_route(project_id: int):
push or an explicit refresh).
"""
from scribe.services.coverage import cached_coverage
from scribe.services.forge import get_forge
from scribe.services.forge import get_forges
uid = get_current_user_id()
result = await get_project_for_user(uid, project_id)
@@ -139,7 +139,9 @@ async def get_coverage_route(project_id: int):
project, _ = result
owner_uid = project.user_id or uid
return jsonify({
"configured": await get_forge() is not None,
# The OWNER's keyring (#2778) — whether refresh could do anything,
# regardless of who is looking.
"configured": (await get_forges(owner_uid, project_id)).configured,
"coverage": await cached_coverage(owner_uid, project_id),
})
@@ -153,7 +155,7 @@ async def refresh_coverage_route(project_id: int):
and wants the new number, and the forge timeout bounds the wait.
"""
from scribe.services.coverage import refresh_coverage
from scribe.services.forge import ForgeError, get_forge
from scribe.services.forge import ForgeError, get_forges
uid = get_current_user_id()
result = await get_project_for_user(uid, project_id)
@@ -161,20 +163,56 @@ async def refresh_coverage_route(project_id: int):
return not_found("Project")
project, _ = result
owner_uid = project.user_id or uid
if await get_forge() is None:
return jsonify({"error": "No git forge is configured (Settings → Config → Git Forge)"}), 400
selector = await get_forges(owner_uid, project_id)
if not selector.configured:
return jsonify({
"error": "The project owner has no forge connection "
"(Settings → Git forges)"
}), 400
try:
coverage = await refresh_coverage(owner_uid, project_id)
coverage = await refresh_coverage(owner_uid, project_id, selector=selector)
except ForgeError as exc:
return jsonify({"error": str(exc)}), 502
if coverage is None:
return jsonify({
"error": "No bound repo is served by the configured forge"
"bind the project's repo (bind_repo) on a remote the forge hosts"
"error": "No bound repo is served by the owner's forge connections"
"bind the project's repo (bind_repo) on a remote a connection hosts"
}), 400
return jsonify({"coverage": coverage})
@projects_bp.route("/<int:project_id>/forge", methods=["PUT"])
@login_required
async def set_project_forge_route(project_id: int):
"""Pin the project to one forge connection, or clear the pin (#2778).
Body: {"connection_id": <id> | null}. Owner-or-admin may ask; either way
the pin can only reference a connection the project OWNER holds — the
service enforces that, so a collaborator's token can never end up serving
someone else's project.
"""
from scribe.services.forge_connections import set_project_pin
uid = get_current_user_id()
result = await get_project_for_user(uid, project_id)
if result is None:
return not_found("Project")
project, permission = result
if permission != "owner" and g.user.role != "admin":
return jsonify({"error": "Only the project owner can change its forge"}), 403
data = await request.get_json() or {}
raw = data.get("connection_id")
if raw is not None and not isinstance(raw, int):
return jsonify({"error": "connection_id must be an integer or null"}), 400
owner_uid = project.user_id or uid
if not await set_project_pin(owner_uid, project_id, raw):
return jsonify({
"error": "That connection does not belong to the project owner"
}), 400
return jsonify({"forge_connection_id": raw})
@projects_bp.route("/<int:project_id>/notes", methods=["GET"])
@login_required
async def get_project_notes_route(project_id: int):
+102 -1
View File
@@ -20,7 +20,8 @@ settings_bp = Blueprint("settings", __name__, url_prefix="/api/settings")
# read and skip the mask on write; this generic KV surface has to apply the
# same treatment, or it silently un-masks what those endpoints masked — the
# rows live on the admin's own user_id, so the plain GET returned them raw.
_SECRET_KEYS = frozenset({"smtp_password", "forge_token", "forge_webhook_secret"})
# (forge_token left with 0078: forge credentials are keyring rows now, #2778.)
_SECRET_KEYS = frozenset({"smtp_password", "forge_webhook_secret"})
_SECRET_MASK = "********"
@@ -73,3 +74,103 @@ async def test_search():
if not Config.searxng_enabled():
return jsonify({"configured": False, "results": [], "searxng_url": ""})
return jsonify({"configured": True, "results": [], "searxng_url": Config.SEARXNG_URL})
# --- forge connections (#2778) ------------------------------------------------
# The user's keyring: read-only forge credentials, one per host, resolved by
# repo host for every server-side forge read on the user's projects. Strictly
# own-rows — a connection is a credential, and there is no admin view of
# another user's keyring. Tokens never leave the server: the model's to_dict
# omits them, and the routes never echo the submitted value back.
@settings_bp.route("/forge-connections", methods=["GET"])
@login_required
async def list_forge_connections_route():
from scribe.services.forge import FORGE_KINDS
from scribe.services.forge_connections import list_connections
uid = get_current_user_id()
rows = await list_connections(uid)
return jsonify({
"connections": [r.to_dict() for r in rows],
"kinds": list(FORGE_KINDS),
})
@settings_bp.route("/forge-connections", methods=["POST"])
@login_required
async def create_forge_connection_route():
from scribe.services.forge_connections import create_connection
uid = get_current_user_id()
data = await request.get_json() or {}
try:
row = await create_connection(
uid,
kind=str(data.get("kind", "")),
base_url=str(data.get("base_url", "")),
token=str(data.get("token", "")),
)
except ValueError as exc:
return jsonify({"error": str(exc)}), 400
return jsonify(row.to_dict()), 201
@settings_bp.route("/forge-connections/<int:connection_id>", methods=["PUT"])
@login_required
async def update_forge_connection_route(connection_id: int):
from scribe.services.forge_connections import update_connection
uid = get_current_user_id()
data = await request.get_json() or {}
token = str(data.get("token", ""))
# The mask coming back means "unchanged" — the form round-trips what the
# list showed, and storing the mask would silently break the connection.
if token == _SECRET_MASK:
token = ""
try:
row = await update_connection(
uid, connection_id,
kind=str(data.get("kind", "")),
base_url=str(data.get("base_url", "")),
token=token,
)
except ValueError as exc:
return jsonify({"error": str(exc)}), 400
if row is None:
return jsonify({"error": "Connection not found"}), 404
return jsonify(row.to_dict())
@settings_bp.route("/forge-connections/<int:connection_id>", methods=["DELETE"])
@login_required
async def delete_forge_connection_route(connection_id: int):
from scribe.services.forge_connections import delete_connection
uid = get_current_user_id()
if not await delete_connection(uid, connection_id):
return jsonify({"error": "Connection not found"}), 404
return "", 204
@settings_bp.route("/forge-connections/<int:connection_id>/test", methods=["POST"])
@login_required
async def test_forge_connection_route(connection_id: int):
"""Probe the SAVED connection: reachability and token acceptance in one
press, so a misconfiguration is visible now rather than as silent
fallbacks later (#2663's lesson, applied per keyring row)."""
from scribe.services.forge import ForgeError, build_adapter
from scribe.services.forge_connections import get_connection
uid = get_current_user_id()
row = await get_connection(uid, connection_id)
if row is None:
return jsonify({"error": "Connection not found"}), 404
adapter = build_adapter(row.kind, row.base_url, row.token)
if adapter is None:
return jsonify({"error": "Connection is not usable — check kind and base URL"}), 400
try:
return jsonify(await adapter.check())
except ForgeError as exc:
return jsonify({"error": str(exc)}), 502
+6
View File
@@ -72,6 +72,12 @@ _NOT_INCLUDED = [
"api_keys", "note_embeddings", "app_logs", "notifications",
"invitation_tokens", "password_reset_tokens", "user_profiles",
"retrieval_logs",
# Sensitive credentials, same reasoning as api_keys: a backup that carries
# forge tokens is a token-exfiltration file. Users re-add connections
# after a restore; the per-project pin (projects.forge_connection_id) is
# deliberately not exported either, so restored projects fall back to
# keyring-by-host resolution — the documented unpinned behavior (#2778).
"forge_connections",
]
+17 -12
View File
@@ -33,7 +33,7 @@ import re
import tarfile
from datetime import datetime, timezone
from scribe.services.forge import ForgeAdapter, get_forge
from scribe.services.forge import ForgeSelector, get_forges
from scribe.services.repo_bindings import keys_for_project
from scribe.services.settings import get_setting, set_setting
@@ -252,27 +252,32 @@ async def _recorded_locations(user_id: int, project_id: int) -> list[tuple[str,
async def compute_coverage(
user_id: int, project_id: int, *, forge: ForgeAdapter | None = None
user_id: int, project_id: int, *, selector: ForgeSelector | None = None
) -> dict | None:
"""Measure a project's pattern-library coverage against its bound repos.
None means "nothing to measure"no forge configured, or none of the
project's bound repos is served by it. That is the ordinary state for a
forge-less install and every caller treats it as silence, not failure.
None means "nothing to measure"the owner's keyring serves none of the
project's bound repos (#2778). That is the ordinary state for a
forge-less user and every caller treats it as silence, not failure.
Forge errors (unreachable, bad token) RAISE — the two callers are a
refresh button and a background task, and both want to know.
``user_id`` is the project OWNER's id: the cache lives there, and the
keyring resolved here must be the same one every other read uses.
"""
forge = forge if forge is not None else await get_forge()
if forge is None:
if selector is None:
selector = await get_forges(user_id, project_id)
if not selector.configured:
return None
repos: list[dict] = []
matched_all: list[tuple[str, str, str, bool]] = []
recorded = await _recorded_locations(user_id, project_id)
for key in await keys_for_project(user_id, project_id):
api_repo = forge.resolve_repo(key)
if api_repo is None:
continue # bound to a host this forge doesn't serve
hit = selector.resolve(key)
if hit is None:
continue # bound to a host no connection serves
forge, api_repo = hit
ref = await forge.default_branch(api_repo)
shapes = shapes_from_archive(await forge.archive(api_repo, ref))
matched = match_shapes(shapes, recorded)
@@ -299,10 +304,10 @@ async def compute_coverage(
async def refresh_coverage(
user_id: int, project_id: int, *, forge: ForgeAdapter | None = None
user_id: int, project_id: int, *, selector: ForgeSelector | None = None
) -> dict | None:
"""Compute and cache. The only writer of the cache key."""
coverage = await compute_coverage(user_id, project_id, forge=forge)
coverage = await compute_coverage(user_id, project_id, selector=selector)
if coverage is not None:
await set_setting(
user_id, f"{_CACHE_KEY_PREFIX}{project_id}", json.dumps(coverage)
+124 -40
View File
@@ -8,10 +8,12 @@ webhooks (step 6), and coverage can be measured (step 7).
Design constraints, in force everywhere below:
- OPTIONAL per instance (rule #115). `get_forge()` returns None when nothing
is configured, and every consumer must treat None as "keep today's
behavior". An install that never configures a forge is not degraded — it
is the baseline.
- OPTIONAL per user (rule #115, sharpened by #2778). Connections are
per-user keyring rows resolved by repo host on the PROJECT OWNER's
keyring; `get_forges()` returns an empty selector when the owner has
nothing configured, and every consumer must treat that as "keep today's
behavior". A user who never configures a forge is not degraded — that is
the baseline.
- READ-ONLY by construction. The adapter exposes reads; there is no write
method to misuse. The token an operator mints for it only ever needs read
scope, and the docs say so.
@@ -40,18 +42,14 @@ from dataclasses import dataclass
from urllib.parse import quote, urlsplit
import httpx
from sqlalchemy import select
from scribe.config import Config
from scribe.services.repo_bindings import normalize_repo_key
from scribe.services.settings import get_admin_setting
logger = logging.getLogger(__name__)
FORGE_KIND_KEY = "forge_kind"
FORGE_BASE_URL_KEY = "forge_base_url"
FORGE_TOKEN_KEY = "forge_token"
# Kinds an instance can configure. Matches _FORGE_CLASSES below.
# Kinds a connection can use. Matches _FORGE_CLASSES below.
FORGE_KINDS = ("gitea", "github")
# Total budget per forge call. Consumers either have a cache to fall back to
@@ -86,7 +84,8 @@ class ForgeFile:
path: str
def _host_of(url: str) -> str:
def host_of(url: str) -> str:
"""The lowercase hostname of a URL — the keyring's lookup key (#2778)."""
return (urlsplit(url).hostname or "").lower()
@@ -111,7 +110,7 @@ class ForgeAdapter:
@property
def host(self) -> str:
return _host_of(self.base_url)
return host_of(self.base_url)
def resolve_repo(self, repo_or_url: str) -> str | None:
"""The forge-API repo path for a recorded repo — or None if this forge
@@ -359,39 +358,124 @@ _FORGE_CLASSES: dict[str, type[ForgeAdapter]] = {
}
async def forge_config() -> dict:
"""The instance's forge configuration, DB-first with env fallback.
def build_adapter(
kind: str, base_url: str, token: str, *, transport=None
) -> ForgeAdapter | None:
"""One validated adapter from raw connection values, or None.
The env channel exists so a deployment can keep the token out of the
database entirely (Docker secret via FORGE_TOKEN_FILE) — the DB value wins
when both are present because the admin UI writes there, and a UI edit
that silently loses to an env var would look exactly like a broken form.
None means "this connection cannot serve reads" — the same contract the
old instance-wide lookup had, applied per keyring row. Misconfigurations
are logged, never raised: a bad row must not break the reads the good
rows can still serve.
"""
return {
"kind": (await get_admin_setting(FORGE_KIND_KEY, "") or Config.FORGE_KIND)
.strip()
.lower(),
"base_url": (
await get_admin_setting(FORGE_BASE_URL_KEY, "") or Config.FORGE_BASE_URL
).rstrip("/"),
"token": await get_admin_setting(FORGE_TOKEN_KEY, "") or Config.FORGE_TOKEN,
}
async def get_forge(*, transport=None) -> ForgeAdapter | None:
"""The configured forge adapter, or None — and None means "behave exactly
as if this module did not exist", which every consumer must honor."""
cfg = await forge_config()
cls = _FORGE_CLASSES.get(cfg["kind"])
kind = (kind or "").strip().lower()
base_url = (base_url or "").rstrip("/")
cls = _FORGE_CLASSES.get(kind)
if cls is None:
if cfg["kind"]:
if kind:
# A kind we don't implement is a misconfiguration, not "off" —
# say so once per lookup rather than silently reading as absent.
logger.warning("unknown forge kind %r configured — forge disabled", cfg["kind"])
logger.warning("unknown forge kind %r configured — connection disabled", kind)
return None
if not cfg["base_url"] or not cfg["token"]:
if not base_url or not token:
return None
if not cfg["base_url"].startswith(("http://", "https://")):
logger.warning("forge base URL %r has no http(s) scheme — forge disabled", cfg["base_url"])
if not base_url.startswith(("http://", "https://")):
logger.warning("forge base URL %r has no http(s) scheme — connection disabled", base_url)
return None
return cls(cfg["base_url"], cfg["token"], transport=transport)
return cls(base_url, token, transport=transport)
@dataclass(frozen=True)
class ForgeSelector:
"""The forge reads available to one project owner (#2778).
Consumers ask it to serve a REPO, not to hand over "the forge": resolve()
walks the owner's adapters and returns the (adapter, api_repo) pair for
the first one whose host serves the repo — or None, which every consumer
treats exactly as the old "no forge configured" state. An empty selector
IS rule #115's baseline.
"""
adapters: tuple[ForgeAdapter, ...] = ()
@property
def configured(self) -> bool:
return bool(self.adapters)
def resolve(self, repo_or_url: str) -> tuple[ForgeAdapter, str] | None:
for adapter in self.adapters:
repo = adapter.resolve_repo(repo_or_url)
if repo is not None:
return adapter, repo
return None
async def get_forges(
owner_id: int, project_id: int | None = None, *, transport=None
) -> ForgeSelector:
"""The forge selector for reads on behalf of ``owner_id``'s records.
The keyring model (#2778): every server-side forge read for a record runs
on the PROJECT OWNER's connections, resolved by repo host — a forge token
is a user's credential, and one user's reads must never ride another
user's token. Pass the record's ``project_id`` so the per-project pin
applies: a pinned project uses ONLY its pinned connection (explicit and
auditable); a pin that no longer belongs to the owner (ownership moved) is
ignored with a warning rather than honored across users.
The env config (FORGE_KIND/FORGE_BASE_URL/FORGE_TOKEN) survives as an
implicit keyring entry for ADMIN owners only — it is the operator's
token, so it must not serve other users' reads — and a stored row for the
same host beats it, because the UI writes rows.
"""
from scribe.models import async_session
from scribe.models.forge_connection import ForgeConnection
from scribe.models.project import Project
from scribe.models.user import User
async with async_session() as session:
pinned_id = None
if project_id:
pinned_id = (
await session.execute(
select(Project.forge_connection_id).where(Project.id == project_id)
)
).scalar_one_or_none()
rows = list(
(
await session.execute(
select(ForgeConnection)
.where(ForgeConnection.user_id == owner_id)
.order_by(ForgeConnection.id)
)
).scalars().all()
)
role = ""
if Config.FORGE_KIND and Config.FORGE_BASE_URL and Config.FORGE_TOKEN:
role = (
await session.execute(select(User.role).where(User.id == owner_id))
).scalar_one_or_none() or ""
if pinned_id:
pin = next((r for r in rows if r.id == pinned_id), None)
if pin is not None:
adapter = build_adapter(pin.kind, pin.base_url, pin.token, transport=transport)
return ForgeSelector((adapter,) if adapter is not None else ())
logger.warning(
"project %s pins forge connection %s the owner (%s) does not hold — pin ignored",
project_id, pinned_id, owner_id,
)
adapters: list[ForgeAdapter] = []
for row in rows:
adapter = build_adapter(row.kind, row.base_url, row.token, transport=transport)
if adapter is not None:
adapters.append(adapter)
if role == "admin":
env = build_adapter(
Config.FORGE_KIND, Config.FORGE_BASE_URL, Config.FORGE_TOKEN,
transport=transport,
)
if env is not None and all(a.host != env.host for a in adapters):
adapters.append(env)
return ForgeSelector(tuple(adapters))
+194
View File
@@ -0,0 +1,194 @@
"""User-level forge connection CRUD — the keyring rows get_forges reads (#2778).
A connection is a user's read-only credential for one forge host; one row per
(user, host) keeps host-keyed resolution deterministic with no default-pointer
machinery. Everything here is own-rows-only: a connection is a credential, and
no caller — admin included — reads or edits another user's. The token never
leaves the server (model.to_dict omits it; routes mask "set/unset").
Validation matches what build_adapter will accept, checked here so a bad
value is a 400 at the form instead of a silently dead keyring row.
"""
from __future__ import annotations
from sqlalchemy import select
from scribe.models import async_session
from scribe.models.forge_connection import ForgeConnection
from scribe.models.project import Project
from scribe.services.forge import FORGE_KINDS, host_of
def validate_connection(kind: str, base_url: str) -> str | None:
"""The error a connection's non-secret values would earn, or None."""
if kind not in FORGE_KINDS:
return f"Unknown forge kind {kind!r} (one of: {', '.join(FORGE_KINDS)})"
if not base_url.startswith(("http://", "https://")):
return "Forge base URL must use http or https"
if not host_of(base_url):
return "Forge base URL carries no hostname"
return None
async def list_connections(user_id: int) -> list[ForgeConnection]:
async with async_session() as session:
rows = await session.execute(
select(ForgeConnection)
.where(ForgeConnection.user_id == user_id)
.order_by(ForgeConnection.host)
)
return list(rows.scalars().all())
async def get_connection(user_id: int, connection_id: int) -> ForgeConnection | None:
async with async_session() as session:
return (
await session.execute(
select(ForgeConnection).where(
ForgeConnection.id == connection_id,
ForgeConnection.user_id == user_id,
)
)
).scalar_one_or_none()
async def create_connection(
user_id: int, *, kind: str, base_url: str, token: str
) -> ForgeConnection:
"""Create a keyring row. Raises ValueError on bad values or a host the
user already holds — one row per (user, host) IS the resolution model,
so a second token for the same host is an update, not a create."""
kind = (kind or "").strip().lower()
base_url = (base_url or "").strip().rstrip("/")
token = token or ""
error = validate_connection(kind, base_url)
if error is None and not token:
error = "A token is required (read scope is enough)"
if error:
raise ValueError(error)
host = host_of(base_url)
async with async_session() as session:
existing = (
await session.execute(
select(ForgeConnection).where(
ForgeConnection.user_id == user_id,
ForgeConnection.host == host,
)
)
).scalar_one_or_none()
if existing is not None:
raise ValueError(
f"You already have a connection for {host} — edit that one; "
"resolution is by host, so a second row could never be reached"
)
row = ForgeConnection(
user_id=user_id, kind=kind, base_url=base_url, host=host, token=token
)
session.add(row)
await session.commit()
await session.refresh(row)
return row
async def update_connection(
user_id: int,
connection_id: int,
*,
kind: str = "",
base_url: str = "",
token: str = "",
) -> ForgeConnection | None:
"""Update own row; empty string = leave unchanged (the settings-form
sentinel convention). None when the row isn't the caller's."""
async with async_session() as session:
row = (
await session.execute(
select(ForgeConnection).where(
ForgeConnection.id == connection_id,
ForgeConnection.user_id == user_id,
)
)
).scalar_one_or_none()
if row is None:
return None
new_kind = (kind or "").strip().lower() or row.kind
new_base = (base_url or "").strip().rstrip("/") or row.base_url
error = validate_connection(new_kind, new_base)
if error:
raise ValueError(error)
new_host = host_of(new_base)
if new_host != row.host:
clash = (
await session.execute(
select(ForgeConnection.id).where(
ForgeConnection.user_id == user_id,
ForgeConnection.host == new_host,
ForgeConnection.id != row.id,
)
)
).scalar_one_or_none()
if clash is not None:
raise ValueError(
f"You already have a connection for {new_host} — edit that one"
)
row.kind = new_kind
row.base_url = new_base
row.host = new_host
if token:
row.token = token
await session.commit()
await session.refresh(row)
return row
async def delete_connection(user_id: int, connection_id: int) -> bool:
"""Delete own row. Project pins pointing at it go NULL (FK SET NULL) —
those projects fall back to keyring resolution, which is the documented
unpinned behavior, not a surprise."""
async with async_session() as session:
row = (
await session.execute(
select(ForgeConnection).where(
ForgeConnection.id == connection_id,
ForgeConnection.user_id == user_id,
)
)
).scalar_one_or_none()
if row is None:
return False
await session.delete(row)
await session.commit()
return True
async def set_project_pin(
owner_id: int, project_id: int, connection_id: int | None
) -> bool:
"""Point a project at one of its OWNER's connections, or clear the pin.
The caller settles WHO may ask (routes check owner-or-admin); this
settles WHOSE connection is eligible: only the project owner's — pinning
a collaborator's token to someone else's project is the confused-deputy
channel this feature exists to close. False = project or connection not
eligible.
"""
async with async_session() as session:
project = (
await session.execute(select(Project).where(Project.id == project_id))
).scalar_one_or_none()
if project is None or (project.user_id or 0) != owner_id:
return False
if connection_id:
held = (
await session.execute(
select(ForgeConnection.id).where(
ForgeConnection.id == connection_id,
ForgeConnection.user_id == owner_id,
)
)
).scalar_one_or_none()
if held is None:
return False
project.forge_connection_id = connection_id or None
await session.commit()
return True
+18 -13
View File
@@ -941,10 +941,10 @@ async def record_verification(
# --- pull-time freshness (#2690) ---------------------------------------------
# A pull is the moment freshness matters: the reader is about to trust the
# cached body. When the instance has a forge configured, the pull fetches the
# recorded file and answers the one mechanically-answerable question — does
# the cached code still appear in the source, verbatim after whitespace
# normalization? The body is a FRAGMENT of the file, so "serve the fetched
# cached body. When the record owner's keyring serves a forge, the pull
# fetches the recorded file and answers the one mechanically-answerable
# question — does the cached code still appear in the source, verbatim after
# whitespace normalization? The body is a FRAGMENT of the file, so "serve the fetched
# file" would clobber the record; confirmation + provenance refresh is what
# fetching can honestly deliver, and divergence is reported, not overwritten.
#
@@ -990,7 +990,8 @@ async def _refresh_provenance(note, commit_sha: str) -> None:
async def attach_live_body(note, data: dict) -> None:
"""Decorate a PULL response with forge-checked freshness (#2690).
Adds, when (and only when) a forge is configured:
Adds, when (and only when) the record owner's keyring serves a forge
(#2778):
- ``body_source``: "forge" (confirmed against the source just now) or
"cache" (the stored body, for whatever reason follows)
- ``body_freshness``: "current" | "diverged" | "missing" |
@@ -1004,14 +1005,17 @@ async def attach_live_body(note, data: dict) -> None:
attention state verify_snippet uses.
"""
from scribe.services.background import spawn
from scribe.services.forge import ForgeError, ForgeNotFound, get_forge
from scribe.services.forge import ForgeError, ForgeNotFound, get_forges
try:
forge = await get_forge()
# The OWNER's keyring, honoring the project pin (#2778) — freshness
# for a record is checked with its owner's credential, never the
# reader's.
selector = await get_forges(note.user_id, getattr(note, "project_id", None))
except Exception:
logger.warning("forge lookup failed during pull", exc_info=True)
return
if forge is None:
if not selector.configured:
return
fields = data.get("snippet") if isinstance(data.get("snippet"), dict) else None
@@ -1033,18 +1037,19 @@ async def attach_live_body(note, data: dict) -> None:
# address a forge API — the project's repo BINDING is the identity that
# can (#2691). Try the location string first (it may be a real remote),
# then fall back to the bindings of the snippet's project.
repo = forge.resolve_repo(loc["repo"])
if repo is None and getattr(note, "project_id", None):
resolved = selector.resolve(loc["repo"])
if resolved is None and getattr(note, "project_id", None):
from scribe.services.repo_bindings import keys_for_project
for key in await keys_for_project(note.user_id, note.project_id):
repo = forge.resolve_repo(key)
if repo is not None:
resolved = selector.resolve(key)
if resolved is not None:
break
if repo is None:
if resolved is None:
data["body_source"] = "cache"
data["body_freshness"] = "repo-not-on-this-forge"
return
forge, repo = resolved
stored_prov_sha = (fields.get("provenance") or {}).get("commit_sha") or ""