Zeren Wang a58ab484a6
feat(projects): Projects MVP Phase 2 — instructions, document shelf, promotion, trash (#5443)
* feat(projects): Projects MVP Phase 2 — instructions, document shelf, promotion, trash

Implements docs/superpowers/specs/2026-09-12-projects-mvp-phase2-design.md
(issue #5160, tracker #5129) in the slice order of the spec's §16.

Slices:
- A: ProjectsConfig + write-time 422 UTF-8 byte cap; PROJECT_CONTEXT_KEY
  admission pinning (both server-owned sets + worker hoist); latest-only
  request-scoped <project> block via DynamicContextMiddleware
  wrap_model_call/awrap_model_call (idempotent reassembly, reserved ID
  prefix + marker + provenance, never persisted); journal audit
  fingerprints; Instructions tab.
- B: ProjectDocumentRow + migration 0023; ProjectDocumentRepository with
  locked check-and-set; hash-qualified immutable shelf storage with
  Paths helpers; upload/list/content/delete-to-trash routes; project
  delete trashes the shelf in-transaction; request-scoped bounded
  <documents> index with honest count/shown + actionable overflow note;
  list_project_documents/read_project_document tools registered only on
  pinned runs; PAT allowlist + drift guards; blocking-IO anchors.
- C: shared thread-upload ingestion service (uploads router refactored to
  parity); POST from-thread with provenance; attach-to-thread with
  lock-staged copy (archived source allowed); read-only thread-files
  view with per-group truncation reporting.
- D: restore (restored/merged/not_found/no_target/content_missing; no
  file moves), purge (continuous row lock across unlink/delete/commit,
  retryable on FS errors), retention sweep (lazy + startup, 24h orphan
  guard, row-side reconciliation never deletes).
- E: Documents tab (shelf + conversation-files browser, provenance,
  archived banner, content-missing rows), /workspace/trash route,
  sidebar entry, composer attach handoff, i18n (en-US/zh-CN), e2e mocks
  + specs.

Review hardening folded in (10 rounds, all with tests):
- force active shelf content (HTML/XML family) to download; nosniff on
  artifact + content responses; unified unsandboxed-iframe PDF preview
  (fixes the pre-existing Chromium sandbox blank in the artifact viewer)
- scope document trash to the URL project under the document lock
- atomic no-overwrite filename reservation for ALL ingestion (seeded
  claims + os.link commit with suffix retry; same-name re-upload now
  unique-names instead of replacing); hidden staging only, no visible
  placeholders; lease cleanup on setup failure
- serialize conversion under the document lock with post-lock active
  revalidation; drain locked filesystem work on cancellation; preserve
  bytes when an insert's commit state is uncertain (including trashed
  rows)
- original-integrity checks before serving text or cached conversions;
  content_missing surfaced in list responses (UI reads the flag, no
  409-probe); downloads always serve original bytes
- bounded streaming document reads with cached char counts; shelf limits
  declared in middleware release identity
- thread-root confinement for from-thread sources; config fallback
  rejects fractional/infinite values; composer counts staged
  attachments; pending attachments persist until submission or removal;
  in-flight instruction/rename edits survive save refetches; shelf and
  trash pagination; conversation-file and thread-files pages stay
  subscribed to refetches

Docs: README/README_zh, backend API.md/ARCHITECTURE.md, AGENTS.md
contracts, config.example.yaml projects block.

Review follow-ups (head b4807477 → this revision):
- The trash retention sweep is split so repeated lazy triggers stay
  bounded: the indexed expiry purge still runs on every trigger
  (GET /api/trash/documents, POST /api/trash/purge) while the
  O(all rows + all files) reconciliation is throttled to one run per
  user per 15 minutes (process-local, per-user window). The startup
  sweep now runs as a background task instead of blocking gateway
  readiness, and shutdown awaits it (bounded).
- The export scrub (stripInternalMarkers) is fence- and indentation-aware
  like the render path, so a pasted, fenced <project>/<documents> snippet
  survives markdown export while real injected blocks (never fenced) are
  still removed. Fence regexes moved to a dependency-free leaf module to
  avoid the messages↔streamdown import cycle.
- The artifact viewer's PDF iframe no longer carries an added title
  attribute (the upstream e2e contract locates it via :not([title])), and
  the upstream artifact-preview spec now pins the new contract: PDFs
  render unsandboxed, images keep sandbox="".

* fix(projects): round-2 review — cancel an overrun trash sweep, restore the PDF frame title

- Shutdown cancelled only the shield around the background startup sweep,
  so an all-users reconciliation that outlived the 5s budget kept walking
  rows and files while the document repo and DB engine were disposed
  underneath it. The wait now lives in `_shutdown_startup_trash_sweep`,
  which cancels the task and drains it before worker exit: the shield
  keeps the wait bounded, the cancel makes it final (CancelledError lands
  at the sweep's next await, and `_run_startup_trash_sweep` only catches
  `Exception`, so nothing swallows it).
- The browser-preview iframe lost `title={getFileName(filepath)}` in the
  previous fix round, leaving the PDF frame without an accessible name
  while its siblings keep theirs. Restore it (WCAG frame titles), assert
  it in the DOM test, and anchor the e2e on `iframe[title="report.pdf"]`
  instead of `iframe:not([title])`.

* fix(projects): round-3 review — report the sweep's late finish, not a phantom cancel

`Task.cancel()` returns False when the sweep already finished inside the
window between the deadline firing and the cancel, so the shutdown log
claimed a cancellation that never happened. Branch on that outcome: the
warning stays for a real cancel, a late finish is logged at info, and both
paths still reap the task before worker exit.

* fix(projects): round-4 review — make Empty trash delete what it confirms

`POST /api/trash/purge` only ran the retention sweep, and the sweep's
candidate selection is age-gated, so a freshly trashed document survived
"Empty trash" even though the confirmation promises that every listed
document is permanently deleted. With one trashed row the route answered
`{"purged": 0}` and left it in place; `GET /api/trash/documents` sweeps
expired rows before listing, so the visible rows were normally ineligible
for the action by construction.

Empty trash now drives `purge_all_trashed`: the caller's trashed rows
(`list_all_trashed`, no age filter) each go through the same guarded,
row-locked `purge` as the single-document delete — bytes first, then the
row, in one transaction — so a row restored mid-flight is skipped instead of
force-deleted, and an unlink failure rolls that row back and answers 500 with
a retryable message. Retention expiry stays where it was: the sweep's
`purge_candidates` is now the only age-gated selection, and the lazy
retention sweep still runs on the listing and at startup.

Tests: the router suite replaces the retention-gated expectation with the
reviewer's repro (fresh row purged, bytes unlinked, shelf and other users'
trash untouched, a failing unlink stays retryable and 500); a blocking-I/O
anchor drives the new entry point through the offload; the mocked e2e covers
the action end to end; a new real-backend spec performs it against the real
gateway and re-reads `GET /api/trash/documents`. README, API, ARCHITECTURE
and the phase-2 design docs (en+zh) state the age-independent contract.
2026-09-16 18:46:18 +08:00

234 lines
11 KiB
Python

"""Personal Access Token (PAT) credentials for programmatic API access.
Tokens are ``dfp_`` + base62(32 CSPRNG bytes), shown exactly once in the
create response and persisted only as a SHA-256 digest. Validation is a
digest-indexed lookup plus a constant-time re-comparison, with a single
generic failure surface so a 401 never reveals which check failed.
v1 scopes are exactly the route-permission strings owned by
``app.gateway.authz`` — a PAT can only narrow its owning user's
permissions, never widen them.
"""
from __future__ import annotations
import functools
import hashlib
import hmac
import re
import secrets
from typing import Any
PAT_TOKEN_PREFIX = "dfp_"
PAT_RANDOM_BYTES = 32
# Best-effort ``last_used_at`` writes are throttled per token so high-volume
# automation does not turn every request into a database write.
PAT_LAST_USED_WRITE_INTERVAL_SECONDS = 300.0
PAT_ALLOWED_SCOPES: frozenset[str] = frozenset(
{
"threads:read",
"threads:write",
"threads:delete",
"runs:create",
"runs:read",
"runs:cancel",
"projects:read",
"projects:write",
"projects:delete",
}
)
PAT_MAX_NAME_LENGTH = 128
# Default-deny route boundary for PAT callers (#5041 review P1-1): scope
# intersection in AuthMiddleware only constrains routes that consult
# ``request.state.auth.permissions`` (``@require_permission``). Authenticated
# mutation routes without that decorator — memory deletion, agent creation,
# credential switching, channel configuration — would otherwise accept a
# PAT holding a single read scope. A route is reachable by PAT only when it
# is explicitly listed here, and only together with the thread/run lifecycle
# the v1 scopes govern; everything else answers 403 regardless of scopes.
_PAT_ROUTE_RULES: tuple[tuple[frozenset[str], re.Pattern[str]], ...] = (
(frozenset({"POST"}), re.compile(r"^/api/threads$")),
(frozenset({"POST"}), re.compile(r"^/api/threads/search$")),
(frozenset({"GET", "PATCH", "DELETE"}), re.compile(r"^/api/threads/[^/]+$")),
(frozenset({"GET", "PUT", "DELETE"}), re.compile(r"^/api/threads/[^/]+/goal$")),
(frozenset({"GET", "POST"}), re.compile(r"^/api/threads/[^/]+/state$")),
(frozenset({"POST"}), re.compile(r"^/api/threads/[^/]+/(compact|history|branches)$")),
(frozenset({"POST"}), re.compile(r"^/api/threads/[^/]+/move$")),
# Projects subtree: same enumerated-no-dead-methods precision as the
# threads/runs rules — only the methods the projects router implements
# are admitted, so a future projects route is default-denied until
# explicitly listed. Scope narrowing (projects:read|write|delete and
# threads:write for move) stays enforced by ``@require_permission``.
(frozenset({"GET", "POST"}), re.compile(r"^/api/projects$")),
# ``/config`` is a literal collection route, not a project id — pinned
# explicitly so its admission never depends on the item regex below.
(frozenset({"GET"}), re.compile(r"^/api/projects/config$")),
(frozenset({"GET", "PATCH", "DELETE"}), re.compile(r"^/api/projects/[^/]+$")),
(frozenset({"POST"}), re.compile(r"^/api/projects/[^/]+/(archive|restore)$")),
(frozenset({"GET"}), re.compile(r"^/api/projects/[^/]+/threads$")),
# Project document shelf (Phase 2 Slice B): enumerated per implemented
# route — list/upload on the collection, content and trash on the item.
(frozenset({"GET", "POST"}), re.compile(r"^/api/projects/[^/]+/documents$")),
(frozenset({"GET"}), re.compile(r"^/api/projects/[^/]+/documents/[^/]+/content$")),
(frozenset({"DELETE"}), re.compile(r"^/api/projects/[^/]+/documents/[^/]+$")),
# Promotion and the conversation-files view (Phase 2 Slice C): save a
# thread file to the shelf, attach a shelf document to a thread, and the
# read-only member-thread file aggregation. Scope narrowing
# (projects:write + threads:read/write as decorated) stays enforced by
# ``@require_permission``.
(frozenset({"POST"}), re.compile(r"^/api/projects/[^/]+/documents/from-thread$")),
(frozenset({"POST"}), re.compile(r"^/api/projects/[^/]+/documents/[^/]+/attach-to-thread/[^/]+$")),
(frozenset({"GET"}), re.compile(r"^/api/projects/[^/]+/thread-files$")),
# Trash tier (Phase 2 Slice D): trash listing, restore, per-document
# purge, and empty-trash — enumerated per implemented route, same
# no-dead-methods precision; scope narrowing (projects:read|write|delete)
# stays enforced by ``@require_permission``.
(frozenset({"GET"}), re.compile(r"^/api/trash/documents$")),
(frozenset({"POST"}), re.compile(r"^/api/trash/documents/[^/]+/(restore|purge)$")),
(frozenset({"POST"}), re.compile(r"^/api/trash/purge$")),
# Runs subtree: enumerated per implemented subroute instead of a
# ``runs(/.*)?`` wildcard, so a route added under /runs is default-denied
# until explicitly listed — the same no-dead-methods precision the
# threads collection rule enforces. The ``{run_id}`` slot necessarily
# matches any single segment; the POST-only collection endpoints sharing
# that depth (stream, wait, regenerate, edit-regenerate) are excluded
# from the GET run-id rule so no unimplemented method is pre-authorized.
(frozenset({"GET", "POST"}), re.compile(r"^/api/threads/[^/]+/runs$")),
(
frozenset({"POST"}),
re.compile(r"^/api/threads/[^/]+/runs/(stream|wait|regenerate/prepare|edit-regenerate/prepare)$"),
),
(
frozenset({"GET"}),
re.compile(r"^/api/threads/[^/]+/runs/(?!stream$|wait$|regenerate$|edit-regenerate$)[^/]+$"),
),
(frozenset({"POST"}), re.compile(r"^/api/threads/[^/]+/runs/[^/]+/cancel$")),
(
frozenset({"GET"}),
re.compile(r"^/api/threads/[^/]+/runs/[^/]+/(join|messages|events|workspace-changes)$"),
),
(frozenset({"GET", "POST"}), re.compile(r"^/api/threads/[^/]+/runs/[^/]+/artifacts/archive$")),
(frozenset({"GET", "POST"}), re.compile(r"^/api/threads/[^/]+/runs/[^/]+/stream$")),
(frozenset({"POST"}), re.compile(r"^/api/runs/(stream|wait)$")),
(frozenset({"GET"}), re.compile(r"^/api/runs/[^/]+/(messages|feedback)$")),
)
_BASE62_ALPHABET = "0123456789ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz"
def is_pat_allowed_route(method: str, path: str) -> bool:
"""Return whether the PAT route policy admits *method* + *path*.
Trailing slashes are normalized away so the mounted route and its
redirect-style twin resolve identically.
"""
normalized = path.rstrip("/") or "/"
return any(method in methods and pattern.match(normalized) for methods, pattern in _PAT_ROUTE_RULES)
@functools.cache
def _base62_width(byte_length: int) -> int:
"""Digits sufficient for any *byte_length*-byte value (exact integer math)."""
width = 1
limit = 1 << (8 * byte_length)
while 62**width < limit:
width += 1
return width
def _base62(data: bytes) -> str:
"""Fixed-width big-endian base62 of *data*, ``0``-padded on the left.
``int.from_bytes`` discards leading zero bytes, so an unpadded encoding
would be variable-length (and empty for all-zero input) — any draw below
62**39 would have produced a shorter-than-expected token. The fixed width
keeps every token body exactly ``_base62_width(len(data))`` characters
and makes the format test deterministic.
"""
value = int.from_bytes(data, "big")
digits: list[str] = []
while value:
value, remainder = divmod(value, 62)
digits.append(_BASE62_ALPHABET[remainder])
body = "".join(reversed(digits))
return body.rjust(_base62_width(len(data)), "0")
def generate_pat_token() -> str:
"""Generate a show-once raw token: ``dfp_`` + base62(CSPRNG bytes)."""
return PAT_TOKEN_PREFIX + _base62(secrets.token_bytes(PAT_RANDOM_BYTES))
def pat_token_digest(token: str) -> str:
"""Return the hex SHA-256 digest persisted for *token*."""
return hashlib.sha256(token.encode("utf-8")).hexdigest()
def digest_matches(stored_digest: str | None, token: str) -> bool:
"""Constant-time comparison of *token* against a stored digest."""
if not isinstance(stored_digest, str) or not stored_digest:
return False
return hmac.compare_digest(stored_digest, pat_token_digest(token))
def extract_bearer_token(authorization: str | None) -> str | None:
"""Return the Bearer credential from an Authorization header value.
``None`` means the request carries no Authorization header at all, so the
caller should fall through to the session-cookie path. Any other unusable
value (non-Bearer scheme, empty credential) returns ``""`` so callers
treat it as an invalid credential rather than an absent one.
"""
if authorization is None:
return None
scheme, _, value = authorization.partition(" ")
if scheme.lower() != "bearer":
return ""
return value.strip()
async def authenticate_pat(app: Any, authorization: str | None) -> tuple[Any, frozenset[str]]:
"""Validate the Bearer credential and resolve its owning user.
Returns ``(user, scopes)``. Every token-verdict failure mode — malformed
token, unknown/revoked/expired token, PAT store not configured, missing
owning user — raises the same generic 401 so responses cannot serve as an
oracle on which check failed. Infrastructure errors (store I/O failures)
propagate and fail closed; they are not part of the token verdict.
"""
from fastapi import HTTPException
token = extract_bearer_token(authorization)
if not token or not token.startswith(PAT_TOKEN_PREFIX):
raise HTTPException(status_code=401, detail="Invalid token")
pat_repo = getattr(app.state, "pat_repo", None)
if pat_repo is None:
raise HTTPException(status_code=401, detail="Invalid token")
record = await pat_repo.get_active_by_digest(pat_token_digest(token))
if record is None or not digest_matches(record.get("token_digest"), token):
raise HTTPException(status_code=401, detail="Invalid token")
from app.gateway.deps import get_local_provider
user = await get_local_provider().get_user(str(record["user_id"]))
if user is None:
# The owning user was deleted or became unresolvable; the token is
# dead even though its row survives (deleting a user revokes their
# PATs, without needing a FK cascade).
raise HTTPException(status_code=401, detail="Invalid token")
await pat_repo.touch_last_used(str(record["id"]))
return user, frozenset(record.get("scopes") or ())
def validate_scopes(scopes: list[str]) -> list[str]:
"""Validate a creation-time scope list; returns the deduplicated order."""
unknown = sorted(set(scopes) - PAT_ALLOWED_SCOPES)
if unknown:
raise ValueError(f"Unknown PAT scopes: {', '.join(unknown)}")
deduplicated = sorted(set(scopes))
if not deduplicated:
raise ValueError("A PAT must request at least one scope")
return deduplicated