Files
Alex 795e39a6bc fix: source authorization, silent retrieval failures, and prompt structure
Source access control
---------------------
`active_docs` is client-supplied and reached the retriever unchecked, and the
retriever queries `WHERE source_id = <id>` with no owner predicate — so any
caller could pass any source id to /stream or /api/answer and have another
tenant's documents quoted back, while /api/sources/<id>/search correctly
refused the same id. Gate it through `can_access`, the helper the guarded
endpoints already use, and filter `self.source` down to the authorized set.
Fails closed: no principal, or a check that errors, drops the source.

Three sibling paths had the same gap:

- workflow agent nodes: `AgentNodeConfig.sources` is written verbatim from
  client JSON at save time and nothing validated it, so a node could name any
  tenant's source. Gate against the workflow owner, so shared workflows keep
  reading their owner's sources like shared agents do.
- /api/share: `_resolve_source_pg_id` resolved any id with no ownership
  predicate and baked it into the agent the share creates; /api/search then
  searched it. Authorize before attaching.
- search_service: re-resolve the ids stored on an agent row instead of
  trusting them, so a row written by any future path with the same gap cannot
  be read back.

Team grantees previously lost their source's retrieval config: the post-check
read was still owner-scoped, so it missed and fell back to defaults (an
`agentic_tool` source was bulk-prefetched for every grantee). Read unscoped
after `can_access` passes.

Retrieval
---------
`PGVectorStore._ensure_table_exists` created an IVFFlat index on the empty
table it had just created. IVFFlat computes centroids at build time, so those
centroids were random, and combined with the `source_id` post-filter a source
with hundreds of embedded chunks returned zero rows — retrieval reported no
documents, the model answered from memory, and nothing was logged. Stop
creating the index (exact search is correct and fast well past the sizes most
deployments reach); raise `ivfflat.probes` to sqrt(lists) where an index still
exists; and re-run a short indexed search exactly, since post-filtering means
no index setting can guarantee a full result. `graphrag` had the same
empty-table index with no fallback at all.

Also: bound `chunks` to 0-500 on both the request and agent paths (0 still
means "skip retrieval"), let a source's configured `retrieval.chunks` outrank
the request body, and cap ClassicRAG's per-source floor at
max(top_k, n_sources) so attaching sources cannot inflate the result set.

Silent failures
---------------
An empty retrieval was invisible to both the model and the client: the `source`
event was suppressed when the list was empty, so "searched and found nothing"
looked identical to "no source attached", and the prompt said nothing at all.
Emit the event always, and tell the model when a search ran and returned
nothing. A file that parses to nothing now fails ingest with a message naming
the cause instead of storing an embedding of the empty string. `score_threshold`
returns warnings when the active store or retriever cannot honour it.

Prompt structure
----------------
Retrieved documents move from the system prompt into the user turn, with the
injection guard restated next to them: they change every turn (defeating prefix
caching), they are third-party text that should not carry system authority, and
routing them through the query budget makes them truncatable rather than
silently crowding it out. Documents are shed lowest-ranked-first before the
question is touched.

The six chat presets (3 tones x 2 retrieval modes) differed only in their
Answering section; they are now composed from single-source fragments at load
time, not through Jinja inheritance, which would have opened a file-read
surface in the template sandbox and broken the tool-prefetch parser. Per-tool
guidance moves out of the prompt into tool schemas, so it travels with the tool
and cannot render when the tool is absent. A plain-text custom prompt is staged
as a persona value inside the skeleton instead of replacing it wholesale — it
used to silently lose the injection guard, platform block, memory and
attachments, and its braces are now inert.

Other fixes
-----------
- agents/base: an oversized system prompt drove the query budget negative and
  dispatched a full-price request with an empty question; raise instead.
- llm/anthropic: migrate off the retired Text Completions API. It flattened
  history to first+last message and ignored tools entirely. Adds the missing
  Anthropic handler, without which every tool call was silently dropped.
- sources/upload: `sitemap` had no branch, so every sitemap ingest died on a
  TypeError; `validate_url` now rejects a falsy URL cleanly.
- workflow nodes: retrieved documents never reached the node agent, so a
  classic node with a source and an ordinary prompt answered "I have no
  documents" while the run reported completed.
- parser/bulk: copy the metadata dict, or every chunk reports the last chunk's
  token_count.
- crawler_loader: carry the page title, or citations render the whole chunk
  body as the label.
2026-08-08 10:21:52 +01:00

184 lines
5.0 KiB
Python

"""
URL validation utilities to prevent SSRF (Server-Side Request Forgery) attacks.
This module provides functions to validate URLs before making HTTP requests,
blocking access to internal networks, cloud metadata services, and other
potentially dangerous endpoints.
"""
import ipaddress
import socket
from urllib.parse import urlparse
from typing import Optional, Set
class SSRFError(Exception):
"""Raised when a URL fails SSRF validation."""
pass
# Blocked hostnames that should never be accessed
BLOCKED_HOSTNAMES: Set[str] = {
"localhost",
"localhost.localdomain",
"metadata.google.internal",
"metadata",
}
# Cloud metadata IP addresses (AWS, GCP, Azure, etc.)
METADATA_IPS: Set[str] = {
"169.254.169.254", # AWS, GCP, Azure metadata
"169.254.170.2", # AWS ECS task metadata
"fd00:ec2::254", # AWS IPv6 metadata
}
# Allowed schemes for external requests
ALLOWED_SCHEMES: Set[str] = {"http", "https"}
def is_private_ip(ip_str: str) -> bool:
"""
Check if an IP address is private, loopback, or link-local.
Args:
ip_str: IP address as a string
Returns:
True if the IP is private/internal, False otherwise
"""
try:
ip = ipaddress.ip_address(ip_str)
return (
ip.is_private or
ip.is_loopback or
ip.is_link_local or
ip.is_reserved or
ip.is_multicast or
ip.is_unspecified
)
except ValueError:
# If we can't parse it as an IP, return False
return False
def is_metadata_ip(ip_str: str) -> bool:
"""
Check if an IP address is a cloud metadata service IP.
Args:
ip_str: IP address as a string
Returns:
True if the IP is a metadata service, False otherwise
"""
return ip_str in METADATA_IPS
def resolve_hostname(hostname: str) -> Optional[str]:
"""
Resolve a hostname to an IP address.
Args:
hostname: The hostname to resolve
Returns:
The resolved IP address, or None if resolution fails
"""
try:
return socket.gethostbyname(hostname)
except socket.gaierror:
return None
def validate_url(url: str, allow_localhost: bool = False) -> str:
"""
Validate a URL to prevent SSRF attacks.
This function checks that:
1. The URL has an allowed scheme (http or https)
2. The hostname is not a blocked hostname
3. The resolved IP is not a private/internal IP
4. The resolved IP is not a cloud metadata service
Args:
url: The URL to validate
allow_localhost: If True, allow localhost connections (for testing only)
Returns:
The validated URL (with scheme added if missing)
Raises:
SSRFError: If the URL fails validation
"""
if not url or not isinstance(url, str):
raise SSRFError("No URL was provided.")
# Ensure URL has a scheme
if not urlparse(url).scheme:
url = "http://" + url
parsed = urlparse(url)
# Check scheme
if parsed.scheme not in ALLOWED_SCHEMES:
raise SSRFError(f"URL scheme '{parsed.scheme}' is not allowed. Only HTTP(S) is permitted.")
hostname = parsed.hostname
if not hostname:
raise SSRFError("URL must have a valid hostname.")
hostname_lower = hostname.lower()
# Check blocked hostnames
if hostname_lower in BLOCKED_HOSTNAMES and not allow_localhost:
raise SSRFError(f"Access to '{hostname}' is not allowed.")
# Check if hostname is an IP address directly
try:
ip = ipaddress.ip_address(hostname)
ip_str = str(ip)
if is_metadata_ip(ip_str):
raise SSRFError("Access to cloud metadata services is not allowed.")
if is_private_ip(ip_str) and not allow_localhost:
raise SSRFError("Access to private/internal IP addresses is not allowed.")
return url
except ValueError:
# Not an IP address, it's a hostname - resolve it
pass
# Resolve hostname and check the IP
resolved_ip = resolve_hostname(hostname)
if resolved_ip is None:
raise SSRFError(f"Unable to resolve hostname: {hostname}")
if is_metadata_ip(resolved_ip):
raise SSRFError("Access to cloud metadata services is not allowed.")
if is_private_ip(resolved_ip) and not allow_localhost:
raise SSRFError("Access to private/internal networks is not allowed.")
return url
def validate_url_safe(url: str, allow_localhost: bool = False) -> tuple[bool, str, Optional[str]]:
"""
Validate a URL and return a tuple with validation result.
This is a non-throwing version of validate_url for cases where
you want to handle validation failures gracefully.
Args:
url: The URL to validate
allow_localhost: If True, allow localhost connections (for testing only)
Returns:
Tuple of (is_valid, validated_url_or_original, error_message_or_none)
"""
try:
validated = validate_url(url, allow_localhost)
return (True, validated, None)
except SSRFError as e:
return (False, url, str(e))