Files
hermes-agent/tests/tools/test_website_policy.py

513 lines
17 KiB
Python
Raw Permalink Normal View History

import json
from pathlib import Path
import pytest
import yaml
from tools.website_policy import WebsitePolicyError, check_website_access, load_website_blocklist
def test_load_website_blocklist_merges_config_and_shared_file(tmp_path):
shared = tmp_path / "community-blocklist.txt"
shared.write_text("# comment\nexample.org\nsub.bad.net\n", encoding="utf-8")
config_path = tmp_path / "config.yaml"
config_path.write_text(
yaml.safe_dump(
{
"security": {
"website_blocklist": {
"enabled": True,
"domains": ["example.com", "https://www.evil.test/path"],
"shared_files": [str(shared)],
}
}
},
sort_keys=False,
),
encoding="utf-8",
)
policy = load_website_blocklist(config_path)
assert policy["enabled"] is True
assert {rule["pattern"] for rule in policy["rules"]} == {
"example.com",
"evil.test",
"example.org",
"sub.bad.net",
}
def test_check_website_access_matches_parent_domain_subdomains(tmp_path):
config_path = tmp_path / "config.yaml"
config_path.write_text(
yaml.safe_dump(
{
"security": {
"website_blocklist": {
"enabled": True,
"domains": ["example.com"],
}
}
},
sort_keys=False,
),
encoding="utf-8",
)
blocked = check_website_access("https://docs.example.com/page", config_path=config_path)
assert blocked is not None
assert blocked["host"] == "docs.example.com"
assert blocked["rule"] == "example.com"
def test_check_website_access_supports_wildcard_subdomains_only(tmp_path):
config_path = tmp_path / "config.yaml"
config_path.write_text(
yaml.safe_dump(
{
"security": {
"website_blocklist": {
"enabled": True,
"domains": ["*.tracking.example"],
}
}
},
sort_keys=False,
),
encoding="utf-8",
)
assert check_website_access("https://a.tracking.example", config_path=config_path) is not None
assert check_website_access("https://www.tracking.example", config_path=config_path) is not None
assert check_website_access("https://tracking.example", config_path=config_path) is None
def test_default_config_exposes_website_blocklist_shape():
from hermes_cli.config import DEFAULT_CONFIG
website_blocklist = DEFAULT_CONFIG["security"]["website_blocklist"]
assert website_blocklist["enabled"] is False
assert website_blocklist["domains"] == []
assert website_blocklist["shared_files"] == []
def test_load_website_blocklist_uses_enabled_default_when_section_missing(tmp_path):
config_path = tmp_path / "config.yaml"
config_path.write_text(yaml.safe_dump({"display": {"tool_progress": "all"}}, sort_keys=False), encoding="utf-8")
policy = load_website_blocklist(config_path)
assert policy == {"enabled": False, "rules": []}
def test_load_website_blocklist_raises_clean_error_for_invalid_domains_type(tmp_path):
config_path = tmp_path / "config.yaml"
config_path.write_text(
yaml.safe_dump(
{
"security": {
"website_blocklist": {
"enabled": True,
"domains": "example.com",
}
}
},
sort_keys=False,
),
encoding="utf-8",
)
with pytest.raises(WebsitePolicyError, match="security.website_blocklist.domains must be a list"):
load_website_blocklist(config_path)
def test_load_website_blocklist_raises_clean_error_for_invalid_shared_files_type(tmp_path):
config_path = tmp_path / "config.yaml"
config_path.write_text(
yaml.safe_dump(
{
"security": {
"website_blocklist": {
"enabled": True,
"shared_files": "community-blocklist.txt",
}
}
},
sort_keys=False,
),
encoding="utf-8",
)
with pytest.raises(WebsitePolicyError, match="security.website_blocklist.shared_files must be a list"):
load_website_blocklist(config_path)
def test_load_website_blocklist_raises_clean_error_for_invalid_top_level_config_type(tmp_path):
config_path = tmp_path / "config.yaml"
config_path.write_text(yaml.safe_dump(["not", "a", "mapping"], sort_keys=False), encoding="utf-8")
with pytest.raises(WebsitePolicyError, match="config root must be a mapping"):
load_website_blocklist(config_path)
def test_load_website_blocklist_raises_clean_error_for_invalid_security_type(tmp_path):
config_path = tmp_path / "config.yaml"
config_path.write_text(yaml.safe_dump({"security": []}, sort_keys=False), encoding="utf-8")
with pytest.raises(WebsitePolicyError, match="security must be a mapping"):
load_website_blocklist(config_path)
def test_load_website_blocklist_raises_clean_error_for_invalid_website_blocklist_type(tmp_path):
config_path = tmp_path / "config.yaml"
config_path.write_text(
yaml.safe_dump(
{
"security": {
"website_blocklist": "block everything",
}
},
sort_keys=False,
),
encoding="utf-8",
)
with pytest.raises(WebsitePolicyError, match="security.website_blocklist must be a mapping"):
load_website_blocklist(config_path)
def test_load_website_blocklist_raises_clean_error_for_invalid_enabled_type(tmp_path):
config_path = tmp_path / "config.yaml"
config_path.write_text(
yaml.safe_dump(
{
"security": {
"website_blocklist": {
"enabled": "false",
}
}
},
sort_keys=False,
),
encoding="utf-8",
)
with pytest.raises(WebsitePolicyError, match="security.website_blocklist.enabled must be a boolean"):
load_website_blocklist(config_path)
def test_load_website_blocklist_raises_clean_error_for_malformed_yaml(tmp_path):
config_path = tmp_path / "config.yaml"
config_path.write_text("security: [oops\n", encoding="utf-8")
with pytest.raises(WebsitePolicyError, match="Invalid config YAML"):
load_website_blocklist(config_path)
def test_load_website_blocklist_wraps_shared_file_read_errors(tmp_path, monkeypatch):
shared = tmp_path / "community-blocklist.txt"
shared.write_text("example.org\n", encoding="utf-8")
config_path = tmp_path / "config.yaml"
config_path.write_text(
yaml.safe_dump(
{
"security": {
"website_blocklist": {
"enabled": True,
"shared_files": [str(shared)],
}
}
},
sort_keys=False,
),
encoding="utf-8",
)
def failing_read_text(self, *args, **kwargs):
raise PermissionError("no permission")
monkeypatch.setattr(Path, "read_text", failing_read_text)
# Unreadable shared files are now warned and skipped (not raised),
# so the blocklist loads successfully but without those rules.
result = load_website_blocklist(config_path)
assert result["enabled"] is True
assert result["rules"] == [] # shared file rules skipped
def test_check_website_access_uses_dynamic_hermes_home(monkeypatch, tmp_path):
hermes_home = tmp_path / "hermes-home"
hermes_home.mkdir()
(hermes_home / "config.yaml").write_text(
yaml.safe_dump(
{
"security": {
"website_blocklist": {
"enabled": True,
"domains": ["dynamic.example"],
}
}
},
sort_keys=False,
),
encoding="utf-8",
)
monkeypatch.setenv("HERMES_HOME", str(hermes_home))
# Invalidate the module-level cache so the new HERMES_HOME is picked up.
# A prior test may have cached a default policy (enabled=False) under the
# old HERMES_HOME set by the autouse _isolate_hermes_home fixture.
from tools.website_policy import invalidate_cache
invalidate_cache()
blocked = check_website_access("https://dynamic.example/path")
assert blocked is not None
assert blocked["rule"] == "dynamic.example"
def test_check_website_access_blocks_scheme_less_urls(tmp_path):
config_path = tmp_path / "config.yaml"
config_path.write_text(
yaml.safe_dump(
{
"security": {
"website_blocklist": {
"enabled": True,
"domains": ["blocked.test"],
}
}
},
sort_keys=False,
),
encoding="utf-8",
)
blocked = check_website_access("www.blocked.test/path", config_path=config_path)
assert blocked is not None
assert blocked["host"] == "www.blocked.test"
assert blocked["rule"] == "blocked.test"
def test_browser_navigate_returns_policy_block(monkeypatch):
from tools import browser_tool
# Allow SSRF check to pass so the policy check is reached
monkeypatch.setattr(browser_tool, "_is_safe_url", lambda url: True)
monkeypatch.setattr(
browser_tool,
"check_website_access",
lambda url: {
"host": "blocked.test",
"rule": "blocked.test",
"source": "config",
"message": "Blocked by website policy",
},
)
monkeypatch.setattr(
browser_tool,
"_run_browser_command",
lambda *args, **kwargs: pytest.fail("browser command should not run for blocked URL"),
)
result = json.loads(browser_tool.browser_navigate("https://blocked.test"))
assert result["success"] is False
assert result["blocked_by_policy"]["rule"] == "blocked.test"
def test_browser_navigate_allows_when_shared_file_missing(monkeypatch, tmp_path):
"""Missing shared blocklist files are warned and skipped, not fatal."""
from tools import browser_tool
config_path = tmp_path / "config.yaml"
config_path.write_text(
yaml.safe_dump(
{
"security": {
"website_blocklist": {
"enabled": True,
"shared_files": ["missing-blocklist.txt"],
}
}
},
sort_keys=False,
),
encoding="utf-8",
)
# check_website_access should return None (allow) — missing file is skipped
result = check_website_access("https://allowed.test", config_path=config_path)
assert result is None
@pytest.mark.asyncio
async def test_web_extract_short_circuits_blocked_url(monkeypatch):
from tools import web_tools
fix(security): add SSRF protection to vision_tools and web_tools (hardened) * fix(security): add SSRF protection to vision_tools and web_tools Both vision_analyze and web_extract/web_crawl accept arbitrary URLs without checking if they target private/internal network addresses. A prompt-injected or malicious skill could use this to access cloud metadata endpoints (169.254.169.254), localhost services, or private network hosts. Adds a shared url_safety.is_safe_url() that resolves hostnames and blocks private, loopback, link-local, and reserved IP ranges. Also blocks known internal hostnames (metadata.google.internal). Integrated at the URL validation layer in vision_tools and before each website_policy check in web_tools (extract, crawl). * test(vision): update localhost test to reflect SSRF protection The existing test_valid_url_with_port asserted localhost URLs pass validation. With SSRF protection, localhost is now correctly blocked. Update the test to verify the block, and add a separate test for valid URLs with ports using a public hostname. * fix(security): harden SSRF protection — fail-closed, CGNAT, multicast, redirect guard Follow-up hardening on top of dieutx's SSRF protection (PR #2630): - Change fail-open to fail-closed: DNS errors and unexpected exceptions now block the request instead of allowing it (OWASP best practice) - Block CGNAT range (100.64.0.0/10): Python's ipaddress.is_private does NOT cover this range (returns False for both is_private and is_global). Used by Tailscale/WireGuard and carrier infrastructure. - Add is_multicast and is_unspecified checks: multicast (224.0.0.0/4) and unspecified (0.0.0.0) addresses were not caught by the original four-check chain - Add redirect guard for vision_tools: httpx event hook re-validates each redirect target against SSRF checks, preventing the classic redirect-based SSRF bypass (302 to internal IP) - Move SSRF filtering before backend dispatch in web_extract: now covers Parallel and Tavily backends, not just Firecrawl - Extract _is_blocked_ip() helper for cleaner IP range checking - Add 24 new tests (CGNAT, multicast, IPv4-mapped IPv6, fail-closed behavior, parametrized blocked/allowed IP lists) - Fix existing tests to mock DNS resolution for test hostnames --------- Co-authored-by: dieutx <dangtc94@gmail.com>
2026-03-23 15:40:42 -07:00
# Allow test URLs past SSRF check so website policy is what gets tested
monkeypatch.setattr(web_tools, "is_safe_url", lambda url: True)
monkeypatch.setattr(
web_tools,
"check_website_access",
lambda url: {
"host": "blocked.test",
"rule": "blocked.test",
"source": "config",
"message": "Blocked by website policy",
},
)
monkeypatch.setattr(
web_tools,
"_get_firecrawl_client",
lambda: pytest.fail("firecrawl should not run for blocked URL"),
)
monkeypatch.setattr("tools.interrupt.is_interrupted", lambda: False)
result = json.loads(await web_tools.web_extract_tool(["https://blocked.test"], use_llm_processing=False))
assert result["results"][0]["url"] == "https://blocked.test"
assert "Blocked by website policy" in result["results"][0]["error"]
def test_check_website_access_fails_open_on_malformed_config(tmp_path, monkeypatch):
"""Malformed config with default path should fail open (return None), not crash."""
config_path = tmp_path / "config.yaml"
config_path.write_text("security: [oops\n", encoding="utf-8")
# With explicit config_path (test mode), errors propagate
with pytest.raises(WebsitePolicyError):
check_website_access("https://example.com", config_path=config_path)
# Simulate default path by pointing HERMES_HOME to tmp_path
monkeypatch.setenv("HERMES_HOME", str(tmp_path))
from tools import website_policy
website_policy.invalidate_cache()
# With default path, errors are caught and fail open
result = check_website_access("https://example.com")
assert result is None # allowed, not crashed
@pytest.mark.asyncio
async def test_web_extract_blocks_redirected_final_url(monkeypatch):
from tools import web_tools
fix(security): add SSRF protection to vision_tools and web_tools (hardened) * fix(security): add SSRF protection to vision_tools and web_tools Both vision_analyze and web_extract/web_crawl accept arbitrary URLs without checking if they target private/internal network addresses. A prompt-injected or malicious skill could use this to access cloud metadata endpoints (169.254.169.254), localhost services, or private network hosts. Adds a shared url_safety.is_safe_url() that resolves hostnames and blocks private, loopback, link-local, and reserved IP ranges. Also blocks known internal hostnames (metadata.google.internal). Integrated at the URL validation layer in vision_tools and before each website_policy check in web_tools (extract, crawl). * test(vision): update localhost test to reflect SSRF protection The existing test_valid_url_with_port asserted localhost URLs pass validation. With SSRF protection, localhost is now correctly blocked. Update the test to verify the block, and add a separate test for valid URLs with ports using a public hostname. * fix(security): harden SSRF protection — fail-closed, CGNAT, multicast, redirect guard Follow-up hardening on top of dieutx's SSRF protection (PR #2630): - Change fail-open to fail-closed: DNS errors and unexpected exceptions now block the request instead of allowing it (OWASP best practice) - Block CGNAT range (100.64.0.0/10): Python's ipaddress.is_private does NOT cover this range (returns False for both is_private and is_global). Used by Tailscale/WireGuard and carrier infrastructure. - Add is_multicast and is_unspecified checks: multicast (224.0.0.0/4) and unspecified (0.0.0.0) addresses were not caught by the original four-check chain - Add redirect guard for vision_tools: httpx event hook re-validates each redirect target against SSRF checks, preventing the classic redirect-based SSRF bypass (302 to internal IP) - Move SSRF filtering before backend dispatch in web_extract: now covers Parallel and Tavily backends, not just Firecrawl - Extract _is_blocked_ip() helper for cleaner IP range checking - Add 24 new tests (CGNAT, multicast, IPv4-mapped IPv6, fail-closed behavior, parametrized blocked/allowed IP lists) - Fix existing tests to mock DNS resolution for test hostnames --------- Co-authored-by: dieutx <dangtc94@gmail.com>
2026-03-23 15:40:42 -07:00
# Allow test URLs past SSRF check so website policy is what gets tested
monkeypatch.setattr(web_tools, "is_safe_url", lambda url: True)
def fake_check(url):
if url == "https://allowed.test":
return None
if url == "https://blocked.test/final":
return {
"host": "blocked.test",
"rule": "blocked.test",
"source": "config",
"message": "Blocked by website policy",
}
pytest.fail(f"unexpected URL checked: {url}")
class FakeFirecrawlClient:
def scrape(self, url, formats):
return {
"markdown": "secret content",
"metadata": {
"title": "Redirected",
"sourceURL": "https://blocked.test/final",
},
}
monkeypatch.setattr(web_tools, "check_website_access", fake_check)
monkeypatch.setattr(web_tools, "_get_firecrawl_client", lambda: FakeFirecrawlClient())
monkeypatch.setattr("tools.interrupt.is_interrupted", lambda: False)
result = json.loads(await web_tools.web_extract_tool(["https://allowed.test"], use_llm_processing=False))
assert result["results"][0]["url"] == "https://blocked.test/final"
assert result["results"][0]["content"] == ""
assert result["results"][0]["blocked_by_policy"]["rule"] == "blocked.test"
@pytest.mark.asyncio
async def test_web_crawl_short_circuits_blocked_url(monkeypatch):
from tools import web_tools
feat(web): add Parallel as alternative web search/extract backend (#1696) * feat(web): add Parallel as alternative web search/extract backend Adds Parallel (parallel.ai) as a drop-in alternative to Firecrawl for web_search and web_extract tools using the official parallel-web SDK. - Backend selection via WEB_SEARCH_BACKEND env var (auto/parallel/firecrawl) - Auto mode prefers Firecrawl when both keys present; Parallel when sole backend - web_crawl remains Firecrawl-only with clear error when unavailable - Lazy SDK imports, interrupt support, singleton clients - 16 new unit tests for backend selection and client config Co-authored-by: s-jag <s-jag@users.noreply.github.com> * fix: add PARALLEL_API_KEY to config registry and fix web_crawl policy tests Follow-up for Parallel backend integration: - Add PARALLEL_API_KEY to OPTIONAL_ENV_VARS (hermes doctor, env blocklist) - Add to set_config_value api_keys list (hermes config set) - Add to doctor keys display - Fix 2 web_crawl policy tests that didn't set FIRECRAWL_API_KEY (needed now that web_crawl has a Firecrawl availability guard) * refactor: explicit backend selection via hermes tools, not auto-detect Replace the auto-detect backend selection with explicit user choice: - hermes tools saves WEB_SEARCH_BACKEND to .env when user picks a provider - _get_backend() reads the explicit choice first - Fallback only for manual/legacy config (uses whichever key is present) - _is_provider_active() shows [active] for the selected web backend - Updated tests, docs, and .env.example to remove 'auto' mode language * refactor: use config.yaml for web backend, not env var Match the TTS/browser pattern — web.backend is stored in config.yaml (set by hermes tools), not as a WEB_SEARCH_BACKEND env var. - _load_web_config() reads web: section from config.yaml - _get_backend() reads web.backend from config, falls back to key detection - _configure_provider() saves to config dict (saved to config.yaml) - _is_provider_active() reads from config dict - Removed WEB_SEARCH_BACKEND from .env.example, set_config_value, docs - Updated all tests to mock _load_web_config instead of env vars --------- Co-authored-by: s-jag <s-jag@users.noreply.github.com>
2026-03-17 04:02:02 -07:00
# web_crawl_tool checks for Firecrawl env before website policy
monkeypatch.setenv("FIRECRAWL_API_KEY", "fake-key")
fix(security): add SSRF protection to vision_tools and web_tools (hardened) * fix(security): add SSRF protection to vision_tools and web_tools Both vision_analyze and web_extract/web_crawl accept arbitrary URLs without checking if they target private/internal network addresses. A prompt-injected or malicious skill could use this to access cloud metadata endpoints (169.254.169.254), localhost services, or private network hosts. Adds a shared url_safety.is_safe_url() that resolves hostnames and blocks private, loopback, link-local, and reserved IP ranges. Also blocks known internal hostnames (metadata.google.internal). Integrated at the URL validation layer in vision_tools and before each website_policy check in web_tools (extract, crawl). * test(vision): update localhost test to reflect SSRF protection The existing test_valid_url_with_port asserted localhost URLs pass validation. With SSRF protection, localhost is now correctly blocked. Update the test to verify the block, and add a separate test for valid URLs with ports using a public hostname. * fix(security): harden SSRF protection — fail-closed, CGNAT, multicast, redirect guard Follow-up hardening on top of dieutx's SSRF protection (PR #2630): - Change fail-open to fail-closed: DNS errors and unexpected exceptions now block the request instead of allowing it (OWASP best practice) - Block CGNAT range (100.64.0.0/10): Python's ipaddress.is_private does NOT cover this range (returns False for both is_private and is_global). Used by Tailscale/WireGuard and carrier infrastructure. - Add is_multicast and is_unspecified checks: multicast (224.0.0.0/4) and unspecified (0.0.0.0) addresses were not caught by the original four-check chain - Add redirect guard for vision_tools: httpx event hook re-validates each redirect target against SSRF checks, preventing the classic redirect-based SSRF bypass (302 to internal IP) - Move SSRF filtering before backend dispatch in web_extract: now covers Parallel and Tavily backends, not just Firecrawl - Extract _is_blocked_ip() helper for cleaner IP range checking - Add 24 new tests (CGNAT, multicast, IPv4-mapped IPv6, fail-closed behavior, parametrized blocked/allowed IP lists) - Fix existing tests to mock DNS resolution for test hostnames --------- Co-authored-by: dieutx <dangtc94@gmail.com>
2026-03-23 15:40:42 -07:00
# Allow test URLs past SSRF check so website policy is what gets tested
monkeypatch.setattr(web_tools, "is_safe_url", lambda url: True)
monkeypatch.setattr(
web_tools,
"check_website_access",
lambda url: {
"host": "blocked.test",
"rule": "blocked.test",
"source": "config",
"message": "Blocked by website policy",
},
)
monkeypatch.setattr(
web_tools,
"_get_firecrawl_client",
lambda: pytest.fail("firecrawl should not run for blocked crawl URL"),
)
monkeypatch.setattr("tools.interrupt.is_interrupted", lambda: False)
result = json.loads(await web_tools.web_crawl_tool("https://blocked.test", use_llm_processing=False))
assert result["results"][0]["url"] == "https://blocked.test"
assert result["results"][0]["blocked_by_policy"]["rule"] == "blocked.test"
@pytest.mark.asyncio
async def test_web_crawl_blocks_redirected_final_url(monkeypatch):
from tools import web_tools
feat(web): add Parallel as alternative web search/extract backend (#1696) * feat(web): add Parallel as alternative web search/extract backend Adds Parallel (parallel.ai) as a drop-in alternative to Firecrawl for web_search and web_extract tools using the official parallel-web SDK. - Backend selection via WEB_SEARCH_BACKEND env var (auto/parallel/firecrawl) - Auto mode prefers Firecrawl when both keys present; Parallel when sole backend - web_crawl remains Firecrawl-only with clear error when unavailable - Lazy SDK imports, interrupt support, singleton clients - 16 new unit tests for backend selection and client config Co-authored-by: s-jag <s-jag@users.noreply.github.com> * fix: add PARALLEL_API_KEY to config registry and fix web_crawl policy tests Follow-up for Parallel backend integration: - Add PARALLEL_API_KEY to OPTIONAL_ENV_VARS (hermes doctor, env blocklist) - Add to set_config_value api_keys list (hermes config set) - Add to doctor keys display - Fix 2 web_crawl policy tests that didn't set FIRECRAWL_API_KEY (needed now that web_crawl has a Firecrawl availability guard) * refactor: explicit backend selection via hermes tools, not auto-detect Replace the auto-detect backend selection with explicit user choice: - hermes tools saves WEB_SEARCH_BACKEND to .env when user picks a provider - _get_backend() reads the explicit choice first - Fallback only for manual/legacy config (uses whichever key is present) - _is_provider_active() shows [active] for the selected web backend - Updated tests, docs, and .env.example to remove 'auto' mode language * refactor: use config.yaml for web backend, not env var Match the TTS/browser pattern — web.backend is stored in config.yaml (set by hermes tools), not as a WEB_SEARCH_BACKEND env var. - _load_web_config() reads web: section from config.yaml - _get_backend() reads web.backend from config, falls back to key detection - _configure_provider() saves to config dict (saved to config.yaml) - _is_provider_active() reads from config dict - Removed WEB_SEARCH_BACKEND from .env.example, set_config_value, docs - Updated all tests to mock _load_web_config instead of env vars --------- Co-authored-by: s-jag <s-jag@users.noreply.github.com>
2026-03-17 04:02:02 -07:00
# web_crawl_tool checks for Firecrawl env before website policy
monkeypatch.setenv("FIRECRAWL_API_KEY", "fake-key")
fix(security): add SSRF protection to vision_tools and web_tools (hardened) * fix(security): add SSRF protection to vision_tools and web_tools Both vision_analyze and web_extract/web_crawl accept arbitrary URLs without checking if they target private/internal network addresses. A prompt-injected or malicious skill could use this to access cloud metadata endpoints (169.254.169.254), localhost services, or private network hosts. Adds a shared url_safety.is_safe_url() that resolves hostnames and blocks private, loopback, link-local, and reserved IP ranges. Also blocks known internal hostnames (metadata.google.internal). Integrated at the URL validation layer in vision_tools and before each website_policy check in web_tools (extract, crawl). * test(vision): update localhost test to reflect SSRF protection The existing test_valid_url_with_port asserted localhost URLs pass validation. With SSRF protection, localhost is now correctly blocked. Update the test to verify the block, and add a separate test for valid URLs with ports using a public hostname. * fix(security): harden SSRF protection — fail-closed, CGNAT, multicast, redirect guard Follow-up hardening on top of dieutx's SSRF protection (PR #2630): - Change fail-open to fail-closed: DNS errors and unexpected exceptions now block the request instead of allowing it (OWASP best practice) - Block CGNAT range (100.64.0.0/10): Python's ipaddress.is_private does NOT cover this range (returns False for both is_private and is_global). Used by Tailscale/WireGuard and carrier infrastructure. - Add is_multicast and is_unspecified checks: multicast (224.0.0.0/4) and unspecified (0.0.0.0) addresses were not caught by the original four-check chain - Add redirect guard for vision_tools: httpx event hook re-validates each redirect target against SSRF checks, preventing the classic redirect-based SSRF bypass (302 to internal IP) - Move SSRF filtering before backend dispatch in web_extract: now covers Parallel and Tavily backends, not just Firecrawl - Extract _is_blocked_ip() helper for cleaner IP range checking - Add 24 new tests (CGNAT, multicast, IPv4-mapped IPv6, fail-closed behavior, parametrized blocked/allowed IP lists) - Fix existing tests to mock DNS resolution for test hostnames --------- Co-authored-by: dieutx <dangtc94@gmail.com>
2026-03-23 15:40:42 -07:00
# Allow test URLs past SSRF check so website policy is what gets tested
monkeypatch.setattr(web_tools, "is_safe_url", lambda url: True)
feat(web): add Parallel as alternative web search/extract backend (#1696) * feat(web): add Parallel as alternative web search/extract backend Adds Parallel (parallel.ai) as a drop-in alternative to Firecrawl for web_search and web_extract tools using the official parallel-web SDK. - Backend selection via WEB_SEARCH_BACKEND env var (auto/parallel/firecrawl) - Auto mode prefers Firecrawl when both keys present; Parallel when sole backend - web_crawl remains Firecrawl-only with clear error when unavailable - Lazy SDK imports, interrupt support, singleton clients - 16 new unit tests for backend selection and client config Co-authored-by: s-jag <s-jag@users.noreply.github.com> * fix: add PARALLEL_API_KEY to config registry and fix web_crawl policy tests Follow-up for Parallel backend integration: - Add PARALLEL_API_KEY to OPTIONAL_ENV_VARS (hermes doctor, env blocklist) - Add to set_config_value api_keys list (hermes config set) - Add to doctor keys display - Fix 2 web_crawl policy tests that didn't set FIRECRAWL_API_KEY (needed now that web_crawl has a Firecrawl availability guard) * refactor: explicit backend selection via hermes tools, not auto-detect Replace the auto-detect backend selection with explicit user choice: - hermes tools saves WEB_SEARCH_BACKEND to .env when user picks a provider - _get_backend() reads the explicit choice first - Fallback only for manual/legacy config (uses whichever key is present) - _is_provider_active() shows [active] for the selected web backend - Updated tests, docs, and .env.example to remove 'auto' mode language * refactor: use config.yaml for web backend, not env var Match the TTS/browser pattern — web.backend is stored in config.yaml (set by hermes tools), not as a WEB_SEARCH_BACKEND env var. - _load_web_config() reads web: section from config.yaml - _get_backend() reads web.backend from config, falls back to key detection - _configure_provider() saves to config dict (saved to config.yaml) - _is_provider_active() reads from config dict - Removed WEB_SEARCH_BACKEND from .env.example, set_config_value, docs - Updated all tests to mock _load_web_config instead of env vars --------- Co-authored-by: s-jag <s-jag@users.noreply.github.com>
2026-03-17 04:02:02 -07:00
def fake_check(url):
if url == "https://allowed.test":
return None
if url == "https://blocked.test/final":
return {
"host": "blocked.test",
"rule": "blocked.test",
"source": "config",
"message": "Blocked by website policy",
}
pytest.fail(f"unexpected URL checked: {url}")
class FakeCrawlClient:
def crawl(self, url, **kwargs):
return {
"data": [
{
"markdown": "secret crawl content",
"metadata": {
"title": "Redirected crawl page",
"sourceURL": "https://blocked.test/final",
},
}
]
}
monkeypatch.setattr(web_tools, "check_website_access", fake_check)
monkeypatch.setattr(web_tools, "_get_firecrawl_client", lambda: FakeCrawlClient())
monkeypatch.setattr("tools.interrupt.is_interrupted", lambda: False)
result = json.loads(await web_tools.web_crawl_tool("https://allowed.test", use_llm_processing=False))
assert result["results"][0]["content"] == ""
assert result["results"][0]["error"] == "Blocked by website policy"
assert result["results"][0]["blocked_by_policy"]["rule"] == "blocked.test"