fix: Update Lasso plugin to v3 API and support non-chat argument formats#13
Conversation
- Update Lasso API endpoint from v2 to v3 - Add messageType field to API payload (required by v3) - Add _extract_string_values() to recursively extract all string values from nested dicts/lists - Update _extract_messages_from_request() to extract arguments from any MCP tool format, not just chat-style messages array - Remove deprecated license classifier from pyproject.toml - Remove unsupported 'version' argument from FastMCP init This enables Lasso guardrails to work with MCP servers like falcon-mcp that use filter/query style arguments instead of chat messages format. Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
|
Note Other AI code review bot(s) detectedCodeRabbit has detected other AI code review bot(s) in this pull request and will avoid duplicating their findings in the review comments. This may lead to a less comprehensive review. 📝 WalkthroughWalkthroughRemoved explicit version from FastMCP instantiation, upgraded Lasso guardrail to use the v3 classify endpoint with enhanced message extraction (including recursive string extraction and optional message_type in payload), and deleted the MIT license classifier from project metadata. Changes
Sequence Diagram(s)sequenceDiagram
participant Client
participant Guardrail as LassoGuardrailPlugin
participant LassoAPI as External Lasso API
Client->>Guardrail: call guardrail with args / messages
alt messages present
Guardrail->>Guardrail: use provided messages list
else no messages
Guardrail->>Guardrail: _extract_string_values(args) → list of strings
Guardrail->>Guardrail: combine into single user message
end
Guardrail->>LassoAPI: POST /v3/classify (payload includes messageType)
LassoAPI-->>Guardrail: classification response
Guardrail-->>Client: return classification result
Estimated code review effort🎯 3 (Moderate) | ⏱️ ~20 minutes Poem
🚥 Pre-merge checks | ✅ 3✅ Passed checks (3 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. ✨ Finishing touches
📜 Recent review detailsConfiguration used: Organization UI Review profile: ASSERTIVE Plan: Pro 📒 Files selected for processing (2)
🧰 Additional context used🧬 Code graph analysis (1)tests/test_lasso_guardrail.py (1)
🪛 Ruff (0.14.13)mcp_gateway/plugins/guardrails/lasso.py157-157: Dynamically typed expressions (typing.Any) are disallowed in (ANN401) 174-174: Logging statement uses f-string (G004) 204-204: Logging statement uses f-string (G004) ⏰ Context from checks skipped due to timeout of 90000ms. You can increase the timeout in your CodeRabbit configuration to a maximum of 15 minutes (900000ms). (1)
🔇 Additional comments (6)
✏️ Tip: You can disable this entire section by setting Comment |
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 1
🤖 Fix all issues with AI agents
In `@mcp_gateway/plugins/guardrails/lasso.py`:
- Line 38: Update the module/class docstring that still references the v2 Lasso
endpoint so it matches the actual configured API base (self.api_base =
"https://server.lasso.security/gateway/v3/classify"); locate the docstring in
lasso.py that mentions the v2 URL and replace it with the v3 URL and any
corresponding wording (e.g., "gateway/v3/classify") to keep the docstring
consistent with the self.api_base value.
📜 Review details
Configuration used: Organization UI
Review profile: ASSERTIVE
Plan: Pro
📒 Files selected for processing (3)
mcp_gateway/gateway.pymcp_gateway/plugins/guardrails/lasso.pypyproject.toml
💤 Files with no reviewable changes (1)
- pyproject.toml
🧰 Additional context used
🪛 Ruff (0.14.13)
mcp_gateway/plugins/guardrails/lasso.py
159-159: Dynamically typed expressions (typing.Any) are disallowed in obj
(ANN401)
176-176: Logging statement uses f-string
(G004)
202-202: Logging statement uses f-string
(G004)
⏰ Context from checks skipped due to timeout of 90000ms. You can increase the timeout in your CodeRabbit configuration to a maximum of 15 minutes (900000ms). (1)
- GitHub Check: Cursor Bugbot
🔇 Additional comments (4)
mcp_gateway/gateway.py (1)
452-452: LGTM!Removing the unsupported
versionargument from FastMCP initialization is correct. The change aligns with the PR objective.mcp_gateway/plugins/guardrails/lasso.py (3)
159-169: LGTM!The recursive helper correctly extracts all non-empty string values from arbitrary nested structures. The
Anytype annotation is appropriate here since MCP tool arguments can have varying structures. The in-place mutation pattern is efficient for this use case.
193-206: LGTM!The fallback logic to extract all string values from non-chat arguments enables support for MCP tools like falcon-mcp that use filter/query-style arguments. Combining strings into a single user message is a reasonable approach for content scanning purposes.
103-105: Verify messageType parameter handling for response processing against Lasso v3 API specification.The
_prepare_payloadmethod defaults tomessage_type="PROMPT", which is semantically appropriate for request processing. However, at line 320 inprocess_response, assistant-generated content is prepared with the same"PROMPT"default. Verify with Lasso v3 API documentation whether responses should use a different messageType (e.g.,"RESPONSE","COMPLETION", or another value) to properly categorize content based on origin.Also applies to: 320
✏️ Tip: You can disable this entire section by setting review_details to false in your review settings.
![cursor[bot]](https://avatars.githubusercontent.com/in/1210556?s=60&v=4){kind=small}
There was a problem hiding this comment.
Cursor Bugbot has reviewed your changes and found 1 potential issue.
Bugbot Autofix is OFF. To automatically fix reported issues with Cloud Agents, enable Autofix in the Cursor dashboard.
There was a problem hiding this comment.
Actionable comments posted: 0
Caution
Some comments are outside the diff and can’t be posted inline due to platform limitations.
⚠️ Outside diff range comments (1)
tests/test_lasso_guardrail.py (1)
98-115: Missing test coverage for new non-chat argument extraction.The PR adds
_extract_string_values()and updates_extract_messages_from_request()to handle MCP-style argument formats (filter/query-style arguments), but this test only covers the standardmessageslist format. Consider adding test cases for:
- Fallback extraction when
messageskey is absent- Nested dict/list structures with string values
- Mixed argument formats used by MCP servers like falcon-mcp
📝 Suggested additional test cases
def test_extract_messages_from_request_fallback_format() -> None: """Test extracting messages from non-chat MCP argument formats.""" plugin = LassoGuardrailPlugin() # Test filter/query-style arguments without "messages" key arguments: Dict[str, Any] = { "query": "SELECT * FROM users", "filters": { "name": "test_user", "nested": {"value": "nested_string"} } } messages = plugin._extract_messages_from_request(arguments) # Verify string values are extracted from nested structures assert len(messages) > 0 # Verify extracted content contains the string values contents = [m["content"] for m in messages] assert any("SELECT * FROM users" in c for c in contents)
📜 Review details
Configuration used: Organization UI
Review profile: ASSERTIVE
Plan: Pro
📒 Files selected for processing (1)
tests/test_lasso_guardrail.py
⏰ Context from checks skipped due to timeout of 90000ms. You can increase the timeout in your CodeRabbit configuration to a maximum of 15 minutes (900000ms). (1)
- GitHub Check: Cursor Bugbot
🔇 Additional comments (2)
tests/test_lasso_guardrail.py (2)
66-66: LGTM!The test assertion correctly reflects the API endpoint upgrade from v2 to v3, aligning with the changes in
mcp_gateway/plugins/guardrails/lasso.py.
187-189: Verify type compatibility with updated_prepare_payloadsignature.According to the summary,
_prepare_payloadsignature changed to acceptmessages: List[str], but this test passes aList[Dict]. Verify the test is compatible with the updated implementation.#!/bin/bash # Description: Check the actual _prepare_payload signature in the implementation # Expected: Confirm parameter type to verify test compatibility ast-grep --pattern $'def _prepare_payload(self, $_) { $$$ }' # Fallback: search with ripgrep for more context rg -n -A 10 'def _prepare_payload' --type py
✏️ Tip: You can disable this entire section by setting review_details to false in your review settings.
- Remove dead code for non-existent 'outputs' attribute - Fix docstring to reference v3 API instead of v2 - Change verbose INFO logging to DEBUG level - Remove unnecessary 'pass' statements from exception classes - Add comprehensive tests for _extract_string_values()
There was a problem hiding this comment.
Actionable comments posted: 2
Caution
Some comments are outside the diff and can’t be posted inline due to platform limitations.
⚠️ Outside diff range comments (1)
tests/test_lasso_guardrail.py (1)
128-145: Missing test coverage for non-chat argument extraction fallback.The existing
test_extract_messages_from_requestonly tests the chat-style"messages"format. Consider adding a test case for the new fallback path that handles non-chat arguments (e.g., filter/query-style args used by MCP servers like falcon-mcp).📝 Suggested additional test
def test_extract_messages_from_request_fallback() -> None: """Test extracting messages from non-chat request arguments.""" plugin = LassoGuardrailPlugin() # Non-chat style arguments (e.g., filter/query args) arguments: Dict[str, Any] = { "filter": "status=active", "query": "SELECT * FROM users", "nested": {"value": "test"} } messages = plugin._extract_messages_from_request(arguments) assert len(messages) == 1 assert messages[0]["role"] == "user" # All string values should be combined assert "status=active" in messages[0]["content"] assert "SELECT * FROM users" in messages[0]["content"] assert "test" in messages[0]["content"]
🤖 Fix all issues with AI agents
In `@mcp_gateway/plugins/guardrails/lasso.py`:
- Around line 191-203: The fallback branch that builds a single user message by
joining extracted strings loses per-argument context and uses the wrong log
level; update the behavior in the block handling when not messages and arguments
by (1) changing logger.info(...) to logger.debug(...) for verbosity consistency,
and (2) preserve argument context by instead creating messages that include
argument keys (e.g., produce one message per argument or include a mapping of
argument name→value) using the existing _extract_string_values helper to obtain
values and then attaching them to messages (refer to variables messages,
arguments, combined_content and method _extract_string_values to locate the code
to modify).
In `@tests/test_lasso_guardrail.py`:
- Around line 98-125: Add a test asserting that None values are ignored by the
helper: in tests/test_lasso_guardrail.py extend test_extract_string_values to
call plugin._extract_string_values with a structure containing None (e.g., {"a":
None, "b": "text"}) and assert the resulting values list contains only the
string entries; reference the helper by name _extract_string_values and ensure
the new assertion follows the existing patterns for values initialization and
equality checks.
♻️ Duplicate comments (1)
mcp_gateway/plugins/guardrails/lasso.py (1)
99-103:messageTypedefaults to "PROMPT" for response validation.The
process_responsemethod at line 308 calls_prepare_payload(messages)without specifyingmessage_type, causing responses to be validated with"PROMPT"type. If the v3 API distinguishes between prompt and response validation, this could apply incorrect rules to assistant outputs.Consider passing an appropriate message type for response validation:
# Prepare and make API call headers = self._prepare_headers() - payload = self._prepare_payload(messages) + payload = self._prepare_payload(messages, message_type="RESPONSE") response = await self._call_lasso_api(headers, payload)Lasso Security API v3 messageType field values for classify endpointAlso applies to: 306-309
📜 Review details
Configuration used: Organization UI
Review profile: ASSERTIVE
Plan: Pro
📒 Files selected for processing (2)
mcp_gateway/plugins/guardrails/lasso.pytests/test_lasso_guardrail.py
🧰 Additional context used
🧬 Code graph analysis (1)
tests/test_lasso_guardrail.py (1)
mcp_gateway/plugins/guardrails/lasso.py (2)
LassoGuardrailPlugin(22-334)_extract_string_values(157-167)
🪛 Ruff (0.14.13)
mcp_gateway/plugins/guardrails/lasso.py
157-157: Dynamically typed expressions (typing.Any) are disallowed in obj
(ANN401)
174-174: Logging statement uses f-string
(G004)
201-201: Logging statement uses f-string
(G004)
⏰ Context from checks skipped due to timeout of 90000ms. You can increase the timeout in your CodeRabbit configuration to a maximum of 15 minutes (900000ms). (1)
- GitHub Check: Cursor Bugbot
🔇 Additional comments (6)
mcp_gateway/plugins/guardrails/lasso.py (5)
34-34: Docstring now correctly references v3 API.The API base URL and documentation have been consistently updated to reference the v3 endpoint.
Also applies to: 45-45
157-167: Recursive string extraction implementation is correct.The helper correctly handles nested structures. The
Anytype annotation (flagged by Ruff ANN401) is appropriate here since the function is designed to process arbitrary nested data structures.
209-229: Response text extraction is well-structured.The simplified implementation correctly handles
CallToolResultwithTextContentand uses appropriate logging levels.
231-278: Request processing logic is sound.The fail-open behavior on unexpected errors and the use of
"PROMPT"message type for request validation are appropriate.
280-334: Response processing implementation is correct.The fail-open behavior and error response construction are appropriate. The
messageTypeconcern was noted above.tests/test_lasso_guardrail.py (1)
61-66: LGTM!Test correctly updated to verify the v3 API endpoint.
✏️ Tip: You can disable this entire section by setting review_details to false in your review settings.
- Include argument keys in extracted content (e.g., 'filter: value') - Change log level from INFO to DEBUG for consistency - Add test for fallback extraction behavior
2 participants
This enables Lasso guardrails to work with MCP servers like falcon-mcp that use filter/query style arguments instead of chat messages format.
Note
Lasso guardrail updates
v3and includemessageTypein request payloads_extract_string_valuesand enhance_extract_messages_from_requestto support non-chat argument formats (e.g.,filter/query), with improved debug loggingcontentand removing deprecatedoutputshandlingGateway and project maintenance
versionargument fromFastMCPinitialization ingateway.pypyproject.tomlclassifiersTests
Written by Cursor Bugbot for commit 97b3d0d. This will update automatically on new commits. Configure here.
Summary by CodeRabbit
✏️ Tip: You can customize this high-level summary in your review settings.