fail-kit-lsp 2.0.1

F.A.I.L. Kit Language Server for Python - Forensic analysis for LangChain, CrewAI, and AutoGen

F.A.I.L. Kit Python Language Server

Language Server Protocol (LSP) implementation for Python AI agent frameworks.

Features

• LangChain Support: Analyze LangChain agents, tools, and chains
• CrewAI Support: Analyze CrewAI agents, tasks, and crews
• AutoGen Support: Analyze AutoGen agents and group chats
• Real-time Diagnostics: Issues appear as you type
• Quick Fixes: One-click fixes for common issues
• Hover Information: Detailed issue descriptions on hover

Installation

pip install fail-kit-lsp

Usage

VS Code

Install the F.A.I.L. Kit VS Code extension, which automatically uses this LSP for Python files.

Neovim (nvim-lspconfig)

require('lspconfig').fail_kit_lsp.setup{}

Other Editors

Start the language server:

fail-kit-lsp

The server communicates over stdin/stdout using the Language Server Protocol.

Detected Issues

Rule	Category	Description
FK001	Audit	Tool call without receipt generation
FK002	Reliability	LLM call without error handling
FK003	Security	Hardcoded secrets in code
FK004	Safety	Destructive operation without confirmation
FK005	Resilience	LLM call without retry/timeout
FK006	Audit	Missing provenance metadata

Example

# This code will trigger FK001 and FK002 warnings

from langchain.agents import AgentExecutor
from langchain_openai import ChatOpenAI

llm = ChatOpenAI()  # FK002: No error handling

# FK001: No receipt generation
result = agent_executor.invoke({"input": "delete all files"})

With F.A.I.L. Kit fixes:

from langchain.agents import AgentExecutor
from langchain_openai import ChatOpenAI
from fail_kit import generate_receipt, audit_logger

llm = ChatOpenAI()

try:
    result = agent_executor.invoke({"input": "delete all files"})
    
    receipt = generate_receipt(
        tool_name="agent_executor",
        input_data={"input": "delete all files"},
        result=result,
    )
    audit_logger.log_action(receipt)
    
except Exception as e:
    audit_logger.log_failure({
        "error_type": type(e).__name__,
        "error_message": str(e),
    })
    raise

Configuration

Create a fail-kit.toml in your project root:

[fail-kit]
# Skip patterns
skip_patterns = ["*_test.py", "test_*.py", "conftest.py"]

# Severity overrides
[fail-kit.severity]
FK001 = "error"  # Treat as error instead of warning
FK005 = "hint"   # Treat as hint instead of info

# Custom patterns
[fail-kit.patterns]
tool_calls = ["my_custom_tool\\.run"]

Development

# Install dev dependencies
pip install -e ".[dev]"

# Run tests
pytest

# Type checking
mypy fail_kit_lsp

# Format code
black fail_kit_lsp
ruff check fail_kit_lsp

License

MIT