pyagentshield 0.1.0__tar.gz

pyagentshield-0.1.0/.env.example

@@ -0,0 +1,119 @@
+# AgentGuard Environment Variables
+# Copy this file to .env and fill in your values
+# NEVER commit .env to version control!
+
+# =============================================================================
+# REQUIRED: OpenAI API Key
+# =============================================================================
+
+# Your OpenAI API key - REQUIRED for:
+#   1. LLM cleaning (recommended for best accuracy ~90%)
+#   2. Finetuning the embedding model (data preparation step)
+#
+# Get your key from: https://platform.openai.com/api-keys
+# Cost: ~$0.0003 per document for cleaning, ~$3-5 total for finetuning
+OPENAI_API_KEY=sk-...
+
+# =============================================================================
+# Embedding Model Configuration
+# =============================================================================
+
+# Provider: "local" uses sentence-transformers (recommended)
+# AGENTGUARD_EMBEDDINGS__PROVIDER=local
+
+# Model path - use your finetuned model for best accuracy (~95%)
+# After running: python scripts/finetune_local.py
+# AGENTGUARD_EMBEDDINGS__MODEL=./agentguard-embeddings-finetuned
+#
+# Or use a pre-trained model (less accurate but no finetuning needed):
+#   - all-MiniLM-L6-v2    (fast, ~70% accuracy)
+#   - all-mpnet-base-v2   (better, ~80% accuracy)
+# AGENTGUARD_EMBEDDINGS__MODEL=all-MiniLM-L6-v2
+
+# =============================================================================
+# Text Cleaning Configuration
+# =============================================================================
+
+# Cleaning method - how AgentGuard removes injection patterns before comparing
+#
+# Options:
+#   - "heuristic"  : Free, fast, regex-based (~70% accuracy)
+#   - "llm"        : Uses GPT-4o-mini, ~$0.0003/doc (~90% accuracy) [RECOMMENDED]
+#
+# AGENTGUARD_CLEANING__METHOD=llm
+
+# LLM model for cleaning (when method=llm)
+# gpt-4o-mini is cheapest and works great for this task
+# AGENTGUARD_CLEANING__LLM_MODEL=gpt-4o-mini
+
+# =============================================================================
+# ZEDD Detection Threshold
+# =============================================================================
+
+# Threshold determines when text is flagged as suspicious
+#
+# How it works:
+#   - ZEDD computes "drift" = how much text changes after cleaning
+#   - If drift > threshold → suspicious
+#
+# Options:
+#   - Leave empty/null : Auto-load from model's calibration.json [RECOMMENDED]
+#   - Set explicit value: e.g., 0.0083 (from your finetuned model)
+#
+# Higher threshold = fewer false positives, might miss some attacks
+# Lower threshold  = catches more attacks, but more false positives
+#
+# AGENTGUARD_ZEDD__THRESHOLD=
+
+# =============================================================================
+# Behavior on Detection
+# =============================================================================
+
+# What to do when a prompt injection is detected
+#
+# Options:
+#   - "block"  : Raise PromptInjectionDetected exception (strict)
+#   - "filter" : Remove suspicious documents silently (recommended for RAG)
+#   - "flag"   : Add metadata but pass through (for logging/monitoring)
+#   - "warn"   : Log warning but pass through (permissive)
+#
+# AGENTGUARD_BEHAVIOR__ON_DETECT=filter
+
+# Minimum confidence to trigger the on_detect action (0.0 to 1.0)
+# Higher = only act on high-confidence detections
+# AGENTGUARD_BEHAVIOR__CONFIDENCE_THRESHOLD=0.5
+
+# =============================================================================
+# Performance Settings
+# =============================================================================
+
+# Batch size for embedding operations (higher = faster but more memory)
+# AGENTGUARD_PERFORMANCE__BATCH_SIZE=32
+
+# Cache embeddings to avoid recomputing for repeated texts
+# AGENTGUARD_PERFORMANCE__CACHE_EMBEDDINGS=true
+
+# =============================================================================
+# Logging
+# =============================================================================
+
+# Log level: DEBUG, INFO, WARNING, ERROR
+# AGENTGUARD_LOGGING__LEVEL=INFO
+
+# =============================================================================
+# Finetuning Configuration (for scripts/finetune_local.py)
+# =============================================================================
+
+# These are used when running the finetuning script, not at runtime
+
+# Max samples to process (more = better model but higher API cost)
+# Default: 5000 samples costs ~$3-5 in OpenAI API calls
+# FINETUNE_MAX_SAMPLES=5000
+
+# Training batch size (reduce if you get OOM errors)
+# For 16GB Mac: use 8
+# For 8GB Mac: use 4
+# FINETUNE_BATCH_SIZE=8
+
+# Output directory for finetuned model
+# FINETUNE_OUTPUT_DIR=./agentguard-embeddings-finetuned

pyagentshield-0.1.0/.gitignore

@@ -0,0 +1,234 @@
+# Byte-compiled / optimized / DLL files
+__pycache__/
+*.py[codz]
+*$py.class
+
+.claude/
+ARCHITECTURE.md
+CLAUDE.md
+.DS_Store
+# C extensions
+*.so
+
+# Distribution / packaging
+.Python
+build/
+develop-eggs/
+dist/
+downloads/
+eggs/
+.eggs/
+lib/
+lib64/
+parts/
+sdist/
+var/
+wheels/
+share/python-wheels/
+*.egg-info/
+.installed.cfg
+*.egg
+MANIFEST
+
+# PyInstaller
+#  Usually these files are written by a python script from a template
+#  before PyInstaller builds the exe, so as to inject date/other infos into it.
+*.manifest
+*.spec
+
+# Installer logs
+pip-log.txt
+pip-delete-this-directory.txt
+
+# Unit test / coverage reports
+htmlcov/
+.tox/
+.nox/
+.coverage
+.coverage.*
+.cache
+nosetests.xml
+coverage.xml
+*.cover
+*.py.cover
+.hypothesis/
+.pytest_cache/
+cover/
+
+# Translations
+*.mo
+*.pot
+
+# Django stuff:
+*.log
+local_settings.py
+db.sqlite3
+db.sqlite3-journal
+
+# Flask stuff:
+instance/
+.webassets-cache
+
+# Scrapy stuff:
+.scrapy
+
+# Sphinx documentation
+docs/_build/
+
+# PyBuilder
+.pybuilder/
+target/
+
+# Jupyter Notebook
+.ipynb_checkpoints
+
+# IPython
+profile_default/
+ipython_config.py
+
+# pyenv
+#   For a library or package, you might want to ignore these files since the code is
+#   intended to run in multiple environments; otherwise, check them in:
+# .python-version
+
+# pipenv
+#   According to pypa/pipenv#598, it is recommended to include Pipfile.lock in version control.
+#   However, in case of collaboration, if having platform-specific dependencies or dependencies
+#   having no cross-platform support, pipenv may install dependencies that don't work, or not
+#   install all needed dependencies.
+#Pipfile.lock
+
+# UV
+#   Similar to Pipfile.lock, it is generally recommended to include uv.lock in version control.
+#   This is especially recommended for binary packages to ensure reproducibility, and is more
+#   commonly ignored for libraries.
+#uv.lock
+
+# poetry
+#   Similar to Pipfile.lock, it is generally recommended to include poetry.lock in version control.
+#   This is especially recommended for binary packages to ensure reproducibility, and is more
+#   commonly ignored for libraries.
+#   https://python-poetry.org/docs/basic-usage/#commit-your-poetrylock-file-to-version-control
+#poetry.lock
+#poetry.toml
+
+# pdm
+#   Similar to Pipfile.lock, it is generally recommended to include pdm.lock in version control.
+#   pdm recommends including project-wide configuration in pdm.toml, but excluding .pdm-python.
+#   https://pdm-project.org/en/latest/usage/project/#working-with-version-control
+#pdm.lock
+#pdm.toml
+.pdm-python
+.pdm-build/
+
+# pixi
+#   Similar to Pipfile.lock, it is generally recommended to include pixi.lock in version control.
+#pixi.lock
+#   Pixi creates a virtual environment in the .pixi directory, just like venv module creates one
+#   in the .venv directory. It is recommended not to include this directory in version control.
+.pixi
+
+# PEP 582; used by e.g. github.com/David-OConnor/pyflow and github.com/pdm-project/pdm
+__pypackages__/
+
+# Celery stuff
+celerybeat-schedule
+celerybeat.pid
+
+# SageMath parsed files
+*.sage.py
+
+# Environments
+.env
+.envrc
+.venv
+env/
+venv/
+ENV/
+env.bak/
+venv.bak/
+
+# Spyder project settings
+.spyderproject
+.spyproject
+
+# Rope project settings
+.ropeproject
+
+# mkdocs documentation
+/site
+
+# mypy
+.mypy_cache/
+.dmypy.json
+dmypy.json
+
+# Pyre type checker
+.pyre/
+
+# pytype static type analyzer
+.pytype/
+
+# Cython debug symbols
+cython_debug/
+
+# PyCharm
+#  JetBrains specific template is maintained in a separate JetBrains.gitignore that can
+#  be found at https://github.com/github/gitignore/blob/main/Global/JetBrains.gitignore
+#  and can be added to the global gitignore or merged into this file.  For a more nuclear
+#  option (not recommended) you can uncomment the following to ignore the entire idea folder.
+#.idea/
+
+# Abstra
+# Abstra is an AI-powered process automation framework.
+# Ignore directories containing user credentials, local state, and settings.
+# Learn more at https://abstra.io/docs
+.abstra/
+
+# Visual Studio Code
+#  Visual Studio Code specific template is maintained in a separate VisualStudioCode.gitignore 
+#  that can be found at https://github.com/github/gitignore/blob/main/Global/VisualStudioCode.gitignore
+#  and can be added to the global gitignore or merged into this file. However, if you prefer, 
+#  you could uncomment the following to ignore the entire vscode folder
+# .vscode/
+
+# Ruff stuff:
+.ruff_cache/
+
+# PyPI configuration file
+.pypirc
+
+# Cursor
+#  Cursor is an AI-powered code editor. `.cursorignore` specifies files/directories to
+#  exclude from AI features like autocomplete and code analysis. Recommended for sensitive data
+#  refer to https://docs.cursor.com/context/ignore-files
+.cursorignore
+.cursorindexingignore
+
+# Marimo
+marimo/_static/
+marimo/_lsp/
+__marimo__/
+
+# AgentShield specific
+# Model cache and outputs
+*.pt
+*.bin
+*.safetensors
+agentshield-cleaner-lora/
+agentshield-cleaner-lora-merged/
+models/
+checkpoints/
+
+# HuggingFace cache
+.cache/huggingface/
+
+# Training outputs
+wandb/
+outputs/
+runs/
+
+# Never commit credentials
+.env.local
+credentials.json
+secrets.yaml