{"id":48391558,"url":"https://github.com/langchain-samples/framework-agnostic-guardrails","last_synced_at":"2026-04-06T00:15:05.166Z","repository":{"id":329355428,"uuid":"1119235266","full_name":"langchain-samples/framework-agnostic-guardrails","owner":"langchain-samples","description":"Framework-agnostic guardrails modeled after LangChain pre-built guardrails middleware","archived":false,"fork":false,"pushed_at":"2025-12-19T03:08:22.000Z","size":178,"stargazers_count":0,"open_issues_count":1,"forks_count":0,"subscribers_count":0,"default_branch":"main","last_synced_at":"2025-12-21T23:46:55.865Z","etag":null,"topics":["beginner","guardrails","middleware"],"latest_commit_sha":null,"homepage":"","language":"Python","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"mit","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/langchain-samples.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":"LICENSE","code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null,"zenodo":null,"notice":null,"maintainers":null,"copyright":null,"agents":null,"dco":null,"cla":null}},"created_at":"2025-12-19T00:52:54.000Z","updated_at":"2025-12-19T03:04:19.000Z","dependencies_parsed_at":null,"dependency_job_id":null,"html_url":"https://github.com/langchain-samples/framework-agnostic-guardrails","commit_stats":null,"previous_names":["christineastoria/framework-agnostic-guardrails"],"tags_count":null,"template":false,"template_full_name":null,"purl":"pkg:github/langchain-samples/framework-agnostic-guardrails","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/langchain-samples%2Fframework-agnostic-guardrails","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/langchain-samples%2Fframework-agnostic-guardrails/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/langchain-samples%2Fframework-agnostic-guardrails/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/langchain-samples%2Fframework-agnostic-guardrails/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/langchain-samples","download_url":"https://codeload.github.com/langchain-samples/framework-agnostic-guardrails/tar.gz/refs/heads/main","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/langchain-samples%2Fframework-agnostic-guardrails/sbom","scorecard":null,"host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":286080680,"owners_count":31454268,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2026-04-05T21:22:52.476Z","status":"ssl_error","status_checked_at":"2026-04-05T21:22:51.943Z","response_time":75,"last_error":"SSL_connect returned=1 errno=0 peeraddr=140.82.121.6:443 state=error: unexpected eof while reading","robots_txt_status":"success","robots_txt_updated_at":"2025-07-24T06:49:26.215Z","robots_txt_url":"https://github.com/robots.txt","online":false,"can_crawl_api":true,"host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":["beginner","guardrails","middleware"],"created_at":"2026-04-06T00:15:03.800Z","updated_at":"2026-04-06T00:15:05.158Z","avatar_url":"https://github.com/langchain-samples.png","language":"Python","funding_links":[],"categories":[],"sub_categories":[],"readme":"# Framework-Agnostic Guardrails\n\nA simple, framework-agnostic wrapper for adding guardrails to AI agents at the application level.\n\n## Guard() Wrapper\n\n`Guard()` is a wrapper that provides input and output validation for AI agent calls. It operates as application-level middleware:\n1. Validates user input before execution (e.g., prompt injection detection, PII screening)\n2. Executes your agent using any framework\n3. Validates agent output before returning results (e.g., regulatory compliance, hallucination detection)\n\nWe believe **application-level guardrails are more effective** than placing guardrails in the observability layer. Security and compliance decisions should be made where they matter most - in your application code, before and after your agent runs.\n\nThis project demonstrates how easy it is to integrate guardrails into your agent using open-source tools like [OpenEvals](https://github.com/langchain-ai/openevals).\n\n\n## Quick Start\n\n### Installation\n\nUsing [UV](https://docs.astral.sh/uv/) (recommended):\n\n```bash\n# Install dependencies\nuv sync\n\n# Set up your OpenAI API key\ncp .env.example .env\n# Edit .env and add your OPENAI_API_KEY\n```\n\n### Basic Usage\n\n```python\nfrom guard import Guard, PIIDetectionGuardrail, JailbreakGuardrail\n\n# Create guard with multiple guardrails\nguard = Guard([\n    PIIDetectionGuardrail(),     # Pattern-based\n    JailbreakGuardrail(),         # LLM-based\n])\n\n# Works with ANY agent function\ndef my_agent(user_input: str) -\u003e str:\n    # Your agent logic (OpenAI, Anthropic, whatever)\n    return client.chat.completions.create(...)\n\n# Run with guardrails\nresult = guard.run(\n    user_input=\"What's the weather?\",\n    agent_call=my_agent\n)\n\nif result.blocked:\n    print(f\"Blocked: {result.findings}\")\nelse:\n    print(f\"Safe: {result.output}\")\n```\n\n### Flexible Input/Output Validation\n\nEach guardrail can be configured to check inputs, outputs, or both:\n\n```python\nguard = Guard([\n    # Check only user input for jailbreak attempts\n    JailbreakGuardrail(check_input=True, check_output=False),\n    \n    # Check only agent output for compliance\n    RegulatoryComplianceGuardrail(check_input=False, check_output=True),\n    \n    # Check both input and output for PII\n    PIIDetectionGuardrail(check_input=True, check_output=True),\n])\n```\n\n**Smart defaults:** Each guardrail comes with sensible defaults based on its purpose (e.g., jailbreak checks input only, compliance checks output only).\n\n## Available Guardrails\n\n### Pattern-Based (Fast, No API Required)\n- **PIIDetectionGuardrail** - Block emails, SSNs, phone numbers\n- **ToxicLanguageGuardrail** - Block offensive content\n\n### LLM-Based (Semantic, Uses OpenEvals)\n- **JailbreakGuardrail** - Detect manipulation attempts\n- **RegulatoryComplianceGuardrail** - Ensure financial/legal compliance\n- **BiasDetectionGuardrail** - Catch discriminatory content\n- **CorrectnessGuardrail** - Validate factual accuracy\n- **HallucinationGuardrail** - Detect made-up information\n\n### Custom\n- **CustomGuardrail** - Your own business logic\n\n## Examples\n\nSee [`examples.py`](examples.py) for detailed demonstrations of each guardrail, including:\n- Input vs output validation\n- Pattern-based vs LLM-based checks\n- Custom business logic\n- Multiple guardrails working together\n\n## Credits\n\nBuilt using:\n- [OpenEvals](https://github.com/langchain-ai/openevals) - LLM-as-judge framework\n- Inspired by [LangChain's guardrails middleware](https://docs.langchain.com/oss/python/langchain/guardrails)\n\n## License\n\nMIT\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Flangchain-samples%2Fframework-agnostic-guardrails","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Flangchain-samples%2Fframework-agnostic-guardrails","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Flangchain-samples%2Fframework-agnostic-guardrails/lists"}