Imported from SpecterOps/Nemesis (AGENTS.md). Install upstream with npx skills add SpecterOps/Nemesis. Copyright stays with the author.

Repository Guidelines

File Restrictions

NEVER read, open, or access .env files under any circumstances. This includes:

• .env
• .env.local
• .env.development
• .env.production
• Any file matching the pattern .env*

If you need environment variable information, refer to env.example instead.

Project Overview

Nemesis is an open-source, centralized data processing platform (v2.0) that ingests, enriches, and enables collaborative analysis of files collected during offensive security assessments. Built on Docker with Dapr integration, it functions as an "offensive VirusTotal."

Project Structure & Module Organization

Nemesis is organized as a multi-service monorepo:

• projects/: deployable services (for example web_api, file_enrichment, frontend, agents), each typically with its own pyproject.toml and tests/.
• libs/: shared Python libraries (common, file_linking, nemesis_dpapi, file_enrichment_modules).
• infra/: Docker, Dapr, Traefik, and observability configuration.
• tools/: repo-level developer scripts (install_dev_env.sh, lint.sh, test.sh, nemesis-ctl.sh).
• docs/: MkDocs content and generated API documentation assets.

Common Commands

Running Nemesis

# Start production services
./tools/nemesis-ctl.sh start prod

# Start with monitoring, jupyter, and LLM support
./tools/nemesis-ctl.sh start prod --monitoring --jupyter --llm

# Start development environment (always builds)
./tools/nemesis-ctl.sh start dev

# Stop services (use same flags as start)
./tools/nemesis-ctl.sh stop prod --monitoring --jupyter --llm

# Stop and remove volumes
./tools/nemesis-ctl.sh clean prod --monitoring --jupyter --llm

Development Setup

# Install all uv dependencies across projects
./tools/install_dev_env.sh

# Install dependencies for a single project
cd projects/web_api && uv sync

Package Management

For adding, upgrading, or removing packages, use the $managing-packages skill.

Linting & Formatting (Ruff)

Run linting/formatting after each fix/change.

# Check all Python code (configured in root pyproject.toml)
uv run ruff check . --fix

# Format code
uv run ruff format .

# Check specific project
cd projects/web_api && uv run ruff check . --fix

Testing

# Run tests for a specific project
cd projects/file_enrichment && uv run pytest tests/

# Run single test file
cd projects/web_api && uv run pytest tests/test_specific.py

# Run specific test
cd projects/web_api && uv run pytest tests/test_file.py::test_function_name

Docker Commands

Information about building dev/prod container images can be found in the docker compose docs.

General instructions for dev docker containers/images:

# Build base images first (required before building services)
docker compose -f compose.base.yaml build

# View logs for a service
docker compose logs -f web-api

# Rebuild Dapr-enabled services (must include the -dapr sidecar)
docker compose up -d --build file-enrichment file-enrichment-dapr
docker compose up -d --build web-api web-api-dapr

Note: Services using Dapr (file-enrichment, web-api, alerting, etc.) have a companion -dapr sidecar container. When rebuilding these services, always restart both the service and its Dapr sidecar to ensure proper communication.

Testing Guidelines

• Write tests for every new feature. New functionality must include tests before it is considered complete.
• Cover both happy path and unhappy path cases. Tests should verify correct behavior with valid inputs (positive cases) and proper error handling with invalid inputs, missing data, edge cases, and failure conditions (negative cases).
• Mock external services. Do not make real calls to external dependencies (Dapr, SeaweedFS, PostgreSQL, RabbitMQ, etc.) in unit tests. Use mocks or fakes to isolate the code under test.

Architecture

Tech Stack

• Python 3.13 with uv for dependency management
• FastAPI for REST services
• React 18 + Vite + TypeScript for frontend
• PostgreSQL for database
• Dapr for pub/sub, workflows, secrets, and service invocation
• RabbitMQ for message queue
• SeaweedFS for object storage
• Traefik for reverse proxy

Directory Structure

projects/           # Microservices
├── web_api/        # FastAPI REST gateway (port 8000)
├── file_enrichment/# File analysis orchestration with Dapr workflows
├── document_conversion/ # Text extraction, PDF conversion
├── cli/            # Command-line file submission tool
├── frontend/       # React web UI
├── alerting/       # Apprise-based notification service
├── housekeeping/   # Cleanup/retention service
├── agents/         # AI-powered alert triage
├── jupyter/        # Jupyter notebook service
├── dotnet_service/ # .NET assembly analysis service
├── titus_scanner/  # Secret scanning with Titus
├── velociraptor_connector/ # Velociraptor integration
└── ...

libs/               # Shared Python libraries
├── common/         # DB connections, logging, models, Dapr wrappers
├── file_enrichment_modules/ # 20+ enrichment module implementations
├── chromium/       # Chrome/Edge data extraction
├── file_linking/   # File association logic
└── nemesis_dpapi/  # Windows DPAPI credential handling

infra/              # Infrastructure configuration
├── dapr/           # Dapr components and workflows
├── postgres/       # Database schema (01-schema.sql)
├── grafana/        # Monitoring dashboards
└── ...

Data Flow

1. Files uploaded via web_api or cli → stored in SeaweedFS
2. file_enrichment receives file events via Dapr pub/sub
3. Dapr workflow orchestrates enrichment modules in parallel
4. Results written to PostgreSQL, findings published for alerting
5. frontend displays results via Hasura GraphQL

Key Configuration

• Root pyproject.toml: Ruff linting config (line-length: 120, Python 3.12)
• .env: Passwords, URLs, feature flags (copy from env.example)
• infra/postgres/01-schema.sql: Database schema

Adding Enrichment Modules

New file enrichment modules go in libs/file_enrichment_modules/. Each module implements a standard interface for detecting applicable files and extracting data.

Codex Skills

Repo-local Codex skills live under .codex/skills/.

• Use $enrichment-module-builder <file type description> to run the guided module workflow for new modules.
• Use $managing-packages <task> when adding, upgrading, removing, or syncing Python dependencies.

Coding Style & Naming Conventions

• Python version target is 3.13 (>=3.13,<3.14 in project manifests).
• Follow root Ruff config: 4-space indentation, 120-char line length, double quotes, and sorted imports.
• Naming: Python modules/functions snake_case, classes PascalCase, constants UPPER_SNAKE_CASE.
• In frontend code, keep component filenames PascalCase under projects/frontend/src/components.

Commit Guidelines

• Match existing commit style: short, imperative subjects like fix docs, bump deps, linting, lint+test scripts; optional issue/PR refs (for example (#100)).

Kubernetes (k3d) Deployment

See k8s/README.md and docs/kubernetes.md for full documentation.

# Setup cluster (k3d + Traefik + Dapr + KEDA via Helm)
./k8s/scripts/setup-cluster.sh

# Deploy with pre-built images
./k8s/scripts/deploy.sh install

# Build locally and deploy
./k8s/scripts/deploy.sh install --build

# Check status
./k8s/scripts/deploy.sh status

# Verify deployment
./k8s/scripts/verify.sh

# Teardown
./k8s/scripts/teardown-cluster.sh

Helm chart is at k8s/helm/nemesis/. Configuration in values.yaml. KEDA autoscales file-enrichment and document-conversion based on RabbitMQ queue depth.

Configuration Tips

• Do not commit secrets; keep .env local and derive it from env.example.
• When changing runtime topology, update related compose*.yaml and infra/* files in the same commit.
• For k8s changes, update the Helm chart under k8s/helm/nemesis/ and test with ./k8s/scripts/deploy.sh install --dry-run.