BMAD-METHOD/bmad-core/templates/architecture-tmpl.md

{Project Name} Architecture Document

High Level Architecture

LLM: Produce the following child sections, and then derscribe why you made the choices you did if any choices were taken that you made that were unique or deviated from what was given as initial context from the PRD or prior discussions, and immediately execute tasks#advanced-elicitation display

Introduction / Preamble

{This document outlines the overall project architecture, including backend systems, shared services, and non-UI specific concerns. Its primary goal is to serve as the guiding architectural blueprint for AI-driven development, ensuring consistency and adherence to chosen patterns and technologies.

Relationship to Frontend Architecture: If the project includes a significant user interface, a separate Frontend Architecture Document (typically named front-end-architecture-tmpl.txt or similar, and linked in the "Key Reference Documents" section) details the frontend-specific design and MUST be used in conjunction with this document. Core technology stack choices documented herein (see "Definitive Tech Stack Selections") are definitive for the entire project, including any frontend components.}

Technical Summary

{ Provide a brief paragraph overview of the system's architecture, key components, technology choices, and architectural patterns used. Reference the goals from the PRD. }

High Level Overview

LLM: Describe the main architectural style (e.g., Monolith, Microservices, Serverless, Event-Driven), reflecting the decision made in the PRD. Explain the repository structure (Monorepo/Polyrepo). Explain the primary user interaction or data flow at a conceptual level.

High Level Project Diagram

LLM: Produce a Mermaid High Level Diagram that best Describes the Overall PRD Backend Architecture

Architectural and Design Patterns

{ List the key high-level patterns chosen for the architecture. These foundational patterns should be established early as they guide component design, interactions, and technology choices. }

• Pattern 1: {e.g., Serverless, Event-Driven, Microservices, CQRS} - Rationale/Reference: {Briefly why, or link to a more detailed explanation if needed}
• Pattern 2: {e.g., Dependency Injection, Repository Pattern, Module Pattern} - Rationale/Reference: {...}
• Pattern N: {...}

Components

LLM: Describe the major logical components or services of the system and their responsibilities, reflecting the decided overall architecture (e.g., distinct microservices, modules within a monolith, packages within a monorepo) and the architectural patterns adopted. Explain how they collaborate. Produce a component diagram or diagrams as needed in mermaid as a child section. Immediately execute tasks#advanced-elicitation display

• Component A: {Description of responsibility}

{Insert component diagram here if it helps - e.g., using Mermaid graph TD or C4 Model Container/Component Diagram}

• Component N...: {Description of responsibility}

{ Insert component diagram here if it helps - e.g., using Mermaid graph TD or C4 Model Container/Component Diagram }

Component Diagrams

Source Tree

LLM: Provide an ASCII or Mermaid diagram representing the project's proposed folder structure based on information we have aligned on to this point such as monorepo vs polyrepo vs monolith vs microservices. The following is a general example, but is very project and technology dependant. Immediately upon render of the section, execute tasks#advanced-elicitation display.

{project-root}/
├── .github/                    # CI/CD workflows (e.g., GitHub Actions)
│   └── workflows/
│       └── main.yml
├── .vscode/                    # VSCode settings (optional)
│   └── settings.json
├── build/                      # Compiled output (if applicable, often git-ignored)
├── config/                     # Static configuration files (if any)
├── docs/                       # Project documentation (PRD, Arch, etc.)
│   ├── index.md
│   └── ... (other .md files)
├── infra/                      # Infrastructure as Code (e.g., CDK, Terraform)
│   └── lib/
│   └── bin/
├── node_modules/ / venv / target/ # Project dependencies (git-ignored)
├── scripts/                    # Utility scripts (build, deploy helpers, etc.)
├── src/                        # Application source code
│   ├── backend/                # Backend-specific application code (if distinct frontend exists)
│   │   ├── core/               # Core business logic, domain models
│   │   ├── services/           # Business services, orchestrators
│   │   ├── adapters/           # Adapters to external systems (DB, APIs)
│   │   ├── controllers/ / routes/ # API endpoint handlers
│   │   └── main.ts / app.py    # Backend application entry point
│   ├── frontend/               # Placeholder: See Frontend Architecture Doc for details if used
│   ├── shared/ / common/       # Code shared (e.g., types, utils, domain models if applicable)
│   │   └── types/
│   └── main.ts / index.ts / app.ts # Main application entry point (if not using backend/frontend split above)
├── stories/                    # Generated story files for development (optional)
│   └── epic1/
├── test/                       # Automated tests
│   ├── unit/                   # Unit tests (mirroring src structure)
│   ├── integration/            # Integration tests
│   └── e2e/                    # End-to-end tests
├── .env.example                # Example environment variables
├── .gitignore                  # Git ignore rules
├── package.json / requirements.txt / pom.xml # Project manifest and dependencies
├── tsconfig.json / pyproject.toml # Language-specific configuration (if applicable)
├── Dockerfile                  # Docker build instructions (if applicable)
└── README.md                   # Project overview and setup instructions

External APIs

LLM: List the External API name proposed if needed - if unsure you can suggest or ask the user when it seems we will need to rely on some external services depending on the project requirements. List the URL to doc or API Spec (openapi / swagger url) - if you are unsure as user to provide after you explain when we need a given API or similar. For each API we need the following information

{External API Name} API

• Purpose: {What service does this API provide?}
• Base URL(s): {e.g., https://some-saas/api/v1/}
• Authentication/Authorization: {Describe how access is controlled.}
• Endpoints Used
  • Identify the endpoint - short statement of purpose to use it

REST API Spec

LLM: Create OpenAPI Spec (Swagger) proposed if proposing a rest api

Data Models

LLM: List major data models, discussing with the user why you have listed the specific data models

Database Schema

LLM: Create a database schema model proposed to meet the project requirements. For each schema, include the purpose of each schema model

Core Workflow

{ Illustrate key or complex workflows using mermaid sequence diagrams. Can have high level tying the full project together, and also smaller epic level sequence diagrams. }

Tech Stack

LLM: This section outlines the definitive technology choices for the project. These selections should be made after a thorough understanding of the project's requirements, components, data models, and core workflows. The Architect Agent should guide the user through these decisions, ensuring each choice is justified and recorded accurately in the table below.

[[LLM: This table is the single source of truth for all technology selections. Other architecture documents (e.g., Frontend Architecture) must refer to these choices and elaborate on their specific application rather than re-defining them.]]

[[LLM: Key decisions to discuss and finalize here, which will then be expanded upon and formally documented in the detailed stack table below, include considerations such as:

• Preferred Starter Template Frontend: { Url to template or starter, if used }
• Preferred Starter Template Backend: { Url to template or starter, if used }
• Primary Language(s) & Version(s): {e.g., TypeScript 5.x, Python 3.11 - Specify exact versions, e.g., 5.2.3}
• Primary Runtime(s) & Version(s): {e.g., Node.js 22.x - Specify exact versions, e.g., 22.0.1}

Must be definitive selections; do not list open-ended choices (e.g., for web scraping, pick one tool, not two). Specify exact versions (e.g., 18.2.0). If 'Latest' is used, it implies the latest stable version at the time of this document's last update, and the specific version (e.g., xyz-library@2.3.4) should be recorded. Pinning versions is strongly preferred to avoid unexpected breaking changes for the AI agent.]]

• Cloud Provider(s): {e.g., AWS, Azure, GCP, On-premise}
• Core Services Used: {List key managed services - e.g., Lambda, S3, Kubernetes Engine, RDS, Kafka}

Category	Technology	Version	Purpose

Infrastructure and Deployment

• Infrastructure as Code (IaC): {Tool used - e.g., AWS CDK, Terraform...} - Location: {Link to IaC code repo/directory}
• Deployment Strategy: {e.g., CI/CD pipeline with automated promotions, Blue/Green, Canary} - Tools: {e.g., Jenkins, GitHub Actions, GitLab CI}
• Environments: {List environments - e.g., Development, Staging, Production}
• Environment Promotion: {Describe steps, e.g., dev -> staging (manual approval / automated tests pass) -> production (automated after tests pass and optional manual approval)}
• Rollback Strategy: {e.g., Automated rollback on health check failure post-deployment, Manual trigger via CI/CD job, IaC state rollback. Specify primary mechanism.}

Error Handling Strategy

• General Approach: {e.g., Use exceptions as primary mechanism, return error codes/tuples for specific modules, clearly defined custom error types hierarchy.}
• Logging:
  • Library/Method: {e.g., console.log/error (Node.js), Python logging module with structlog, dedicated logging library like Pino or Serilog. Specify the chosen library.}
  • Format: {e.g., JSON, plain text with timestamp and severity. JSON is preferred for structured logging.}
  • Levels: {e.g., DEBUG, INFO, WARN, ERROR, CRITICAL. Specify standard usage for each.}
  • Context: {What contextual information must be included? e.g., Correlation ID, User ID (if applicable and safe), Service Name, Operation Name, Key Parameters (sanitized).}
• Specific Handling Patterns:
  • External API Calls: {Define retry mechanisms (e.g., exponential backoff, max retries - specify library if one is mandated like Polly or tenacity), circuit breaker pattern usage (e.g., using resilience4j or equivalent - specify if and how), timeout configurations (connect and read timeouts). How are API errors (4xx, 5xx) translated or propagated?}
  • Internal Errors / Business Logic Exceptions: {How to convert internal errors to user-facing errors if applicable (e.g., generic error messages with a unique ID for support, specific error codes). Are there defined business exception classes?}
  • Transaction Management: {Approach to ensure data consistency in case of errors during multi-step operations, e.g., database transactions (specify isolation levels if non-default), Saga pattern for distributed transactions (specify orchestrator/choreography and compensation logic).}

Coding Standards

LLM: These standards are mandatory for all code generation by AI agents and human developers. Deviations are not permitted unless explicitly approved and documented as an exception in this section or a linked addendum. Ensure you let the user know they should really review this and also make sections to it as needed during development once this is sharded out to the coding-standards file as this will always be used by the dev agent to ensure it follows standards. Also keep this section as short and lean as possible to not bloat the context with stuff the agent will probably know anyways. The following are possible sections, it all depends on the project and what the user things should be included - use your best judgement to keep it lean but clear to keep a dumb dev agent on the rails. This applies to the child sections also, do not add them if not needed, or strive to keep them succinct.

• Primary Runtime(s): {e.g., Node.js 22.x, Python Runtime for Lambda - refer to Definitive Tech Stack}
• Style Guide & Linter: {e.g., ESLint with Airbnb config + Prettier; Black, Flake8, MyPy; Go fmt, golint. Specify chosen tools and link to configuration files (e.g., .eslintrc.js, pyproject.toml). Linter rules are mandatory and must not be disabled without cause.}
• Naming Conventions:
  • Variables: {e.g., camelCase (JavaScript/TypeScript/Java), snake_case (Python/Ruby)}
  • Functions/Methods: {e.g., camelCase (JavaScript/TypeScript/Java), snake_case (Python/Ruby)}
  • Classes/Types/Interfaces: {e.g., PascalCase}
  • Constants: {e.g., UPPER_SNAKE_CASE}
  • Files: {e.g., kebab-case.ts (TypeScript), snake_case.py (Python), PascalCase.java (Java). Be specific per language.}
  • Modules/Packages: {e.g., camelCase or snake_case. Be specific per language.}
• File Structure: Adhere to the layout defined in the "Project Structure" section and the Frontend Architecture Document if applicable.
• Unit Test File Organization: {e.g., *.test.ts/*.spec.ts co-located with source files; test_*.py in a parallel tests/ directory. Specify chosen convention.}
• Asynchronous Operations: {e.g., Always use async/await in TypeScript/JavaScript/Python for promise-based operations; Goroutines/Channels in Go with clear patterns for error propagation and completion; Java CompletableFuture or Project Reactor/RxJava if used.}
• Type Safety: {e.g., Leverage TypeScript strict mode (all flags enabled); Python type hints (enforced by MyPy); Go static typing; Java generics and avoidance of raw types. All new code must be strictly typed.}
  • Type Definitions: {Location, e.g., src/common/types.ts, shared packages, or co-located. Policy on using any or equivalent (strongly discouraged, requires justification).}
• Comments & Documentation:
  • Code Comments: {Expectations for code comments: Explain why, not what, for complex logic. Avoid redundant comments. Use standard formats like JSDoc, TSDoc, Python docstrings (Google/NumPy style), GoDoc, JavaDoc.}
  • READMEs: {Each module/package/service should have a README explaining its purpose, setup, and usage if not trivial.}
• Dependency Management: {Tool used - e.g., npm/yarn, pip/poetry, Go modules, Maven/Gradle. Policy on adding new dependencies (e.g., approval process, check for existing alternatives, security vulnerability scans). Specify versioning strategy (e.g., prefer pinned versions, use tilde ~ for patches, caret ^ for minor updates - be specific).}

Detailed Language & Framework Conventions

[[LLM: For each primary language and framework selected in the "Definitive Tech Stack Selections", the following specific conventions must be adhered to. If a chosen technology is not listed below, it implies adherence to its standard, widely accepted best practices and the general guidelines in this document.]]

{Language/Framework 1 Name, e.g., TypeScript/Node.js} Specifics

• Immutability: {e.g., "Always prefer immutable data structures. Use Readonly<T>, ReadonlyArray<T>, as const for object/array literals. Avoid direct mutation of objects/arrays passed as props or state. Consider libraries like Immer for complex state updates."}
• Functional vs. OOP: {e.g., "Favor functional programming constructs (map, filter, reduce, pure functions) for data transformation and business logic where practical. Use classes for entities, services with clear state/responsibilities, or when framework conventions (e.g., NestJS) demand."}
• Error Handling Specifics: {e.g., "Always use Errorobjects or extensions thereof forthrow. Ensure Promiserejections are alwaysErrorobjects. Use custom error classes inheriting from a baseAppError for domain-specific errors."}
• Null/Undefined Handling: {e.g., "Strict null checks (strictNullChecks) must be enabled. Avoid ! non-null assertion operator; prefer explicit checks, optional chaining (?.), or nullish coalescing (??). Define clear strategies for optional function parameters and return types."}
• Module System: {e.g., "Use ESModules (import/export) exclusively. Avoid CommonJS (require/module.exports) in new code."}
• Logging Specifics: {e.g., "Use the chosen structured logging library. Log messages must include a correlation ID. Do not log sensitive PII. Use appropriate log levels."}
• Framework Idioms (e.g., for NestJS/Express): {e.g., "NestJS: Always use decorators for defining modules, controllers, services, DTOs. Adhere strictly to the defined module structure and dependency injection patterns. Express: Define middleware patterns, routing structure."}
• Key Library Usage Conventions: {e.g., "When using Axios, create a single configured instance. For date/time, use {date-fns/Luxon/Day.js} and avoid native Date object for manipulations."}
• Code Generation Anti-Patterns to Avoid: {e.g., "Avoid overly nested conditional logic (max 2-3 levels). Avoid single-letter variable names (except for trivial loop counters like i, j, k). Do not write code that bypasses framework security features (e.g., ORM query builders)."}

{Language/Framework 2 Name, e.g., Python} Specifics:

• Immutability: {e.g., "Use tuples for immutable sequences. For classes, consider @dataclass(frozen=True). Be mindful of mutable default arguments."}
• Functional vs. OOP: {e.g., "Employ classes for representing entities and services. Use functions for stateless operations. List comprehensions/generator expressions are preferred over map/filter for readability."}
• Error Handling Specifics: {e.g., "Always raise specific, custom exceptions inheriting from a base AppException. Use try-except-else-finallyblocks appropriately. Avoid broadexcept Exception: clauses without re-raising or specific handling."}
• Resource Management: {e.g., "Always use with statements for resources like files or DB connections to ensure they are properly closed."}
• Type Hinting: {e.g., "All new functions and methods must have full type hints. Run MyPy in CI. Strive for disallow_untyped_defs = True."}
• Logging Specifics: {e.g., "Use the loggingmodule configured for structured output (e.g., withpython-json-logger). Include correlation IDs."}
• Framework Idioms (e.g., for Django/Flask/FastAPI): {e.g., "Django: Follow fat models, thin views pattern. Use ORM conventions. FastAPI: Utilize Pydantic for request/response models and dependency injection for services."}
• Key Library Usage Conventions: {e.g., "For HTTP requests, use httpxorrequestswith explicit timeout settings. For data manipulation, preferpandas where appropriate but be mindful of performance."}

{Add more Language/Framework sections as needed...}

• {Consider other things that the trained LLM Dev Agent could potentially be random about specific to the chosen language technologies and platforms that it should be reminded of here}

Test Strategy and Standards

LLM: Work with the user to determine the test strategy. Unit test framework, test organization, Test coverage desired, Build then test or TDD, integration testing approach, integration testing with local or remote resources, any mocking for integration tests and then what? for example if using dynamo, will we use in memory ddb, docker ddb, remote test db instance etc... Consider this for all proposed technologies and what makes sense to test locally or remote. Use any info that follows as needed. This section is NOT going to be read directly by the dev agent as it can get long, the minimum basics the agent needs to know will be listed in coding standards such as what the unit or integration test library is and any specific standard of style the dev needs to know. If the qa agent is used, that agent will be familiar and review this section.

• Tools: {Reiterate primary testing frameworks and libraries from Tech Stack, e.g., Jest, Playwright, PyTest, JUnit, Testcontainers.}
• Unit Tests:
  • Scope: {Test individual functions, methods, classes, or small modules in isolation. Focus on business logic, algorithms, and transformation rules.}
  • Location: {e.g., *.test.ts/*.spec.ts co-located with source files; test_*.py in a parallel tests/ directory, following language conventions.}
  • Mocking/Stubbing: {Specify chosen mocking library (e.g., Jest mocks, unittest.mock in Python, Mockito for Java). Mock all external dependencies (network calls, file system, databases, time).}
  • AI Agent Responsibility: {AI Agent must generate unit tests covering all public methods, significant logic paths, edge cases, and error conditions for any new or modified code.}
• Integration Tests:
  • Scope: {Test the interaction between several components or services within the application boundary. E.g., API endpoint to service layer to database (using a test database or in-memory version).}
  • Location: {e.g., /tests/integration or /src/integration-test (Java).}
  • Environment: {Specify how dependencies are handled (e.g., Testcontainers for databases/external services, in-memory databases, dedicated test environment).}
  • AI Agent Responsibility: {AI Agent may be tasked with generating integration tests for key service interactions or API endpoints based on specifications.}
• End-to-End (E2E) Tests:
  • Scope: {Validate complete user flows or critical paths through the system from the user's perspective (e.g., UI interaction, API call sequence).}
  • Tools: {Reiterate E2E testing tools from Tech Stack (e.g., Playwright, Cypress, Selenium).}
  • AI Agent Responsibility: {AI Agent may be tasked with generating E2E test stubs or scripts based on user stories or BDD scenarios. Focus on critical happy paths and key error scenarios.}
• Test Coverage:
  • Target: {Specify target code coverage if any (e.g., 80% line/branch coverage for unit tests). This is a guideline; quality of tests is paramount over raw coverage numbers.}
  • Measurement: {Tool used for coverage reports (e.g., Istanbul/nyc, Coverage.py, JaCoCo).}
• Mocking/Stubbing Strategy (General): {Beyond specific test types, outline general principles. e.g., "Prefer fakes or test doubles over extensive mocking where it improves test clarity and maintainability. Strive for tests that are fast, reliable, and isolated."}
• Test Data Management: {How is test data created, managed, and isolated? E.g., factories, fixtures, setup/teardown scripts, dedicated test data generation tools.}

Security

LLM: Outline key security considerations relevant to the codebase. These are mandatory and must be actively addressed by the AI agent during development.

• Input Sanitization/Validation: {Specify library/method for ALL external inputs (API requests, user-provided data, file uploads). E.g., 'Use class-validator with NestJS DTOs for all API inputs; all validation rules must be defined in DTOs.' For other languages, 'Use {validation_library} for all external inputs; define schemas and constraints.' Validation must occur at the boundary before processing.}
• Output Encoding: {Specify where and how output encoding should be performed to prevent XSS and other injection attacks. E.g., 'All dynamic data rendered in HTML templates must be contextually auto-escaped by the template engine (specify engine and confirm default behavior). If generating HTML/XML/JSON manually, use approved encoding libraries like {encoder_library_name}.'}
• Secrets Management: {Reference docs/environment-vars.md regarding storage for different environments. In code, access secrets only through a designated configuration module/service. Never hardcode secrets, include them in source control, or log them. Use specific tools for local development if applicable (e.g., Doppler, .env files NOT committed).}
• Dependency Security: {Policy on checking for vulnerable dependencies. E.g., 'Run automated vulnerability scans (e.g., npm audit, pip-audit, Snyk, Dependabot alerts) as part of CI. Update vulnerable dependencies promptly based on severity.' Policy on adding new dependencies (vetting process).}
• Authentication/Authorization Checks: {Where and how should these be enforced? E.g., 'All API endpoints (except explicitly public ones) must enforce authentication using the central auth module/middleware. Authorization (permission/role checks) must be performed at the service layer or entry point for protected resources.' Define patterns for implementing these checks.}
• Principle of Least Privilege (Implementation): {e.g., 'Database connection users must have only the necessary permissions (SELECT, INSERT, UPDATE, DELETE) for the specific tables/schemas they access. IAM roles for cloud services must be narrowly scoped to the required actions and resources.'}
• API Security (General): {e.g., 'Enforce HTTPS. Implement rate limiting and throttling (specify tool/method). Use standard HTTP security headers (CSP, HSTS, X-Frame-Options, etc. - specify which ones and their configuration). Follow REST/GraphQL security best practices.'}
• Error Handling & Information Disclosure: {Ensure error messages do not leak sensitive information (stack traces, internal paths, detailed SQL errors) to the end-user. Log detailed errors server-side, provide generic messages or error IDs to the client.}
• Regular Security Audits/Testing: {Mention if planned, e.g., penetration testing, static/dynamic analysis tool usage in CI (SAST/DAST).}
• {Other relevant practices, e.g., File upload security, Session management security, Data encryption at rest and in transit beyond HTTPS if specific requirements exist.}

Change Log

Change	Date	Version	Description	Author

--- Below, Prompt for Design Architect (If Project has UI) To Produce Front End Architecture ----