Skip to main content
Tech Packs are the modular knowledge units that make SoftArchitect AI stack-aware. Each pack is a self-contained Technical Encyclopedia entry for a specific technology: a profile describing when and why to use it, plus a set of enforceable governance rules that the AI applies when generating recommendations for your project. When you select a Tech Pack at project creation, the AI knows to ask Flutter-specific questions for a Flutter project, Python FastAPI patterns for a Python backend, and so on — rather than giving generic advice.

​
What a Tech Pack contains

Every Tech Pack follows the Standard Schema defined in packages/knowledge_base/02-TECH-PACKS/_STANDARD_SCHEMA/ and consists of two files:

00-TECH_PROFILE.md

Suitability analysis, trade-off matrix (learning curve, performance, scalability, ecosystem, LTS), system requirements, and compatible stack combinations.

01-RULES.md

Mandatory governance rules: naming conventions, architectural principles (SOLID), security patterns, testing targets, linting tools, and CI/CD enforcement.
Many packs also include a KNOWLEDGE_BASE/ subdirectory with supplementary reference material — code patterns, anti-patterns, and worked examples — that the RAG engine retrieves during document generation.

​
Available Tech Packs

The knowledge base ships with 32 Tech Packs across six categories:

​
Frontend

PackPath
Flutter (mobile/desktop)FRONTEND/mobile-flutter/
Android — KotlinFRONTEND/mobile-android-kotlin/
iOS — SwiftFRONTEND/mobile-ios-swift/
React NativeFRONTEND/mobile-react-native/
ReactFRONTEND/web-react/
VueFRONTEND/web-vue/
AngularFRONTEND/web-angular/
Web (general)FRONTEND/web-general/

​
Backend

PackPath
Python — FastAPIBACKEND/python-fastapi/
Python — DjangoBACKEND/python-django/
Python — FlaskBACKEND/python-flask/
Node.js — ExpressBACKEND/node-express/
Node.js — NestJSBACKEND/node-nest/
Java — Spring BootBACKEND/java-springboot/
GoBACKEND/go-lang/
C# — .NETBACKEND/csharp-dotnet/
PHP — LaravelBACKEND/php-laravel/

​
AI engineering

PackPath
LangChainAI_ENGINEERING/llm-langchain/
Semantic KernelAI_ENGINEERING/llm-semantic-kernel/
OllamaAI_ENGINEERING/model-ollama/
ChromaDBAI_ENGINEERING/vector-chromadb/
PineconeAI_ENGINEERING/vector-pinecone/

​
Data

PackPath
Firebase / FirestoreDATA/nosql-firebase/
PostgreSQLDATA/sql-postgresql/
MySQLDATA/sql-mysql/
RedisDATA/cache-redis/

​
DevOps & Cloud

PackPath
GitHub ActionsDEVOPS_CLOUD/ci-github-actions/
DockerDEVOPS_CLOUD/container-docker/
KubernetesDEVOPS_CLOUD/orchestration-k8s/
AWSDEVOPS_CLOUD/cloud-aws/
AzureDEVOPS_CLOUD/cloud-azure/

​
General

PackPath
Git conventionsgeneral/GIT_CONVENTIONS.md
OWASP Top 10general/OWASP_TOP_10.md
TDD methodologygeneral/TDD_METHODOLOGY.md

​
How a Tech Pack is structured

The Flutter Tech Pack (FRONTEND/mobile-flutter/) is a representative example.

​
Tech profile (00-TECH_PROFILE.md)

The profile answers the key selection question: “Is this technology the right fit?” 
# Tech Profile: Google Flutter

Category:   Cross-Platform UI Framework
License:    BSD-3-Clause
Language:   Dart 3.0+ (Sound Null Safety)
Version:    Flutter 3.19+ (Stable Channel)
It includes a suitability matrix that the AI uses to either recommend or steer away from a technology:
DimensionRatingComment
Development speed5/5Hot Reload changes code in milliseconds without losing state
UI performance5/5Stable 60/120 FPS via AOT compilation
Learning curve3/5~4–6 weeks for productivity (Dart + Riverpod + declarative paradigm)
Ecosystem4/5~80k packages on pub.dev; accelerated growth
LTS & viability5/5Backed by Google; used by Google Ads, Alibaba, BMW
The profile also documents anti-patterns (cases where the technology should not be used) and known stack integration compatibility with backends, auth providers, databases, and state management libraries.

​
Rules file (01-RULES.md)

The rules file contains the governance layer that the AI injects directly into RULES.md and AGENTS.md when generating your project’s root documents. For Flutter this includes:
  • Dart naming conventions (files: snake_case.dart, classes: PascalCase, private members: _prefix)
  • Riverpod 2.0 state management patterns (and why setState is an anti-pattern for complex logic)
  • SOLID principles mapped to Flutter’s widget and provider architecture
  • Mandatory linting with flutter_lints (strict mode)
  • Test coverage target ≥80% with flutter_test and mockito
  • Security rules: never use deprecated withOpacity(), always use withValues(alpha: x.x)

​
How the AI uses Tech Packs

During the interview (Phase 1 – Context and Phase 2 – Requirements), the AI queries the vector store for Tech Pack content relevant to the technologies you mention. This drives stack-specific questions:
  • Flutter pack loaded → AI asks about target platforms (iOS/Android/Desktop/Web), offline requirements, and state management preferences
  • FastAPI pack loaded → AI asks about async requirements, dependency injection patterns, and whether you need WebSocket or SSE streaming
  • Firebase pack loaded → AI asks about real-time sync requirements, authentication strategy, and data residency constraints
In Phase 3 (Architecture), the retrieved pack rules are injected into the RAG prompt so that TECH_STACK_DECISION.md and ARCH_DECISION_RECORDS.md cite concrete evidence from the knowledge base rather than generic opinions.

​
Selecting a Tech Pack

When you create a new project in the Flutter desktop app, the project setup dialog lists all available Tech Packs grouped by category. You can select multiple packs — for example, Flutter + FastAPI + Firebase + GitHub Actions.
You can also mention technologies during the interview conversation and the AI will automatically retrieve the relevant pack content from ChromaDB without requiring an explicit selection.

​
Adding custom Tech Packs

Custom packs are on the roadmap. The schema is already open: copy _STANDARD_SCHEMA/00-TECH_PROFILE.template.md and _STANDARD_SCHEMA/01-RULES.template.md, fill in the {{PLACEHOLDERS}}, place your files in a new subdirectory under the appropriate category folder, then re-ingest the knowledge base. 
packages/knowledge_base/02-TECH-PACKS/
└── BACKEND/
    └── my-custom-framework/       ← new directory
        ├── 00-TECH_PROFILE.md
        ├── 01-RULES.md
        └── KNOWLEDGE_BASE/        ← optional supplementary docs
            └── patterns.md
After ingestion, the AI will use your custom pack exactly like the built-in ones.

Build docs developers (and LLMs) love

Get started for freeTalk to us