convez.ai
MCP ServerAutomaçãoIntermediário
VenkateshDoijode

AITestCaseGen MCP — Gerador de Casos de Teste com IA

VenkateshDoijode/ai-testcase-generator-mcp

Gera, importa e gerencia casos de teste automaticamente no Zephyr Scale

Ver no GitHub
InicialPluginScore · de 0 a 100

Servidor MCP que gera casos de teste abrangentes (positivos, negativos, edge cases) a partir de metadados de endpoints, alimentado por IA/LLMs. Integra-se nativamente com Jira Zephyr Scale, funcionando offline via Ollama ou HuggingFace, sem dependência de APIs pagas.

Instalação

$npx convez-ai VenkateshDoijode/ai-testcase-generator-mcp

1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.

Métricas

Estrelas
3
Forks
0
Acompanhando
0
Saúde
28%
Último commit
há 5 meses
Versão
—
Issues abertas
0
Contribuidores
1

Casos de uso

  • Automatizar geração de casos de teste a partir de requisitos Jira/Confluence
  • Integração completa com Jira Zephyr Scale para ciclos de teste
  • Ambientes air-gapped sem acesso a APIs de IA em nuvem

Veredito

Ferramenta robusta de automação QA com foco em ambientes enterprise offline-first. A abordagem de degradação graciosa (regra-base → Ollama → HuggingFace) e integração nativa Zephyr a distinguem; requer experiência Python/CLI e conhecimento de QA para máximo aproveito.

README

A tradução para o português deste README ainda está na fila.

🧪 AITestCaseGen

AI-powered test case management toolkit for Jira Zephyr Scale

Generate, import, update, link, and report on test cases — fully automated, with or without cloud AI.

Python License Zephyr Scale HuggingFace PRs Welcome


🚀 Why AITestCaseGen?

Writing test cases manually from requirements is one of the biggest time sinks in QA. AITestCaseGen reads your requirements — from Jira, Confluence, documents, or plain text — and generates structured positive, negative, and edge-case test cases, then pushes them directly into Zephyr Scale.

It's not just a generator. It's a full test case lifecycle CLI:

🧠 GenerateCreate test cases from Jira issues, Confluence pages, docs, or text
📥 ImportPush generated cases directly into Zephyr Scale
✏️ UpdateBulk-edit existing test cases
🔗 LinkConnect test cases to Jira issues
🗑️ DeleteRemove test cases (with dry-run safety)
📤 ExportPull existing cases out of Zephyr
🔄 Test CyclesUpdate execution results in bulk

🤖 AI Engine Priority

The script automatically picks the best available engine — zero config required:

PriorityEngineCondition
1HuggingFacetransformers installed + model downloadable
2OllamaOllama installed and running locally
3Rule-basedAlways works — no AI needed

This means the tool never hard-fails — even with no internet, no API key, and no GPU.


✨ Why This Is Different

Most "AI test case generator" tools are tightly coupled to a single cloud provider — if your API key expires or your org blocks external LLM calls, the tool stops working entirely. AITestCaseGen is built around graceful degradation:

  • No hard dependency on a paid API — HuggingFace models run via your Artifactory mirror or locally, Ollama runs fully offline, and the rule-based engine guarantees the script keeps working even with zero AI infrastructure.
  • Enterprise-friendly by default — designed for locked-down environments with no access to OpenAI/Anthropic APIs; works fully air-gapped.
  • End-to-end lifecycle, not just generation — generates and imports, updates, deletes, links, exports, and updates test cycle executions directly via the Zephyr Scale REST API.
  • Multi-source input — Jira issues, Confluence pages, document folders (.docx / .pdf / .txt), or raw text all feed the same pipeline.

📊 Comparison with Similar Approaches

CapabilityAITestCaseGenChatGPT / Copy-PasteCommercial Zephyr AI Add-ons
Works offline / air-gapped✅ Ollama or local model❌ Cloud API required❌ Usually cloud-only
No paid API key required✅ Rule-based fallback❌ Subscription needed❌ License cost
Direct Zephyr Scale integration✅ Full API pipeline❌ Manual copy-paste✅ Vendor UI only
Multi-source input✅ Jira, Confluence, docs, text⚠️ One source at a time⚠️ Often limited
Bulk test cycle updates✅ CSV/Excel bulk update❌ Not supported⚠️ Vendor-dependent
Open source & customizable✅ Edit anythingN/A❌ Closed source
CI/CD friendly✅ Pure CLI❌ Manual workflow⚠️ Varies
CostFreeSubscriptionLicense/subscription

The core differentiator: AI generation is just one stage of a connected pipeline — generation, Jira/Zephyr sync, and execution reporting all flow together instead of ending at "here's some text, paste it yourself."


📁 Project Structure

AITestCaseGen/
├── resources/
│   ├── config.ini                    # Jira credentials and settings
│   ├── import_test_case.xlsx         # Input for import / output of AI generator
│   ├── update_test_case.xlsx         # Input for update script
│   ├── delete_Test_case.xlsx         # Input for delete script
│   └── update_test_cycle.xlsx        # Input for test cycle update
├── src/
│   ├── main/
│   │   ├── config.py                 # Reads config.ini
│   │   └── jira_client.py            # Jira / Zephyr Scale REST API client
│   └── test/
│       ├── generate_tc_via_AI.py     # AI test case generator
│       ├── import_testcases.py       # Import test cases to Zephyr Scale
│       ├── update_testcases.py       # Update existing test cases
│       ├── delete_testcases.py       # Delete test cases
│       ├── link_testcases_to_jira.py # Link test cases to Jira issues
│       ├── export_test_cases.py      # Export test cases from Zephyr
│       ├── project_info.py           # Display Jira project information
│       ├── update_test_cycle.py      # Update test cycle execution results
│       └── create_excel_templates.py # Generate blank Excel input templates
└── requirements.txt

⚙️ Prerequisites

  • Python 3.10+
  • Install dependencies:
pip3 install -r requirements.txt
pip3 install transformers accelerate huggingface_hub

🔧 Configuration

Edit resources/config.ini:

[jira]
base_url   = https://your-jira-instance.com
jira_token = YOUR_JIRA_PAT_TOKEN
project_key = YOUR_PROJECT_KEY
project_id  = YOUR_PROJECT_ID

[testcase]
owner_account_id = YOUR_ACCOUNT_ID
test_cycle       = YOUR_CYCLE_KEY

[ai]
hf_token    = YOUR_HUGGINGFACE_API_TOKEN
hf_endpoint = https://your-artifactory-host/artifactory/api/huggingfaceml/huggingface

[defaults]
type_of_test      = Regression Test
automation_status = Manual

🔒 config.ini is in .gitignore — credentials are never committed to git.


🧩 Setup — HuggingFace (Primary)

Step 1 — Install dependencies

pip3 install transformers accelerate huggingface_hub

Step 2 — Configure HuggingFace Token

  1. Get your HuggingFace or Artifactory identity token
  2. Add it to resources/config.ini:
[ai]
hf_token = your-artifactory-identity-token

Step 3 — First Run (model download ~300MB)

The model downloads automatically on first run from your configured hf_endpoint.

Default model: google/flan-t5-base

ModelSizeSpeed
google/flan-t5-small~80MBFastest
google/flan-t5-base~300MBDefault
google/flan-t5-large~800MBMost accurate

🦙 Setup — Ollama (Optional Fallback)

If HuggingFace is unavailable, install Ollama for fully local AI generation:

  1. Download: https://ollama.com/download/windows
  2. Run OllamaSetup.exe
  3. Pull a model:
ollama pull llama3.2

No code changes needed — the script detects Ollama automatically.


📝 Generating Test Cases

From a folder of documents (.docx / .pdf / .txt)

python -m test.generate_tc_via_AI --input-folder "C:\path\to\docs" --folder "/Generated" --count 10

From a Jira issue key

python -m test.generate_tc_via_AI --issue PROJECT-1234 --folder "/Generated" --count 5

From a Jira issue + documents folder (combined)

python -m test.generate_tc_via_AI --issue PROJECT-1234 --input-folder "C:\path\to\docs" --folder "/Generated" --count 10

From a single file

python -m test.generate_tc_via_AI --file resources/requirements.txt --folder "/Generated"

From a Confluence page

python -m test.generate_tc_via_AI --confluence 123456789 --folder "/Generated" --count 10

Pass the page ID (numeric) or the full page URL:

python -m test.generate_tc_via_AI --confluence "https://your-jira.com/wiki/spaces/PROJECT/pages/123456789" --count 10

Multiple pages (comma-separated):

python -m test.generate_tc_via_AI --confluence "123456789,987654321" --folder "/Generated"

How to get the Confluence page ID: Open the page → click ... (More actions) → Page Information → ID is in the URL: .../pages/<pageId>/...

From inline text

python -m test.generate_tc_via_AI --text "User should be able to login with valid credentials" --count 5

All CLI Options

ArgumentDescriptionDefault
--issueJira issue key (e.g. PROJECT-1234)—
--input-folderFolder containing .docx / .pdf / .txt files—
--fileSingle requirements file path—
--textInline requirement text—
--confluenceConfluence page ID or URL (comma-separated)—
--modelHuggingFace or Ollama model namegoogle/flan-t5-base
--countNumber of test cases to generate per source5
--folderZephyr Scale folder path/Generated
--outputOutput .xlsx file pathresources/import_test_case.xlsx

📤 Import Generated Test Cases to Jira

python -m test.import_testcases --file ..\resources\import_test_case.xlsx

🛠️ Other Scripts

Update existing test cases

python -m test.update_testcases --file ..\resources\update_test_case.xlsx

Delete test cases — dry run first, then live

python -m test.delete_testcases --file ..\resources\delete_Test_case.xlsx --dry-run
python -m test.delete_testcases --file ..\resources\delete_Test_case.xlsx

Link test cases to a Jira issue

python -m test.link_testcases_to_jira --file ..\resources\update_test_case.xlsx

Update test cycle execution results

python -m test.update_test_cycle --file ..\resources\update_test_cycle.xlsx

🤝 Contributing

Issues, feature requests, and PRs are welcome. If AITestCaseGen saves you time, consider giving it a ⭐ — it helps others discover the project.

📜 License

MIT License — see LICENSE for details.

Ver o original no GitHub

Bom para

Relacionados