Agentic guide catalogue¶
Start with the published documentation or follow the same guides here on GitHub.
Start with native DI and lifecycle; capability snippets below assume they run within their application runtime activation scope.
See all typed configuration fields for defaults and environment keys.
Choose a path¶
| Goal | Read |
|---|---|
| Install the native Agentic module | Installation |
| Run a real local model | Local Ollama setup and acceptance |
| Run an agent with a tool and memory | Quick start |
| Understand the framework visually | Visual guide, Architecture |
| Select a model, API and typed options | Models, Migration |
| Build a complete application | Tutorial, IDP walkthrough |
| Define application behavior | Agents, Tools, Skills, Prompts, Templates |
| Retain context and retrieve knowledge | Memory, Content, Embeddings, Vector stores |
| Add reasoning and quality checks | Reasoning, Validation, Evaluation |
| Coordinate work | Pipelines, Workflows |
| Operate an application | Observability, Resilience, Storage, Security, Execution, Explainability |
| Compare approaches | Experiments, Lab |
| Run in a notebook | Jupyter |
Examples and their setup requirements live in the example catalogue. Studio has its own repository; its downstream migration and native-runtime compatibility require separate qualification.
Contributing documentation¶
Keep examples at the PyFly Agentic application boundary unless explaining a native integration. State which behavior is automatic, optional or application-managed. Preserve the distinction between Chat Completions and Responses, and between process-local conversation history and persistent working memory.
Use a diagram when relationships or control flow are clearer visually. Keep each figure focused, explain its meaning in adjacent prose, and link to the relevant API guide. Include descriptive alt text and a full-size link for SVGs. Regenerate branded assets through the asset generator, rather than editing their generated SVGs.
Run from the repository root:
python docs/modules/agentic/assets/tools/build_brand_assets.py --check
uv tool run --with-requirements requirements-docs.txt mkdocs build --strict
python scripts/check_agentic_docs.py
uv run --no-project --with playwright==1.64.1 python -m playwright install chromium
uv run --no-project --with playwright==1.64.1 python scripts/check_agentic_docs.py --browser
The static check follows local links, image paths and HTML fragments in the built site. The browser check renders every Mermaid diagram using the site's actual Material integration and fails on missing or invalid diagrams. It also checks keyboard-operated expand/fit controls at desktop and mobile widths. It requires network access for the same CDN resources used by the site. These checks run in the pull-request gate and before Pages deployment.
For local preview:
See CONTRIBUTING for the full development gate. Release versions use YY.MM.Patch; documentation-only updates can deploy independently of a package release. Update mkdocs.yml when adding a page and keep the root README, landing page and guides consistent.