|
| 1 | +# Docs AI Operating Model (template) |
| 2 | + |
| 3 | +A **docs-as-code + LLM strategy** reference implementation you can fork to bootstrap: |
| 4 | +- a documentation site (MkDocs) |
| 5 | +- a PromptOps library (prompts as code, with review + validation) |
| 6 | +- an evaluation harness (quality gates for AI-assisted writing) |
| 7 | +- a governance pack (policies, risk register, ADRs, metrics) |
| 8 | + |
| 9 | +> **Portfolio note** |
| 10 | +> |
| 11 | +> All content in this repository is **generic and non-proprietary**. It is meant to demonstrate senior-level thinking and execution for documentation leaders and AI-forward technical writers. |
| 12 | +
|
| 13 | +--- |
| 14 | + |
| 15 | +## Why this repo exists |
| 16 | + |
| 17 | +Docs teams are being asked to “use AI” — but the real work is: |
| 18 | +- **deciding** which problems are worth solving with LLMs |
| 19 | +- **controlling risk** (privacy, hallucinations, IP, compliance) |
| 20 | +- **operationalizing quality** (repeatable evaluation, measurable outcomes) |
| 21 | +- **integrating into DocsOps** (CI/CD, style gates, review workflows) |
| 22 | + |
| 23 | +This repo is designed to **show your seniority** by making those decisions visible. |
| 24 | + |
| 25 | +--- |
| 26 | + |
| 27 | +## What it demonstrates |
| 28 | + |
| 29 | +### Docs-as-code (DocsOps) |
| 30 | +- Site authored in Markdown and published with MkDocs |
| 31 | +- CI quality gates: build checks + prose linting |
| 32 | +- A consistent information architecture (strategy → governance → implementation → case studies) |
| 33 | + |
| 34 | +### LLM strategy (operating model) |
| 35 | +- Use-case catalog + decision criteria |
| 36 | +- Guardrails: data classification, risk register, human review lanes |
| 37 | +- Metrics: adoption, quality, efficiency, and customer impact |
| 38 | + |
| 39 | +### PromptOps (prompts as code) |
| 40 | +- Prompt files stored with metadata and versioning |
| 41 | +- Schema validation in CI (no broken prompt definitions) |
| 42 | +- Patterns for prompt structure and output contracts |
| 43 | + |
| 44 | +### Evaluation (what “good” means) |
| 45 | +- Task-based evaluation dataset (small but realistic) |
| 46 | +- Rule-based checks for structure + safety |
| 47 | +- Optional “LLM-as-judge” extension points (vendor-agnostic) |
| 48 | + |
| 49 | +--- |
| 50 | + |
| 51 | +## Quick start (local) |
| 52 | + |
| 53 | +```bash |
| 54 | +# 1) Create and activate a virtual environment |
| 55 | +python -m venv .venv |
| 56 | +source .venv/bin/activate |
| 57 | + |
| 58 | +# 2) Install deps |
| 59 | +pip install -r requirements.txt |
| 60 | +pip install -r requirements-docs.txt |
| 61 | + |
| 62 | +# 3) Validate the prompt library |
| 63 | +python -m docsai_toolkit validate-prompts |
| 64 | + |
| 65 | +# 4) Build the site (strict) |
| 66 | +mkdocs build --strict |
| 67 | + |
| 68 | +# 5) Run a lightweight evaluation pass |
| 69 | +python -m docsai_toolkit eval --dataset eval/datasets/smoke.yml --dry-run |
| 70 | +``` |
| 71 | + |
| 72 | +--- |
| 73 | + |
| 74 | +## Publish the docs site (GitHub Pages) |
| 75 | + |
| 76 | +This repo includes a Pages workflow that: |
| 77 | +1) builds the MkDocs site |
| 78 | +2) uploads it as a Pages artifact |
| 79 | +3) deploys it via `actions/deploy-pages` |
| 80 | + |
| 81 | +In your repo settings: |
| 82 | +- **Settings → Pages → Source → GitHub Actions** |
| 83 | + |
| 84 | +--- |
| 85 | + |
| 86 | +## Repository map |
| 87 | + |
| 88 | +``` |
| 89 | +docs/ # the published site (strategy, governance, implementation) |
| 90 | +prompts/ # prompt library (YAML) |
| 91 | +eval/ # evaluation datasets + rubrics |
| 92 | +adr/ # architecture decision records |
| 93 | +src/docsai_toolkit/ # small CLI utilities (validation + eval scaffolding) |
| 94 | +.github/workflows/ # CI + Pages deployment |
| 95 | +styles/ # Vale prose linting rules (optional but recommended) |
| 96 | +``` |
| 97 | + |
| 98 | +--- |
| 99 | + |
| 100 | +## Suggested ways to personalize (for hiring impact) |
| 101 | + |
| 102 | +- Add a **case study** reflecting your strongest domain (SaaS, MDM, APIs, data governance) |
| 103 | +- Add a **short demo video** (60–90s) showing: PR → CI → Pages → prompt validation → eval report |
| 104 | +- Add a “**Hiring?**” section with your role targets and preferred work style (remote-first, async) |
| 105 | + |
| 106 | +--- |
| 107 | + |
| 108 | +## License |
| 109 | + |
| 110 | +MIT (see `LICENSE`). |
0 commit comments