Back to the catalog

Subdirectories

Bundle OKF 0.2 · 9 conceitos · go-fair-us/ai-blueprint-core

Open source Repository Open in the app JSON README (API)

About

# Subdirectories

* [overview](overview/index.md) - Background, scope, and document status for the NIAID Blueprint
* [audience](audience/index.md) - Guidance for data generators and repository owners
* [implementation](implementation/index.md) - Collaborative development and Discovery Portal integration
* [metadata-schema](metadata-schema/index.md) - Minimal metadata schema — motivation, requirements, and impact
* [persistent-identifiers](persistent-identifiers/index.md) - Persistent identifiers — motivation, requirements, and impact
* [api-specification](api-specification/index.md) - Minimal API specifications — motivation, requirements, and impact
* [citation](citation/index.md) - Citation requirements — motivation, requirements, and impact
* [outreach-training](outreach-training/index.md) - Outreach and training — motivation, requirements, and impact
* [appendix](appendix/index.md) - Supplemental tables, worked examples, and reference material

Details

Kind
OKF bundles
Topic
Developer tools
Publisher
go-fair-us
Origin
okf_github
Category
dados
Version
0.2
Stars
1
Forks
1
Last push
2026-08-24T16:36:08Z
Repository state
ativo
Language
Python
Added
2026-09-08 16:07:08
Updated
2026-09-08 16:07:08
Origin id
go-fair-us/ai-blueprint-core:okf/bundles/niaid_blueprint/index.md

README

# ai-blueprint-core

AI agent tools help NIAID-funded data repositories apply the [NIAID Blueprint for Digital Objects](https://datascience.niaid.nih.gov/resources).

The Blueprint is a FAIR data program from NIAID/ODSET. It defines minimal metadata schemas, persistent identifiers (PIDs), API standards, and citation practices for research data repositories.

This project supplies LLM-driven agents, guided by structured prompt personas. The agents help repository owners and staff assess and apply Blueprint requirements in five areas:

1. **Metadata schema**: schema.org-based metadata elements for digital objects
2. **Persistent identifiers**: DOIs, ORCIDs, RORs, RRIDs, and ontology terms
3. **APIs and machine access**: JSON-LD endpoints, OpenAPI documentation, structured data
4. **Citation guidance**: PID-based citation examples in standard formats
5. **Outreach and training**: Contact Points, training materials, Portal onboarding

## Flipped Interaction Pattern

This is not an agent. It is a pattern that reverses normal chat mode. The language model starts from a prompt and waits for user input to continue the conversation.

An example lives in the `prompts` directory. That example is long and verbose. A shorter, clearer version would help.

| Prompt | Purpose |
|--------|---------|
| `fairAssessmentInterview.md` | Runs a structured 6-phase interview to assess a repository Blueprint alignment and produces a gap report with ranked recommendations |

Paste the document into your prompt. Modern models start the interview and end with a summary of the responses. The exchange can get long. If you want to stop, tell the model: "stop the interview and give me the summary now".

## Usage

This repository has no executable code that you must run. The tooling converts PDF to Markdown. The converted results already live in the `docs` directory.

## Requirements

- Python 3.13+
- [`uv`](https://docs.astral.sh/uv/) for environment and dependency management

## Setup

```bash
# Clone and install
git clone <repo-url>
cd ai-blueprint-core
uv sync
```

## Reference

The authoritative Blueprint specification is in `docs/BluePrint/NIAID_Blueprint_v2_26Sep2025_forExternal.md`. The tools `docling` and `marker-pdf` converted that file from the PDF.

To pass this to a model, use the GitHub raw link: https://raw.githubusercontent.com/go-fair-us/ai-blueprint-core/refs/heads/master/docs/BluePrint/NIAID_Blueprint_v2_26Sep2025_forExternal.md

## Dependencies

- [`docling`](https://github.com/docling-project/docling): structured document parsing and extraction
- [`marker-pdf`](https://github.com/VikParuchuri/marker): PDF-to-Markdown conversion

Both are ML-based libraries. Expect a large `.venv` and model downloads on the first run.

More