https://github.com/mahmood726-cyber/rapidmeta-kit
Full interactive RapidMeta meta-analysis dashboard from a tiny JSON config: edit ~20 lines, run one command, get the complete workbench (screening, extraction, 28-panel stats, GRADE, WebR, PRISMA) running offline. Built for low-API-quota use (Gemini CLI free tier).
https://github.com/mahmood726-cyber/rapidmeta-kit
evidence-synthesis meta-analysis offline-first rapidmeta systematic-review
Last synced: about 20 hours ago
JSON representation
Full interactive RapidMeta meta-analysis dashboard from a tiny JSON config: edit ~20 lines, run one command, get the complete workbench (screening, extraction, 28-panel stats, GRADE, WebR, PRISMA) running offline. Built for low-API-quota use (Gemini CLI free tier).
- Host: GitHub
- URL: https://github.com/mahmood726-cyber/rapidmeta-kit
- Owner: mahmood726-cyber
- License: mit
- Created: 2026-05-28T08:57:15.000Z (about 2 months ago)
- Default Branch: main
- Last Pushed: 2026-06-14T16:43:33.000Z (about 1 month ago)
- Last Synced: 2026-06-14T17:13:02.540Z (about 1 month ago)
- Topics: evidence-synthesis, meta-analysis, offline-first, rapidmeta, systematic-review
- Language: HTML
- Size: 4.6 MB
- Stars: 0
- Watchers: 0
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
- Agents: AGENTS.md
Awesome Lists containing this project
README
# rapidmeta-kit
[](https://github.com/mahmood726-cyber/rapidmeta-kit/actions/workflows/ci.yml)
[](LICENSE)
[](https://mahmood726-cyber.github.io/rapidmeta-kit/)
> **▶ Try it live:** — a
> complete RapidMeta dashboard for finerenone in CKD, generated by this kit
> and served straight from `docs/`. Click through the Screening, Extraction,
> and Analysis Suite tabs.
Spin up the **full interactive RapidMeta dashboard** for any topic by editing
a small JSON file and running one command. You get the complete workbench —
protocol, search, screening, risk-of-bias, editable extraction, live
re-pooling, a 28-panel statistics tab, GRADE Summary of Findings, WebR
R-validation, PRISMA flow — all running in the browser, offline.
The heavy engine (~1 MB) and its stat-panel scripts are **pre-built and
bundled**. The only thing that changes per review is a ~20-line config, so an
AI CLI (Gemini, Claude, etc.) on a free / rate-limited key can produce many
dashboards without burning quota.
> Want a *lightweight* static one-page meta-analysis instead (tiny output, no
> interactive engine)? See the sister project
> **[meta-starter-kit](https://github.com/mahmood726-cyber/meta-starter-kit)**.
> This repo is the *full* interactive version.
---
## ⬇️ Download (one click)
**[Download the whole kit as a ZIP »](https://github.com/mahmood726-cyber/rapidmeta-kit/archive/refs/heads/main.zip)**
Unzip it. Or grab a versioned copy from the
[Releases page](https://github.com/mahmood726-cyber/rapidmeta-kit/releases).
## ▶️ Use it in 3 steps
1. **Install Python** (once): — on
Windows tick **"Add Python to PATH"** during install.
2. **Run the example.**
- **Windows:** double-click **`RUN_EXAMPLE.bat`**.
- **Mac/Linux:** `bash run_example.sh`.
- **Any terminal:** `python clone.py configs/example_finerenone_ckd.json`
3. **Open** `output/finerenone_ckd.html` in a browser. That is a complete,
interactive RapidMeta dashboard for finerenone — click through Screening,
Extraction, and the Analysis Suite tabs.
## Make your own
1. Copy an example config:
```
copy configs\example_finerenone_ckd.json configs\my_review.json (Windows)
cp configs/example_finerenone_ckd.json configs/my_review.json (Mac/Linux)
```
2. Edit `configs/my_review.json` — set the drug, condition, and trials
(see [Config](#config)).
3. Build:
```
python clone.py configs/my_review.json
```
4. Open `output/.html`.
## Config
Minimum:
```json
{
"drug": "Finerenone",
"slug": "finerenone_ckd",
"condition": "CKD in type 2 diabetes",
"title": "Finerenone for CKD — Living Meta-Analysis",
"trials": [
{ "nct": "NCT02540993", "name": "FIDELIO-DKD", "year": 2020,
"tE": 367, "tN": 2833, "cE": 420, "cN": 2841 },
{ "nct": "NCT02545049", "name": "FIGARO-DKD", "year": 2021,
"tE": 458, "tN": 3686, "cE": 519, "cN": 3666 }
]
}
```
- `tE`/`tN` = events/total in the intervention arm; `cE`/`cN` = comparator arm.
- Or give a published effect instead: `"publishedHR": 0.86, "hrLCI": 0.75, "hrUCI": 0.99`.
- Optional extras: `comparator`, `hero_h2`, `nyt_headline`, `pico`,
`acronyms`, per-trial `pmid` / `phase` / `group` / `allOutcomes`.
- See `configs/example_finerenone_ckd.json` for a fully-populated example.
The build prints a `[CONFIG ERROR]` naming exactly what to fix if a field is
missing or malformed.
## How it works
`clone.py` stamps your config onto `template/base_dupilumab_copd.html` (the
validated RapidMeta base, unminified) — it swaps the drug/condition tokens
and replaces the trial dataset — then copies the shared `assets/` folder
(stylesheet, plotly, and the 28 stat-panel scripts) next to your output so it
runs offline. No statistics are recomputed by an LLM; the dashboard does all
analysis in the browser.
## Offline notes
- Everything runs client-side and works offline **once downloaded**, except:
- **Icons** use a Font Awesome CDN (cosmetic; the page is fully usable
without them).
- The live **CT.gov / PubMed lookups** inside the Search tab need internet
(optional — you can paste/extract data manually).
- Network-meta-analysis plots use the bundled Plotly (`assets/plotly.min.js`)
— no internet needed.
## What NOT to edit
Only ever create/edit files in `configs/`. `template/` (engine + `assets/`)
and `clone.py` are finished. `output/` is regenerable.
## Development
The kit is pure-Python stdlib at runtime — `clone.py` has **zero** dependencies
and makes no network calls. Tests and JSON-schema validation are dev-only:
```
pip install -e ".[dev]" # pytest + jsonschema
python -m pytest # unit + end-to-end build tests
```
The suite covers the JS-literal rendering (including apostrophe escaping, which
would otherwise break a dashboard), config validation and exit codes, and a
full subprocess build of both bundled example configs. CI runs it on Python
3.9 / 3.11 / 3.13 and smoke-builds the example dashboard
(`.github/workflows/ci.yml`).
The live demo under `docs/` is regenerable at any time:
```
python clone.py configs/example_finerenone_ckd.json --out docs/index.html
```
## Honesty
Each dashboard is an **auto-generated draft**. The kit wires your trial data
into a working analysis app; it cannot judge whether the trials are
clinically combinable or whether the extracted numbers are correct. Verify
inputs against source publications before using or citing any result.