Apppublic

therenald/humanitarianRAG-explore

sourceHugging Faceupdated 4mo agoView on Hugging Face

0likes

App README

Evidence Synthesis — free explorer

The open-exploration version of the prototype (the demo ui/app.py), so people can use the tool freely outside the structured user study. Same retrieval and generation engine and the same document collection; here all the controls are exposed (search mode, retrieval method, top-k, prompt variant).

—Access: gated by the same `STUDY_PASSCODES` codes as the study, so participants reuse their personal code (and the DEMO code works too). Any valid code unlocks it. Unset → no gate (local dev).
—No data is collected here — there is no consent, background, or response logging. It is purely a sandbox.
—Embedding model: BAAI/bge-large-en-v1.5 (downloads on first run).

Space secrets (Settings → Variables and secrets → Secrets)

Name	Value
`STUDY_PASSCODES`	the same JSON code→id map as the study Space (any value unlocks)
`GOOGLE_API_KEY`	your Gemini API key (used by the RAG mode)

(No SPREADSHEET_ID / GCP_SERVICE_ACCOUNT needed — this app saves nothing.)

Deploy

bash

hf repo create humanitarianRAG-explore --repo-type space --space_sdk docker
cd ~/explore_deploy
git init -b main && git lfs install
git add -A && git commit -m "Deploy Evidence Synthesis explorer"
git remote add origin https://huggingface.co/spaces/<your-username>/humanitarianRAG-explore
git push -f origin main

Set the two secrets, then the Space URL (https://<your-username>-humanitarianrag-explore.hf.space) is what the study's done screen links to. If you name the Space differently, update EXPLORE_URL in the study's ui/study_config.py and redeploy the study.

Assembly note

This directory is assembled from the study deploy base (shared indexes, core, src, corpus map) with ui/app.py as the entry point and study-only files (study_app.py, study_config.py, sheets.py, assets/) removed.