Spaces:
Paused
Paused
Upload Readme.md
#4
by Basementup - opened
Readme.md
ADDED
|
@@ -0,0 +1,86 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
title: Dr Drastic Unified Evidence Engine
|
| 3 |
+
emoji: "\U000026A1"
|
| 4 |
+
colorFrom: blue
|
| 5 |
+
colorTo: green
|
| 6 |
+
sdk: gradio
|
| 7 |
+
sdk_version: 6.18.0
|
| 8 |
+
python_version: "3.12"
|
| 9 |
+
app_file: app.py
|
| 10 |
+
fullWidth: true
|
| 11 |
+
pinned: false
|
| 12 |
+
license: mit
|
| 13 |
+
---
|
| 14 |
+
|
| 15 |
+
# Dr Drastic: Unified Evidence Engine
|
| 16 |
+
|
| 17 |
+
This project fuses the strongest parts of the two original Gradio bots into one
|
| 18 |
+
testable application:
|
| 19 |
+
|
| 20 |
+
- PDF, DOCX, TXT, Markdown and CSV ingestion
|
| 21 |
+
- role classification and source IDs
|
| 22 |
+
- pasted or document-derived chronology extraction
|
| 23 |
+
- event tagging and evidence-pack completeness scoring
|
| 24 |
+
- potential contradiction and omission review
|
| 25 |
+
- source-linked evidence matrix
|
| 26 |
+
- legal and regulatory route suggestions
|
| 27 |
+
- grounded Queen Nala AI rulings with streaming responses and source-ID citations
|
| 28 |
+
- configurable Hugging Face inference model/provider and session knowledge base
|
| 29 |
+
- Markdown ruling transcript export
|
| 30 |
+
- text-to-speech for key panels
|
| 31 |
+
- TXT, JSON, DOCX and ZIP exports
|
| 32 |
+
- per-session Gradio state instead of a process-wide global result
|
| 33 |
+
|
| 34 |
+
## Run
|
| 35 |
+
|
| 36 |
+
```powershell
|
| 37 |
+
python -m pip install -r requirements.txt
|
| 38 |
+
python app.py
|
| 39 |
+
```
|
| 40 |
+
|
| 41 |
+
This workspace already includes a local virtual environment:
|
| 42 |
+
|
| 43 |
+
```powershell
|
| 44 |
+
.\.venv\Scripts\python.exe app.py
|
| 45 |
+
```
|
| 46 |
+
|
| 47 |
+
For AI rulings, set a Hugging Face token locally or add `HF_TOKEN` as a Space
|
| 48 |
+
secret:
|
| 49 |
+
|
| 50 |
+
```powershell
|
| 51 |
+
$env:HF_TOKEN = "hf_..."
|
| 52 |
+
```
|
| 53 |
+
|
| 54 |
+
## Hugging Face Pro
|
| 55 |
+
|
| 56 |
+
With a **Hugging Face Pro** account, Queen Nala automatically prefers
|
| 57 |
+
`hf-inference` then `auto` routing for better HF router access and higher
|
| 58 |
+
limits. Pro status is detected once per process via `whoami` when a token is
|
| 59 |
+
available.
|
| 60 |
+
|
| 61 |
+
| Environment | Token setup |
|
| 62 |
+
|-------------|-------------|
|
| 63 |
+
| Hugging Face Spaces | Space secret `HF_TOKEN` (Inference Providers permission) |
|
| 64 |
+
| GitHub Actions / Codespaces | Repository or user secret `HF_TOKEN` or `HUGGING_FACE_HUB_TOKEN` |
|
| 65 |
+
| Local / Docker | `$env:HF_TOKEN = "hf_..."` or `hf auth login` |
|
| 66 |
+
|
| 67 |
+
The Queen Nala tab shows a subtle routing hint when Pro is active. Pro users
|
| 68 |
+
get fewer provider hops and retries to avoid retry storms on transient outages.
|
| 69 |
+
|
| 70 |
+
### Docker
|
| 71 |
+
|
| 72 |
+
```powershell
|
| 73 |
+
docker build -t dr-drastic .
|
| 74 |
+
docker run --rm -p 7860:7860 -e HF_TOKEN=hf_... dr-drastic
|
| 75 |
+
```
|
| 76 |
+
|
| 77 |
+
## Test
|
| 78 |
+
|
| 79 |
+
The core tests do not require Gradio or document-export packages:
|
| 80 |
+
|
| 81 |
+
```powershell
|
| 82 |
+
python -m unittest discover -s tests -v
|
| 83 |
+
```
|
| 84 |
+
|
| 85 |
+
Automated findings are review leads, not legal conclusions or legal advice.
|
| 86 |
+
Verify every output against the original source documents and current law.
|