Spaces:
Sleeping
Sleeping
File size: 8,434 Bytes
78c1c50 2f024ae 78c1c50 2f024ae 78c1c50 2f024ae 78c1c50 2f024ae 78c1c50 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 | # Website Maintenance and New Case Study Guide
This guide explains how to maintain the RFR Knowledge Platform website and how
to add new managed retreat case studies to the searchable knowledge base.
## Repositories and Live Website
- Shared GitHub repo:
`https://github.com/rfrraguwaterloo-cmyk/UWaterlooManagedRetreatChatbot`
- Shared Hugging Face Space:
`https://huggingface.co/spaces/UWRFR/UWaterlooRFRChatBot`
- Live app:
`https://uwrfr-uwaterloorfrchatbot.hf.space`
GitHub is the easiest shared place for code review. Hugging Face is where the
Streamlit app runs.
## Editing the Website From Another Computer
1. Install Git and Python 3.11 or newer.
2. Clone the shared GitHub repo:
```bash
git clone https://github.com/rfrraguwaterloo-cmyk/UWaterlooManagedRetreatChatbot.git
cd UWaterlooManagedRetreatChatbot
```
3. Create a working branch:
```bash
git checkout -b update-website-text
```
4. Install dependencies:
```bash
python3 -m venv venv
source venv/bin/activate
pip install -r requirements.txt
```
5. Run the website locally:
```bash
streamlit run app/app.py
```
6. Edit the code. Most website text and tabs are in:
```text
app/app.py
```
7. Check the app still compiles:
```bash
python3 -m py_compile app/app.py
```
8. Commit and push:
```bash
git status
git add app/app.py
git commit -m "Update website text"
git push origin update-website-text
```
9. Open a Pull Request on GitHub into `main`.
After the change is merged into `main`, the GitHub Action can deploy to Hugging
Face if the shared repo has `HF_TOKEN` and `HF_SPACE_ID` configured. The team can
also push manually to Hugging Face from a machine that is logged in with `hf`.
## Manual Hugging Face Push
From a local repo with the Hugging Face CLI logged in:
```bash
git remote add uwrfr-chatbot https://huggingface.co/spaces/UWRFR/UWaterlooRFRChatBot
git push uwrfr-chatbot HEAD:main
```
If the remote already exists, update it instead:
```bash
git remote set-url uwrfr-chatbot https://huggingface.co/spaces/UWRFR/UWaterlooRFRChatBot
git push uwrfr-chatbot HEAD:main
```
Or upload the current folder with the CLI:
```bash
hf upload UWRFR/UWaterlooRFRChatBot . \
--type space \
--exclude ".git/*" \
--exclude ".github/*" \
--exclude ".venv/*" \
--exclude "venv/*" \
--exclude "__pycache__/*" \
--exclude "data/raw/*" \
--exclude "*.pdf" \
--commit-message "Update website"
```
Check Space status:
```bash
hf spaces info UWRFR/UWaterlooRFRChatBot
```
## Website Secrets
Set these in Hugging Face Space settings under **Variables and secrets**.
Required for the chatbot:
```text
ANTHROPIC_API_KEY
```
Optional if using OpenAI mode:
```text
OPENAI_API_KEY
OPENAI_MODEL
```
Required for Google Drive case-study request uploads:
```text
GOOGLE_OAUTH_CLIENT_ID
GOOGLE_OAUTH_CLIENT_SECRET
GOOGLE_OAUTH_REFRESH_TOKEN
```
Optional Drive override for new case study requests:
```text
RFR_REQUESTS_DRIVE_FOLDER_ID
```
By default, new case study requests are uploaded to this Drive folder:
```text
10n_-uOCT2GXu_G3r1qi5qJr8hHoy-E_y
```
The OAuth token should be generated by signing in as the Google account that
should own request uploads, usually `rfr.rag.uwaterloo@gmail.com`.
Generate the Hugging Face OAuth secrets locally:
```bash
python3 -m case_study_pipeline.create_drive_oauth_secrets \
--credentials-file /path/to/oauth-client.json \
--space-id UWRFR/UWaterlooRFRChatBot
```
The older fallback `GOOGLE_SERVICE_ACCOUNT_JSON` service-account setup can fail
in regular My Drive folders because service accounts do not have storage quota.
Use OAuth or a Google Shared Drive for request uploads.
## How New Case Study Requests Work
Users can submit candidate papers through the **New Case Study Requests** tab.
The app creates a new Google Drive folder for each request. Each folder contains:
- `request.json`
- uploaded PDFs or supporting documents
The request tab only collects files and notes. It does not run the extraction
pipeline and does not call an LLM.
## Adding a New Case Study to the Knowledge Base
Use this workflow after the team accepts a submitted request or identifies a new
case study.
### 1. Choose the Next Case ID
Check the highest indexed case:
```bash
ls data/extracted/CS*.json | sort -V | tail
```
If the latest indexed case is `CS54`, the next one is usually `CS55`.
### 2. Create the Raw Case Folder
Create a folder under `data/raw/`:
```bash
mkdir -p data/raw/CS55-ShortName_COUNTRY
```
Put source PDFs, `.md`, or `.txt` files in that folder.
Add or confirm `case_meta.json` in that folder. Example:
```json
{
"case_id": "CS55",
"name": "Short case study name",
"location": "City or region",
"country": "Country"
}
```
### 3. Run the One-Provider Extraction Pipeline
Use one provider for all four extraction and audit steps. Claude is the current
simplest option:
```bash
python3 -m case_study_pipeline.run_case_study \
--case-folder data/raw/CS55-ShortName_COUNTRY \
--case-id CS55 \
--llm-provider claude \
--max-chars-per-source 150000 \
--force
```
This writes outputs to:
```text
data/raw/CS55-ShortName_COUNTRY/pipeline_output/
```
Expected output files:
- `CS55_Ver1.md`
- `CS55_Check1_report.md`
- `CS55_Ver2.md`
- `CS55_Check2_report.md`
- matching PDFs
### 4. Select the Best Run
```bash
python3 -m case_study_pipeline.select_best_run --case-id CS55
```
This creates or updates:
```text
data/raw/CS55-ShortName_COUNTRY/pipeline_output/selected.json
```
### 5. Ingest the Case Study
```bash
python3 ingest/ingest_pipeline_outputs.py --case-id CS55
```
This creates:
```text
data/extracted/CS55.json
```
### 6. Rebuild Summaries and Embeddings
Always run both commands after adding or changing extracted case data:
```bash
python3 ingest/create_summary_chunks.py
python3 ingest/embed_and_index.py
```
These update:
```text
data/extracted/case_summaries.json
data/extracted/precomputed_embeddings.json
```
The live website will not know about the new case until these files are updated
and deployed.
### 7. Test Locally
```bash
streamlit run app/app.py
```
Check:
- the new case appears in **Case Studies**
- the app can answer a direct question such as `Summarize CS55`
- source links appear if `source_links` metadata is available
### 8. Commit and Push
Commit only the files needed by the app. Usually this means:
```bash
git add data/extracted/CS55.json
git add data/extracted/case_summaries.json
git add data/extracted/precomputed_embeddings.json
git add data/raw/CS55-ShortName_COUNTRY/case_meta.json
git commit -m "Add CS55 case study"
```
The raw source PDFs are usually not committed. Keep them in Drive or local raw
storage unless the team explicitly decides otherwise.
Push to GitHub:
```bash
git push shared-github HEAD:main
```
Push to Hugging Face:
```bash
git push uwrfr-chatbot HEAD:main
```
## Updating an Existing Case Study
If a case study already exists and you are improving its sources or extraction:
```bash
python3 -m case_study_pipeline.run_case_study \
--case-folder data/raw/CS55-ShortName_COUNTRY \
--case-id CS55 \
--llm-provider claude \
--max-chars-per-source 150000 \
--force
python3 -m case_study_pipeline.select_best_run --case-id CS55
python3 ingest/ingest_pipeline_outputs.py --case-id CS55
python3 ingest/create_summary_chunks.py
python3 ingest/embed_and_index.py
```
Then test, commit, and push as above.
## Common Problems
### New case does not show on the website
Re-run:
```bash
python3 ingest/create_summary_chunks.py
python3 ingest/embed_and_index.py
```
Then commit and deploy the updated files in `data/extracted/`.
### Hugging Face rebuilds but old content appears
Check the Space commit SHA:
```bash
hf spaces info UWRFR/UWaterlooRFRChatBot
```
Confirm it matches the commit you pushed.
### Drive request upload fails
Check:
- `GOOGLE_OAUTH_CLIENT_ID`, `GOOGLE_OAUTH_CLIENT_SECRET`, and
`GOOGLE_OAUTH_REFRESH_TOKEN` exist as Hugging Face secrets
- the Drive folder is owned by or shared with the OAuth Google account
- `RFR_REQUESTS_DRIVE_FOLDER_ID` points to the correct folder
- the Space has restarted after secrets or variables were added
### LLM key fails
Check the relevant Hugging Face secret:
```text
ANTHROPIC_API_KEY
OPENAI_API_KEY
```
The app currently defaults to Claude for answer generation. The extraction
pipeline can use one provider with `--llm-provider claude` or
`--llm-provider openai`.
|