✦ AI Scoping
AI-assisted RI Scoping speeds up pre-filling analyses by automatically generating scenarios from a text description of your facility. It's an optional but effective step before starting the analysis pages.
AI configuration
Before any generation, fill in two parameters in the ⚙ AI Configuration block:
AI engine — configured once for the whole application on /gestion/ia (see AI Configuration), among 8 available engines: Gemini, Claude, Mistral, Groq, HuggingFace, OpenRouter, Ollama (local model, no data leaves your machine) and a custom provider. The most commonly used for RI Scoping:
- Gemini Flash/Pro (Google) — recommended for large scenario lists; free with a generous quota
- Claude Haiku/Sonnet (Anthropic) — better reasoning quality, requires a paid Anthropic account
- Ollama (local) — for sensitive context, runs the model on your machine with no API key and no outgoing data
API key — enter your key in the password field. The key is stored locally in your browser (never sent to EasyRisk servers).
Getting a free Gemini key
- Go to aistudio.google.com
- Sign in with a Google account
- Click Get API key → Create API key
- Copy the key (starts with
AIza…) and paste it into the API key field - The free quota is more than enough for a complete study (hundreds of generations/day)
No credit card required for the free plan.
Getting a Claude key (Anthropic)
- Go to console.anthropic.com
- Create an account and add a payment method (pay-as-you-go billing)
- In API Keys, click Create Key
- Copy the key (starts with
sk-ant-…)
The typical cost for a complete RI study (PHA + FMEA + HAZOP) is on the order of a few cents with Haiku.
Method tabs
Select the method to pre-fill:
| Tab | Status | What is generated |
|---|---|---|
| ⚙ PHA — Physical risks | Available | PHA scenarios (hazard, feared event, causes, consequences, measures) |
| ☣ PHA — Hazard study | Available | Hazard study scenarios (ERC, hazardous phenomena, MMR, PIGC rating) |
| 🔩 FMEA | Available | Failure modes per component (effects, causes, S/O/D rating) |
| 📋 HAZOP | Available | Deviations per node and guide word (causes, consequences, measures) |
| 📊 LOPA | Available | LOPA scenarios with estimated IPLs and PFDs |
| 🔷 Bow Tie | Coming soon | (under development) |
| 🌳 Fault Tree Analysis | Coming soon | (under development) |
Description sheet
Each tab offers a downloadable sheet (.txt) listing the sections to fill in to maximize the quality of the generation:
- ⬇ Download the sheet — retrieves the structured sheet template for the active method
- Load [method] example — inserts a complete, pre-filled example to test the generation without real data
- Clear — empties the text area
Enter at least 400 characters in the description area (indicator shown at the bottom of the area) — below that, the AI lacks context and produces generic scenarios.
Description area
Paste or type the description of your facility into the text area. You can use:
- The content of the downloaded sheet (sections filled in)
- An excerpt from an existing study report
- A free-form description in natural language
Running the generation
- Click ✦ Analyze and pre-fill — [Method]
- A confirmation box appears — it reminds you that the generated scenarios will be added to the existing data (nothing is erased)
- Click Start import to confirm
- Generation takes 10 to 60 seconds depending on the engine and the length of the description (a counter is displayed)
- On success, an import summary is shown with the number of scenarios created and a direct link to the dedicated page
Confidentiality
The text in the description area is sent to the chosen AI service (except with Ollama, where it never leaves your machine). Do not paste real personal data, sensitive industrial secrets or NDA-protected information in a demonstration context.
After generation
Go to the method's page to review the imported scenarios:
- Complete empty or incomplete fields
- Correct ratings if they don't match the project's metrics
- Remove irrelevant scenarios
- Add specific scenarios the AI may not have covered
Generation quality
The more precise the description (named equipment, pressures, temperatures, materials, existing barriers), the better the scenarios. The example provided in each tab illustrates the expected level of detail.