ShaleioAI generation (Amazon Bedrock)
Edition: AI dashboard generation and the AI home page are Enterprise Edition features. The Developer Edition omits them — you can still generate dashboards with your own agent workflow against the published schema.
StrataBI Enterprise can turn a plain-language request — "monthly revenue by region with a region filter" — into an editable, schema-conforming dashboard, using Amazon Bedrock running in your own AWS account. Nothing is sent to Shaleio; the model call goes directly from your runtime to Bedrock under your IAM.
What it does
The generator converts an intent into a single valid StrataBI dashboard JSON document, using the dashboard schema and the core LLM instructions as its contract. It produces intent only — a name plus a layout of tiles — and leaves execution, module lookup, and orchestration to core, exactly as described in Agent integration. The result opens in the builder for you to review and edit; the model drafts, you keep control.
Enabling it
AI is off by default and opt-in. Turn it on with the Enable AI toggle in the admin console → Settings, or by setting the environment variable at deploy time. Two prerequisites in your account:
- Bedrock model access — request access to the model you intend to use in the Amazon
Bedrock console (per-account, per-region).
- IAM permission — the runtime's task role needs
bedrock:InvokeModelfor that model.
Configuration
| Setting | Environment variable | Default |
|---|---|---|
| Enable AI | STRATABI_ENABLE_AI | (off) |
| Model id | STRATABI_BEDROCK_MODEL_ID | us.anthropic.claude-sonnet-4-5-20250929-v1:0 |
| Bedrock region | STRATABI_BEDROCK_REGION | falls back to AWS_REGION |
| Max output tokens | STRATABI_AI_MAX_TOKENS | 4096 |
The model id and related settings are also editable live from the AI / Bedrock controls on the admin console → Settings tab. Point STRATABI_BEDROCK_MODEL_ID at any Bedrock model your account is entitled to; the region can differ from your data plane if you enabled Bedrock elsewhere.
Graceful degradation
The feature is designed to fail safe. When AI is disabled, when the account lacks Bedrock access, or when the model returns something that doesn't validate, StrataBI simply does not offer (or quietly drops) the generated result — it never blocks the app or produces an invalid dashboard. Generated output is always validated against the schema before it is offered for editing, and the generator can fall back to an embedded copy of the schema in air-gapped deployments where the public site is unreachable.
Cost & data
Bedrock usage is billed by AWS, directly to your account, per the model's pricing — Shaleio does not meter or mark it up. Because the call runs inside your account, your prompts and generated dashboards stay in your environment; see Security posture and the cost FAQ.