# Jev (TypeSafe AI) next to Amazon Bedrock — companion artifacts

Companion artifacts for the FactualMinds post
**Jev (TypeSafe AI) on AWS: When System One Decisions Belong Next to Bedrock (2026)**.

**Public URLs:** `https://www.factualminds.com/examples/architecture-blog-2026/jev-typesafe-aws/{path}`

## Demo disclaimer

These files are **decision worksheets and a syntax-checkable stub**, not a production deploy, **not** a live TypeSafe call, and **not** an anonymized client engagement.

- No TypeSafe API key is required. The stub never opens a socket.
- Do **not** treat worksheet dollar totals as measured Jev latency or accuracy.
- TypeSafe’s published card used in the CSV: **$0.042 / MTok** input, output free, claimed **70–500 ms** E2E, launch date **15 Sep 2026**. Workflow-eval **193.6× / 444.6×** remain vendor claims with TypeSafe’s own “higher end” caveat.
- First-party platform signals cited in the post (Gateway ~180→95 ms canary, AgentCore TCO silhouette ~$791/mo at 50K sessions) come from earlier published FactualMinds AgentCore posts — **not** from Jev.

## Status snapshot (verify before you build)

Checked **20 Sep 2026** against TypeSafe and AWS public docs:

| Capability | Status |
| ---------- | ------ |
| Jev as a Bedrock model ID / Marketplace listing | **Not published** |
| TypeSafe HTTP `POST /v1/systemone`, model `jev-1.13.0` | Documented — [Models](https://docs.typesafe.ai/models) |
| Python `typesafe_sdk` / JS `@typesafe-ai/sdk` | Documented — [SDKs](https://docs.typesafe.ai/sdk) |
| Pydantic AI `TypeSafeModel` (`typesafe:jev-1.13.0`) | Documented — [Pydantic AI TypeSafe](https://pydantic.dev/docs/ai/models/typesafe/) |
| Jev writes tool arguments | **No** — picks a tool; `ToolCallProposed` + language-model fallback fills args |
| VPC endpoint / PrivateLink to TypeSafe | **None** — egress to `api.typesafe.ai` |
| Fine-tune / LoRA on customer data | **None** |

## Versions (assumed)

Pin and verify before any live call:

| Component | Assumed version / note |
| --------- | ---------------------- |
| Python | 3.12+ |
| Jev | **`jev-1.13.0`** (do not tune thresholds against `jev-latest`) |
| `typesafe_sdk` | current; live path only |
| `pydantic-ai` | `TypeSafeModel` + `FallbackModel` if you add Bedrock args |
| `boto3` | Secrets Manager for `TYPESAFE_API_KEY` when live |
| Bedrock | `Converse` + Guardrails for generation; not a Jev host |

Local stubs compile without TypeSafe or AWS packages.

## Layout

| Path | Purpose |
| ---- | ------- |
| [`lambda_pydantic_ai_stub.py`](lambda_pydantic_ai_stub.py) | Request shape + confidence gate; `DRY_RUN=1` default |
| [`monday-checklist.md`](monday-checklist.md) | Ship-this-week gate list |
| [`decision-matrix.md`](decision-matrix.md) | Jev vs Bedrock classifier vs Guardrails vs AR Checks vs code |
| [`cost-worksheet.csv`](cost-worksheet.csv) | One worked TypeSafe row; Bedrock cells are formulas / pricing-page placeholders |

## Local checks

```bash
python3 -m py_compile lambda_pydantic_ai_stub.py
```

## Related posts

- [Jev + Bedrock guide](https://www.factualminds.com/blog/jev-typesafe-ai-aws-guide-2026/)
- [LangGraph + AgentCore](https://www.factualminds.com/blog/langgraph-aws-bedrock-agentcore-2026/)
- [Bedrock Guardrails](https://www.factualminds.com/blog/how-to-set-up-amazon-bedrock-guardrails-production/)
- [Automated Reasoning Checks](https://www.factualminds.com/blog/amazon-bedrock-automated-reasoning-checks-hallucination-prevention/)
- [AgentCore pricing calculator](https://www.factualminds.com/tools/amazon-bedrock-agentcore-pricing-calculator/)