Skip to content

DATAMIMIC CE v2.2.0

Latest

Choose a tag to compare

@ake2l ake2l released this 09 Nov 03:59
· 1 commit to development since this release
4a10fee

We’re thrilled to announce DATAMIMIC CE v2.2.0, a small follow-up to v2.1 that finalizes the determinism work (domains & seeds) and adds a minimal MCP entry point.
This release keeps your datasets consistent and verifiable—ready for CI, analytics, and agent workflows.


🔥 What’s New in v2.2.0

1️⃣ Determinism Proof (Reproducibility)

Make equality checks explicit across runs.

  • generate responses now include determinism_proof.content_hash.
  • Byte-identical inputs produce byte-identical outputs.

🧩 Benefit: Simple, machine-checkable regression and CI parity.


2️⃣ Frozen Default Clock

Time-dependent values are reproducible by default.

  • Default clock: 2025-01-01T00:00:00Z.
  • Override per request to re-enable wall-clock variance.

3️⃣ MCP Entry Point (Opt-In)

A lightweight IDE/agent path for deterministic tools.

  • New CLI: datamimic-mcp exposing list_domains and generate.
  • Env: DATAMIMIC_MCP_HOST, DATAMIMIC_MCP_PORT, optional DATAMIMIC_MCP_API_KEY.

📘 MCP Quickstart: docs/mcp_quickstart.md


4️⃣ Safety Cap

Guard against runaway jobs.

  • count is limited to 10,000 per request (exceed → 422).

💼 Why It Matters

  • Stable Baselines: Hash + frozen clock lock in reproducible outputs.
  • Agent-Ready: Simple MCP bridge for IDEs and automation.
  • Safer Defaults: Caps and validation reduce CI/agent foot-guns.

⚙️ Upgrade Notes

Determinism Proof

No input changes required. Compare determinism_proof.content_hash across runs to assert equality.

Clock

If you relied on wall-clock variance, pass a clock explicitly in your requests.

Batching

For > 10,000 records, split into multiple calls and merge downstream.

MCP (Optional)

Nothing changes for library/CLI users unless you run datamimic-mcp.
Configure auth with DATAMIMIC_MCP_API_KEY if needed.


🧩 Compatibility

  • Python: 3.11–3.12 (3.14 experimental).