Hatch v0.6.0

Hatch v0.6.0 expands Beacon Wilds with a replayable survival mode, richer story combat, and flexible pet imports.

Highlights:

• Play Endless Patrol through deterministic streamed rooms, enemy waves, guardians, and temporary field upgrades.
• Choose Explorer, Adventure, or Expert difficulty and keep separate Story and Endless records for each.
• Use Wisp Bolt, Moon Arc, and Beacon Pulse across the expanded Beacon Rescue story.
• Submit Endless Patrol runs to a real-time public leaderboard with constrained, rate-limited writes.
• Import complete Hatch packages, minimal Hermes/Petdex ZIPs, or loose 1536×1872 PNG/WebP atlases.
• Preserve legacy Beacon Rescue records while migrating to the new per-mode record format.

Validated with 46 Node tests, 13 Python tests, production type-check/build and artifact validation, lint, CLI self-test, checksum verification, and a local production container build.

This is a public-preview prerelease; v0.x interfaces may evolve.

Hatch v0.5.0

Hatch v0.5.0 adds direct Hermes Agent-compatible sprite exports and simplifies repository automation.

Highlights:

• Export a minimal Hermes Agent ZIP directly from the Hatch website or saved-pet history.
• Use --package-format hatch|hermes|both in hatch.py.
• Preserve the existing full Hatch portable-package format.
• Use the current Hermes 8×9 atlas contract without frame conversion.
• Disable routine Dependabot pull requests and general-purpose CI runs.
• Keep tagged and manually dispatched multi-platform container publishing.

Validated with 44 Node tests, 13 Python tests, production build, lint, artifact validation, CLI self-test, Docker build, and an import through the Hermes Agent package importer.

This is a public-preview prerelease; v0.x interfaces may evolve.

Hatch v0.4.0

Hatch v0.4.0 brings the standalone Python CLI to provider parity with the web app.

Highlights:

• Run hatch.py directly with fal, OpenAI, xAI, OpenRouter, Google, ComfyUI, or InvokeAI.
• Load .env.local and .env automatically without replacing exported variables.
• Use provider-specific model environment variables or --model for one-run overrides.
• Reuse the canonical identity as a private reference for all motion and egg edits.
• Normalize bounded base64 provider responses into the local PNG pipeline; xAI JPEG conversion uses Pillow.
• Retry rate limits and transient provider failures during bounded-concurrency runs.
• Keep credentials and generated identity references out of exported pet packages.

Validation includes live generation and reference-edit smoke tests for OpenAI, xAI, and OpenRouter; a complete mocked 27-job cloud pipeline; 41 Node tests; 12 Python tests; production build; lint; and the offline CLI self-test.

This is a public-preview prerelease; v0.x interfaces may evolve.

Hatch v0.3.0

Hatch v0.3.0 expands generation choices across hosted APIs and local GPU engines.

Highlights:

• Use ComfyUI or InvokeAI directly from hatch.amayx.com when the engine allows the Hatch origin through CORS.
• Keep direct endpoint, bearer token and workflow JSON only in the active browser tab.
• Retain the self-hosted Hatch server-proxy mode with mounted workflow defaults.
• Add OpenAI, xAI, OpenRouter and Google image providers alongside fal.
• Bound cloud request and provider response bodies before parsing.
• Preserve provider metadata in portable pet packages and browser history.

This is a public-preview prerelease; v0.x interfaces may evolve.

Hatch v0.2.0

Hatch v0.2.0 adds self-hosted ComfyUI and InvokeAI generation directly to the web app.

Highlights:

• Choose ComfyUI, InvokeAI, or fal from the generator.
• Prefer a configured local engine automatically.
• Use server-mounted workflow defaults or paired browser JSON overrides.
• Keep local endpoints and engine tokens server-side.
• Validate bounded workflows, job IDs, same-origin requests, media types, and PNG output.
• Preserve provider metadata in portable packages and browser history.

The public site at https://hatch.amayx.com/ remains fal-only. Self-host with npm, Docker, or ghcr.io/dralkh/hatch:0.2.0 to connect a private engine.

This is a public-preview prerelease; v0.x interfaces may evolve.

Hatch v0.1.0 — First public preview

Hatch’s first public preview turns one prompt into a consistent, transparent pixel-art pet package with 81 playable frames.

Highlights:

• Live web app at https://hatch.amayx.com/
• Dependency-free Python CLI with fal, ComfyUI, and InvokeAI backends
• Portable sprite atlases, Canvas player, local history, and Beacon Rescue validation game
• Multi-platform GHCR preview image with provenance and SBOM

See CHANGELOG.md for the full release notes.