Skip to content

Releases

v0.1.3 — Private JSON credentials

Choose a tag to compare

@seekskyworld seekskyworld released this 17 Nov 08:38

Studio login now uses private local JSON credentials, so CLI and MCP can share a session without macOS Keychain, Windows Credential Manager or Linux Secret Service. Sessions, signed artifact links, upload references and optional LLM keys use the same local storage layer.

  • Protect files with POSIX permissions or Windows ACLs, preserve profile/account binding, and keep credential updates atomic. Old keychain entries are not read, migrated or deleted automatically.
  • Fix Windows data-directory and relative-import paths; add an optional Windows npm/MCP launch example.
  • Fix a login/exit race: a long synchronous permission check can no longer let an older login restore a session after logout.
  • Preserve the CLI → Studio execution layer, MCP tools, and the progressively loaded Skill.

Validation: 1,822 tests passed; one Windows-native ACL test was skipped on macOS. The exact attached TGZ passed strict installation and local npm/global npm/npx, Skill, MCP, JSON persistence/CAS and Sharp checks on macOS arm64 and Linux arm64. Dependency graphs and installed bytes match the reviewed package. Live macOS CLI and MCP authentication reused the same JSON session across separate data directories without another login prompt.

Windows implementation is included, but native acceptance remains pending: the check workflow did not start because GitHub reported an account billing lock. The attached readiness report retains both missing Windows checks; this release does not claim full three-platform verification.

Public-package acceptance: a fresh npm installation from the public GitHub archive completed image generation → image-to-model → download through both the installed Skill/CLI and the default MCP server using a standard MCP SDK client. Both routes produced real 1024×1024 PNGs and textured GLBs; hashes, geometry and 2048×2048 textures were independently validated, and the results appeared in the Studio asset library. This does not claim every Agent host or operating system has completed live-account acceptance.

npm version 0.1.3 is publicly available. The registry tarball was downloaded and matched to the original attached runtime archive by size, SHA-256, SHA-512 and SHA-1. Public-registry installation, Skill installation, CLI session reuse and MCP discovery/read-only account access were subsequently verified. The attached publication receipts preserve the original submission and verification; the package was not uploaded again.

npm install --global @seekskyworld/ai-modeling@0.1.3

The attached GitHub runtime archive remains an alternative installation source. Then choose Skill or MCP in the README. The Skill ZIP contains guidance and requires the runtime separately. Existing Windows task data can be selected with AI_MODELING_DATA_DIR; upgrading does not erase existing tasks.

Source: 70c64ca0286c3f90b8eecec770d673e4be27ba1d. Runtime SHA-256: 43bf10f8472bd6c15a8ac9f80180e1530c1b8c055c505973be98eb57a4b85213.

Reviewed Studio pricing

The source JSON at the fixed commit below was checked byte-for-byte against its public URL. Keep the recorded expiry; matching hashes do not extend the review or guarantee upstream billing.

ai-modeling pricing update \
  "https://raw.githubusercontent.com/seekskyworld/ai-modeling/70c64ca0286c3f90b8eecec770d673e4be27ba1d/studio/api/reviewed-prices.json" \
  --sha256 ff826078d351ddd6fcf993fe17b85e41425dc06d83d8d0bfa06060726acd9e20
ai-modeling pricing status

ai-modeling 0.1.2

Choose a tag to compare

@seekskyworld seekskyworld released this 17 Nov 06:39

ai-modeling 0.1.2

Clarifies authentication recovery after a completed browser login. A stored login-flow result no longer implies a freshly verified credential store or Studio session: status responses identify their scope and verification state, and the MCP/Skill guides bind continuation to the original login ID and request.

This patch also includes concrete image-size and model-parameter examples. It retains the global-install and native-dependency fixes from 0.1.1.

npm install --global @seekskyworld/ai-modeling@0.1.2
ai-modeling --version
ai-modeling doctor --offline --json

The npm publication was accepted, but the exact public version still returned 404 at 2026-11-17 07:00 UTC. Public-registry installation of 0.1.2 remains unverified until indexing completes; 0.1.1 is available. The attached 0.1.2 runtime TGZ is an independently verified installation source: verify SHA256SUMS, then run npm install --global ./ai-modeling-0.1.2.tgz. Do not republish this immutable version while it is processing.

Use the README to choose Skill or MCP. The default MCP set includes six tools and GPT Image 2.5 image generation. MCP invokes the packaged CLI; Studio operations use your own web account and credits.

Verification

  • 1,813 integrated tests passed with zero skips. Type checking, lint, formatting, documentation, build and package inventory passed. Production dependency audit: zero vulnerabilities.

  • The exact attached TGZ passed strict archive installation, local npm/npx, genuine global npm installation, CLI, MCP, Skill management, PNG encode/decode and native Keyring module loading on macOS arm64 and Linux arm64. Credential-store reads are a separate live check.

  • npm/npx matched 185 reviewed dependencies and 11,118 bundled files on each platform. Global checks matched 185 packages and 11,211 dependency files against the separately verified local installation.

  • Independent review confirmed that login-flow resources still avoid credential/network reads and preserve original request recovery. A stale completed login and unavailable credentials are covered by regression tests.

  • Fresh public npm 0.1.1 installation and explicit Agent Skill use, image/model dry runs, and npx MCP discovery passed before this patch. Version 0.1.2 public-registry first-use remains pending: two bounded check windows returned 404 in all 23 checks. Older results are not relabeled as new-version tests.

  • All 11 public GitHub assets were downloaded anonymously and matched their expected sizes and SHA-256 digests. The public 0.1.2 TGZ also passed a fresh npx MCP connection, six-tool discovery, GPT Image 2.5 default, guide reading, and login-flow status verification against an existing historical login record. This archive route is separate from registry installation.

Remaining validation limits

  • Windows has not been validated. GitHub Actions did not start for this commit because the repository account is locked due to a billing issue. Attached readiness retains the missing Windows reports; this release does not claim three-platform acceptance.
  • Live image/model generation remains pending because the current machine cannot read its OS-stored Studio credentials. The bounded request returns CREDENTIAL_STORE_UNAVAILABLE. No generation was submitted and no Studio credits were spent in these blocked release tests.
  • Current Agent calls through an MCP SDK transport and explicit Skill reading are distinct from an independent host automatically discovering the integration and completing a paid workflow. The separate Claude Code login used in earlier tests also needs renewal.
  • Golutra Agent 0.3.9 filters native MCP tools. Use its Skill/CLI route until the host limitation is fixed. Complete official-CLI or Studio-website equivalence is not claimed; see current limitations.

Integrity and price source

Source commit: e24941f59520ec1f424e543504a38261bc5a7682.

Runtime TGZ SHA-256: 7f63a0507a48788cf4559c3c120254c93820345b2d8b57715493dee7b93b97ec.

The original distribution files and actual platform reports are attached. The runtime archive is byte-identical to the one tested on both platforms.

Reviewed price JSON at this commit

Price file SHA-256: ff826078d351ddd6fcf993fe17b85e41425dc06d83d8d0bfa06060726acd9e20. Reviewed 2026-11-17; expires 2026-11-24 04:25:59 UTC. Estimates exclude promotions; Studio billing is authoritative.

ai-modeling 0.1.1

Choose a tag to compare

@seekskyworld seekskyworld released this 17 Nov 05:38

ai-modeling 0.1.1

Fixes the first-use global npm installation failure in 0.1.0. Sharp and Keyring now install with their platform dependencies, and image processing loads its native runtime only when needed. Version/help, Skill management and MCP discovery can start independently of the image-processing binary.

npm install --global @seekskyworld/ai-modeling@0.1.1
ai-modeling --version
ai-modeling doctor --offline --json

Version 0.1.1 is publicly available from npm. Its registry tarball was downloaded anonymously and matched the tested archive by SHA-256, SHA-512 integrity and SHA-1. A clean registry global installation, Skill installation and dry runs, plus a separate npx MCP connection, passed. The attached ai-modeling-0.1.1.tgz remains an alternative: verify SHA256SUMS, then run npm install --global ./ai-modeling-0.1.1.tgz. Do not install the temporary 0.0.0-stage placeholder or the defective 0.1.0 global installation.

Use the README to choose Skill or MCP. One runtime serves all three interfaces; MCP invokes the CLI. The default MCP set includes six tools and GPT Image 2.5 image generation. Studio operations use your own web account and credits.

Verification

  • 1,812 integrated tests passed with zero skipped tests. Type checking, lint, formatting, documentation, build and the package inventory passed. The production dependency audit reported zero vulnerabilities.

  • The exact attached TGZ passed on macOS arm64 and Linux arm64 through locked installation, local npm/npx, and genuine global npm installation. Global checks exercise the CLI, MCP, Skill management, real PNG encode/decode, and native Keyring module loading without reading credentials.

  • Local npm and npx match 185 reviewed dependencies and 11,118 bundled files. Global dependency files are compared against the separately integrity-verified local installation because global npm does not produce a hidden lockfile.

  • The new native-runtime test rejects the original 0.1.0 installation defect. Current platform results and remaining release-readiness requirements are attached without rewriting earlier failures.

  • All 11 public GitHub Release assets were downloaded anonymously and their hashes checked. A separate clean global install of the public TGZ passed SDK discovery of the six default MCP tools and usage-guide reading without task-state creation.

  • A fresh Codex sub-agent followed the public README, installed the project Skill, read its guidance, and validated image and model dry runs. This verifies explicit Skill use, not an independent host's automatic discovery or live generation. The online README now includes concrete -p examples checked against 0.1.1.

Remaining validation limits

  • Windows remains unverified. Hosted CI could not start because of the repository account's Actions configuration. The attached readiness result retains the missing Windows reports; this release does not claim three-platform acceptance.
  • Live Agent image-to-model acceptance is pending. The test environment cannot currently read its OS-stored Studio credentials, and the independent Agent client's login needs renewal. The CLI now fails within a bounded interval with recovery guidance. No new generation was submitted during these blocked tests.
  • Skill instructions, SDK discovery and local installation checks are distinct from an independently authenticated AI client completing a paid workflow.
  • Golutra Agent 0.3.9 currently filters native MCP tools; use its Skill/CLI route until the host limitation is fixed. Complete official CLI or website equivalence is not claimed; see current limitations.

Integrity and price source

Source commit: 575841d17eefd1acba8c73e5fe8c780ae4821b80.

Runtime TGZ SHA-256: ef89b9fb7feaae009f660bfe32aae51880ebb217604d1093bdd3513b67a1e798.

The six original distribution files are attached without rebuilding. Additional platform reports preserve the actual validation boundary.

Reviewed price JSON at the source commit

Price file SHA-256: ff826078d351ddd6fcf993fe17b85e41425dc06d83d8d0bfa06060726acd9e20. Reviewed 2026-11-17; expires 2026-11-24 04:25:59 UTC. Estimates exclude promotions; Studio billing is authoritative. Updating prices requires reviewed evidence and a trusted checksum.

ai-modeling 0.1.0

ai-modeling 0.1.0 Pre-release
Pre-release

Choose a tag to compare

@seekskyworld seekskyworld released this 17 Nov 05:09

ai-modeling 0.1.0

Known installation defect: a fresh global npm install can omit native optional dependencies, causing even ai-modeling --version to fail while loading Sharp. The local-install and npx checks below passed but did not cover this global path. A patch release is being prepared; do not treat this version as ready for a first global install.

Use Tripo Studio through one CLI runtime, an Agent Skill, or MCP tools. Studio operations use your own web account and credits. This is an independent community project under Apache-2.0.

The npm registry has accepted this version and is processing the new package. If the pinned npm version is not yet available, download the attached ai-modeling-0.1.0.tgz, verify SHA256SUMS, and install it with npm install --global ./ai-modeling-0.1.0.tgz. Do not install the temporary 0.0.0-stage placeholder.

npm install --global @seekskyworld/ai-modeling@0.1.0
ai-modeling doctor --offline --json
  • CLI execution with task recovery and caller-selected output paths.
  • One progressively loaded Skill, with explicit client targets and project/user installation.
  • MCP invokes the same CLI; the default set has six tools, including standalone image generation with GPT Image 2.5. Use --toolset studio for broader asset and editing operations.
  • Browser login and OS credential storage, with bounded credential access instead of indefinite hangs.
  • Reviewed portable runtime dependencies are bundled; exact optional native packages install for the current platform.

See the README for Skill and MCP installation and the compatibility guide for feature differences.

Validation and known limits

  • 1,804 local tests passed with no skipped tests; type checking, lint, formatting, documentation and build passed. Production dependency audit reported zero vulnerabilities at release time.
  • This exact runtime archive passed locked installation and npm/npx consumer checks on macOS arm64 and Linux arm64. Each path matched 185 installed dependencies and verified 11,172 bundled dependency files.
  • Windows validation remains pending. Hosted CI could not start because of the repository account's Actions configuration. Attached readiness data correctly remains ready: false for the two missing Windows reports; this release does not claim three-platform acceptance.
  • Real Agent image-to-model acceptance from the public npm package is pending. The local OS credential store is currently unavailable; the fixed helper returns a bounded error after approximately 10 seconds without starting login or submitting a generation. Installation checks do not establish credential-store access, successful login, or generation.
  • Golutra Agent 0.3.9 currently filters native MCP tools; use its Skill/CLI route until the host limitation is fixed.
  • Complete official CLI or website equivalence is not claimed. Review current feature, runtime and account limitations.

Integrity and reviewed prices

Source: 7eaaa0af472fc3597fb847f22e4ddd6f6684e946.

Runtime TGZ SHA-256: 5a32a225d0418f9f9cd739fe07a2aa16eb89989641942794cf9ec1b7199a7271.

The six original distribution files are attached without rebuilding. npm's scoped tarball name may differ; its integrity will be checked against the original runtime TGZ after registry processing completes. Additional platform reports and readiness.json preserve the actual validation boundary.

Reviewed price JSON at the source commit

Price file SHA-256: ff826078d351ddd6fcf993fe17b85e41425dc06d83d8d0bfa06060726acd9e20. Reviewed 2026-11-17; expires 2026-11-24 04:25:59 UTC. Estimates exclude promotions; Studio billing is authoritative. Future price updates require reviewed evidence and a trusted checksum, not an automatic expiry extension.