v1.1.1

Fixed

• Make the materials_advanced.twoThetaRange tool schema compatible with Vertex AI and Gemini by using homogeneous array items with an exact length of two.
• Add a regression test for provider-compatible array schemas.
• Synchronize package and lockfile versions.

v1.1.0

Changes in v1.1.0

• Publish compiled JavaScript from dist and build it automatically during npm pack.
• Expose all Ai4Scholar operations as native Pi tools and remove the MCP adapter and direct MCP SDK dependency.
• Read Ai4Scholar credentials from the unified Pi Scholar config or environment variables only.
• Use separate default timeouts for crawler-backed and structured Ai4Scholar requests, with cancellation and progress reporting.
• Correct Google Scholar, recommendation, dataset, PubMed, and Google Patents request mappings.
• Replace the previous streaming citation inserter with verifiable Semantic Scholar citation candidates.
• Download Ai4Scholar images to the output directory and return supported images inline.
• Rewrite Chinese and English user guides, add compact bilingual Materials Project capability pages, and validate bundled skills against the Agent Skills frontmatter rules.
• Remove obsolete contract notes, draft specifications, generated capability JSON, and local release artifacts.

Pi Scholar 1.0.2

Documentation humanization release:

• Humanize documentation tone across Chinese and English documentation.
• Replace overly defensive audit/contract phrasing with clear, developer-oriented usage guides.
• Streamline configuration and research provider notes around practical developer workflows while preserving all security and error-handling constraints.

Pi Scholar 1.0.1

Documentation cleanup release:

• Keep release and configuration-transition notes in CHANGELOG.md instead of separate upgrade and implementation-report files.
• Remove test-machine and transport-failure wording from the research documentation.
• Keep README and configuration references focused on current behavior.

Pi Scholar 1.0.0

Production release with schemaVersion 3 unified direct-key and environment fallback configuration, preserved Ai4Scholar support, and the complete MP01-MP17 Materials Project workflow. Includes route-specific REST tools, a fixed optional mp-api/pymatgen bridge, phase diagrams, simulated XRD, reproducible exports, generated capability matrices, and rewritten Chinese/English documentation. PubMed live access remains dependent on local NCBI connectivity; sampled complete band/DOS/phonon objects were unavailable from the official helper and are reported as upstream SOURCE_UNAVAILABLE instead of being fabricated.

Pi Scholar v0.5.0

Production release of safe synchronization and multi-source research. Adds Semantic Scholar, OpenAlex, PubMed/PMC, arXiv, Crossref, Unpaywall, easyScholar journal ranks, Materials Project workflows, and the contract-gated CAS Common Chemistry boundary. Preserves Ai4Scholar as an explicit independent service. Includes transactional publication recovery, user-content protection, bounded network handling, schemaVersion 2 migration, updated skills, bilingual documentation, and 149 passing tests.

Pi Scholar v0.4.3

Changed

• Installation now uses npm latest for first-time setup.
• Updates now use npm latest so the installed extension follows the newest published release.
• Fixed-version installation remains available when a release needs to be pinned.

Pi Scholar v0.4.2

Changed

• Unified configuration examples now use direct apiKey entries for image providers.
• Image-provider configuration supports direct API keys with providerOptions..apiKey; environment-variable names remain available through apiKeyEnv.
• Atlas GPT Image 2 documentation and requests use explicit pixel dimensions, with 1024x1024 as the default.
• Updated the Chinese and English README, configuration reference, provider reference, and regression tests.

Pi Scholar v0.4.1

Fixed

• Atlas GPT Image 2 requests now include the upstream billing-required size, defaulting to 1024x1024.
• Explicit pixel dimensions are forwarded to Atlas while 1K, 2K, and 4K aliases are rejected when unsupported.
• Updated the Chinese and English provider compatibility documentation and regression coverage.

Pi Scholar v0.4.0

Added

• Native scientific image generation and editing for Gemini API, Vertex AI, OpenAI, xAI, fal.ai, Qwen/DashScope, Atlas, and custom OpenAI-compatible services.
• Unified text-to-image, image-to-image, multi-reference, mask, aspect-ratio, resolution, count, quality, transparency, format, and compression controls with provider-specific validation.
• Read-only model discovery, non-generating connection checks, configured default models, and automatic selection of the newest compatible candidate.
• Vertex service-account JSON, Application Default Credentials, project, and regional endpoint support.
• Safe reference input handling, asynchronous job polling, origin-bound downloads, bounded response bodies, atomic artifact publishing, and secret redaction.
• Scientific figure workflow guidance covering provider selection, prompt structure, cost control, validation, and research-integrity requirements.

Changed

• Consolidated image provider settings into the existing pi-scholar.config.json discovery and validation flow.
• Extended /pi-scholar status to report configured image providers without exposing credentials.
• Updated Chinese and English configuration documentation with provider compatibility and current official references.

Notes

• Balance lookup remains disabled where no stable official API is verified. Use the corresponding provider billing console.
• Connection checks use read-only catalogs and do not prove image inference entitlement or remaining quota.