v1.0
Current versionmacOS / Windows
Real-time support for interview prep and live practice
This update focuses on interview materials, planning integration, knowledge retrieval, live Q&A, screenshot solving, interview review, and stealth assistance as one more complete workflow. As a v1.0 major release, upgrading is recommended, with continued stability improvements across macOS and Windows.
Complete update list (22 items)
- NewAdded an Interview Materials workspace for candidate, role, and supporting materials, with manual entry, upload, paste, AI extraction, review editing, and version recovery.
- NewAdded full-screen and region screenshot Q&A, answer-direction switching for the same question, and a way to return to Voice Ask for spoken follow-ups after a coding answer.
- NewAdded AI eye-contact correction guidance and Dual-Device Interview Support (Beta), with virtual-camera, opacity, and second-device rehearsal steps for video interviews.
- NewAdded a faster Agent API and MCP setup: create an API key and give it directly to Codex App, Claude App, or another supported Agent to configure the connection.
- NewExpanded live voice sessions and low-latency transcription with automatic or manual submission, continuous follow-ups, and AI / Transcript view switching.
- NewAdded Live English translation: speech from the other side is recognized with English priority and shown with Chinese below the original; Flagship only.
- NewUpgraded the AI Resume Center with online editing, templates, JD tailoring, polish and expansion, and application-version management.
- ImprovedImproved Interview Planning with session material binding and cross-platform synchronization, so the same candidate, role, and supporting materials can be used in the Portal and desktop app; supporting-material usage and plan limits are clearer, and plans remain usable after their scheduled time.
- ImprovedImproved Interview Review reports with evidence-based summaries, scores, strengths, and improvement guidance generated from real Q&A, including support for longer interview records.
- ImprovedImproved resume import and AI editing with support for additional file types; rich-text polishing, content review, and output now preserve paragraphs, lists, and original formatting more reliably.
- ImprovedKnowledge retrieval now combines keyword and semantic matching, with explicit control over whether materials are used only in the Cheatsheet or also for AI answers; it falls back automatically when semantic retrieval is unavailable without blocking ordinary answers.
- ImprovedKnowledge Cheatsheet and answer references now use the materials actually selected for the interview, refresh when candidate, role, or supporting materials change, show the material name in the reader, and support mouse, trackpad, and keyboard scrolling for long content.
- ImprovedExpanded multilingual setup guidance with 40 interview-language and regional/combined options; interview, answer, and speech-recognition languages can be adjusted independently, with a 9-language UI.
- ImprovedAdded clearer stealth-window, click-through, hidden-keyboard-event, and platform-rehearsal guidance so users can verify shared-screen and local-reading behavior before a live interview.
- ImprovedImproved the Resume Center to Interview Materials handoff: turn an edited resume into candidate material, then load it through an Interview Plan in Desktop.
- ImprovedImproved coordination across answer preferences, automatic answers, manual Ask, and voice follow-ups; long interviews now compress earlier context and resume more safely to reduce missing information and repetition across follow-up questions.
- ImprovedReworked Free, Pro, and Flagship model entitlements and updated eight mainstream models across Gemini, GPT, and Claude, keeping catalog visibility, execution routes, and account permissions aligned; the catalog now covers fast, balanced, advanced, and deep-reasoning choices, while Deep screenshot mode can use a higher reasoning level for complex interface, document, and coding questions.
- ImprovedImproved material processing across the Portal and desktop app with continuous progress, clear failure reasons, in-place retry, structured previews, transient network retry, and duplicate-action prevention.
- ImprovedUpgraded the underlying audio, window, shortcut, and screenshot-protection infrastructure with verified native components and matching runtime assets on macOS and Windows, reducing long-session stalls, permission surprises, and platform compatibility issues.
- FixedFixed multiple file upload, pasted extraction, and processing-progress issues to improve material creation success on macOS and Windows.
- FixedFixed multiple form editing, UI responsiveness, shortcut reliability, and long-session issues that could cause validation errors, missing content, or delayed refreshes.
- FixedFixed repeated material indexing and retrieval failures by reusing completed indexes, avoiding duplicate processing and Knowledge Point charges, and keeping ordinary answers available when retrieval fails.