Releases: chrischabot/interviewer
Releases · chrischabot/interviewer
Interviewer v0.4
What's New in v0.4
Writer Quality Improvements
- Completely rewritten writer prompts - Produces sophisticated, flowing prose for educated readers instead of staccato LinkedIn-style sentences
- Sentence complexity: 8/10 with compound/complex sentences and subordinate clauses
- Paragraphs now 4-6 sentences developing ideas fully
- Removed Derek Sivers style anchor (caused short punchy sentences)
- No more bullet points in essays - lists woven into prose
UI Improvements
- Delete buttons on Analysis and Plan screens with confirmation dialogs
- New conversation list icons: sparkles (analysis), chat bubble (continue), chevron (new)
- Scrollable recent conversations - shows all sessions with auto-hiding scrollbar
- All icons now consistent grey color
Bug Fixes
- Fixed AudioDeviceManager initialization race condition (spurious "device not found" messages)
- Removed negative LinkedIn instruction that LLMs fixated on
Fully signed and notarized - macOS will allow it to run without warnings.
Interviewer v0.3
What's New in v0.3
Bug Fixes
- Fixed mic status indicator - The indicator now properly transitions from yellow (muted) to green (ready/listening) after the AI finishes speaking
- Changed from computed property to stored property with timer
- Properly triggers SwiftUI updates when bleed guard period expires
UI Improvements
- Streaming writer output - Essay generation now streams in real-time with a typewriter effect
- Regenerate button - Regenerate essays with a single click
- Analysis button in recent conversations now navigates to the analysis screen
- Extracted view components for better code organization
- Updated button accessibility labels
Documentation
- Updated CLAUDE.md with Swift & SwiftUI coding guidelines
Fully signed and notarized - this release is properly code-signed with Developer ID and notarized by Apple. macOS will allow it to run without any warnings.
Interviewer v0.2
What's New in v0.2
Streaming Writer Output
- Real-time essay generation: Watch the essay being written word-by-word instead of waiting 30-60 seconds
- Auto-scrolling to follow new content as it streams in
- "Writing..." indicator with spinner during generation
New UI Controls
- Regenerate button: Test new writing prompts on existing sessions without restarting
- Cancels any ongoing generation
- Deletes cached draft and generates fresh
- Home button: Quick navigation back from the draft view at any time
Writer Agent Improvements
- Refined writing constraints for more natural, human prose
- Better voice preservation from interview transcripts
- Improved paragraph structure and flow
Technical Improvements
- Streaming support added to LLMClient protocol (OpenAI and Anthropic adapters)
- Task cancellation handling for clean regeneration workflow
- Unified logging and hardened LLM flows
- Fixed research result schema issues
Bug Fixes
- Fixed API key validation caching to avoid double keychain prompts
- Various stability improvements
Interviewer v0.1
Interviewer v0.1 - Initial Release
"You don't know what you know, until you are asked."
A voice-first companion that conducts podcast-style interviews and transforms your spoken expertise into polished essays.
Features
- 🎙️ Voice-first interview experience with OpenAI Realtime API
- 🤖 Multi-agent system (Planner, Note-Taker, Researcher, Orchestrator, Analyst, Writer)
- 📝 Automatic essay generation in your voice
- 🔄 Follow-up sessions to deepen existing conversations
- 💾 Local data storage with SwiftData
Requirements
- macOS 15.0 (Sequoia) or later
- OpenAI API key with Realtime API access
Installation
- Download
Interviewer.dmg - Open the DMG and drag Interviewer to Applications
- Launch and enter your OpenAI API key in Settings
- Choose a topic and start speaking
Built with Swift, SwiftUI, and a profound respect for the difficulty of knowing what you know.