Spaces:
Running
Running
Commit History
-
Make context meter optional (hidden by default) and soften README model refs Running
NeonClary +1: CursorAdd a Settings toggle for the context-use bar and token/summary status, defaulting to hidden and persisting in sessionStorage. Replace dated BrainForge/Security model pins in the README with “most recent model” guidance. Co-authored-by: Cursor <cursoragent@cursor.com>45a489a
-
Merge AskJerry UI refresh + jump-button fixes into dev-cursor
NeonClary +1: CursorKeeps dev-cursor's copy-response button, streaming hardening, and backend vLLM/model fixes; adopts the warm cream/Chumash tribal-tech theme, sticky composer dock, footer, and jump-to-question fixes. Resolved theme conflicts in favor of the new token system and dropped a hardcoded API key from .env.example. Co-authored-by: Cursor <cursoragent@cursor.com>53f0a6c -
Refresh Ask Jerry UI and fix jump-to-question button
NeonClary +1: Cursor- Apply warm cream / Chumash tribal-tech theme via centralized CSS tokens (light + dark), surface layering, single primary action, and a subtle sunburst signature strip under the header. - Center the conversation in a reading column, add a sticky composer dock and a persistent copyright footer. - Fix footer background gap: [data-tip] was overriding the jump button's position:fixed, dropping it into normal flow; pin it with higher specificity. - Position the jump button against the reading column's right edge and stop it from stealing focus into the composer on click. - Right-size the Neon.ai logo; backend error-message and model-id cleanup. Co-authored-by: Cursor <cursoragent@cursor.com>175465b
-
Fix Ask Jerry vLLM auth and model id for current 4090-x1-3 deployment.
NeonClary +1: CursorUse shared.env VLLM_API_KEY and BrainForge/Security@2026.05.13 instead of the retired klatchat password and old model id. Co-authored-by: Cursor <cursoragent@cursor.com>82d23ab
-
Add copy-response button and harden Ask Jerry chat streaming.
NeonClary +1: CursorLoad shared.env for local secrets, point vLLM example at Security on 4090-x1-3, improve SSE error handling, and add dev start script. Co-authored-by: Cursor <cursoragent@cursor.com>658b082
-
fix: render options menu via portal to body (escapes overflow clip)
NeonClaryMade-with: Cursorb12ad77 -
fix: options menu fixed position + measured top so it escapes overflow clip
NeonClaryMade-with: Cursorf52a564 -
style: Neon.ai wordmark solid #0e7490, no gradient
NeonClaryMade-with: Cursord5b777c -
style: slightly darken Neon.ai cyan gradient
NeonClaryMade-with: Cursorda7dbe4 -
style: Neon.ai wordmark cyan gradient to match title brightness
NeonClaryMade-with: Cursor9472d30 -
style: Neon.ai wordmark deep dark turquoise aligned to logo cyan
NeonClaryMade-with: Cursor714eca6 -
style: match Neon.ai brand size to main header title (20px)
NeonClaryMade-with: Cursor4902551 -
style: match header Neon.ai and empty-state text to chat (14px)
NeonClaryMade-with: Cursor94f1cae -
fix: options panel height fits content up to 70vh (remove 520px cap)
NeonClaryMade-with: Cursor60e5f9f -
copy: drop (text-to-speech) from Always speak hint
NeonClaryMade-with: Cursor9459c88 -
style: Options menu lavender background, purple section borders
NeonClaryMade-with: Cursor5551a3f -
ui: split Always speak vs Playback speed in Settings; trim hints
NeonClaryMade-with: Cursor73302b6
-
fix: position header bar tooltips below and left-aligned into view
NeonClaryMade-with: Cursorf4888c6 -
ui: group Settings into boxed sections, move persona to bottom
NeonClaryMade-with: Cursorebb59c0 -
fix: show TTS speed control button on all messages, not just during playback
NeonClaryMade-with: Cursorcaafa14 -
feat: TTS playback speed control (1x / 1.5x / 2x) in message controls and Settings
NeonClaryMade-with: Cursor7e082a0 -
feat: add skip-back button for TTS playback (replay current or previous sentence)
NeonClaryMade-with: Cursordc20b43 -
Merge FEAT_STT-TTS: TTS/STT voice features with streaming autoplay
NeonClaryMade-with: Cursora2abb02 -
refactor: simplify TTS pipeline with stable sentence boundaries, remove prefetch cache
NeonClaryMade-with: Cursorf714780 -
docs: README note for python -m pip on Windows; master has dev ports
NeonClaryMade-with: Cursore7c7cf6 -
chore: configurable dev ports for parallel FEAT + master worktrees
NeonClary- ASKJERRY_DEV_PORT / ASKJERRY_API_PORT in vite.config.js - npm run dev:alt -> 3007/8007; cross-env for Windows - README: worktree + dual-stack runbook; .env.example CORS note Made-with: Cursorddca9c1 -
chore: configurable dev ports for parallel FEAT + master worktrees
NeonClary- ASKJERRY_DEV_PORT / ASKJERRY_API_PORT in vite.config.js - npm run dev:alt -> 3007/8007; cross-env for Windows - README: worktree + dual-stack runbook; .env.example CORS note Made-with: Cursor6b325ba -
feat: STT/TTS (Coqui/Whisper), always-speak, mic input, TTS prefetch
NeonClary- Backend: /api/tts, /api/transcribe, /api/voice/status|wake; ffmpeg in Docker - UI: speaker + search on assistant rows; Options: Always speak responses - STT: mic button; green composer while MediaRecorder is live - Prefetch first-sentence TTS after user has used speak once (session primed) - NLTK punkt fallback; refresh clears TTS prefetch Made-with: Cursor9110948
-
ui: chat layout — wider gutters, 960px row, 30% center overlap
NeonClary- Halve horizontal padding on header, messages, context bar, composer, jump button - Message row max-width 960px; composer inner 1040px - Symmetric 35% side padding for 30% center overlap (documented in CSS) Made-with: Cursored47eea -
ui: wide-layout chat columns, label alignment, remove AI dot
NeonClary- 30% padding overlap layout for AI/user bubbles - Names at outer edges of bubbles - Remove purple shape dot; composer height fix from prior session Made-with: Cursor383506a -
ui: compact sticky footer, custom tooltips, button hover polish, favicon
NeonClary- Move composer outside scroll container so it stays compact during long responses - Add Neon.ai logo as favicon - Custom larger tooltips (data-tip) on all buttons - Lavender hover on all buttons, reversed primary button colors - Auto-focus textarea after streaming finishes - Search popover: all 3 buttons purple, web search button added Made-with: Cursorab56503 -
feat: markdown rendering, jump-to-latest-question, web search button
NeonClary- Render AI responses as formatted markdown (react-markdown + remark-gfm) - Fix jump button: scroll to most recent question at top of view - Add Search button in search popover (opens Google with query) - Prepend 'Provide citations for' to auto-generated search prompt - Remove leftover debug fetch instrumentation throughout Made-with: Cursor8a55c98 -
feat: share menu, export as Markdown or plain text, chat UI updates
NeonClary- Add share popover (LinkedIn, Facebook, email, copy to clipboard) - Add export popover with Markdown (.md) and plain text (.txt) downloads - Composer styling and related backend adjustments Made-with: Cursor5f0d937 -
Fix model bleeding past assistant turn with stop sequences
NeonClary4330ca1
-
README: HF CLI snippet for creating Docker Space
NeonClary41531de -
Add Docker + Hugging Face Spaces layout (LLMComparisons-style)
NeonClary- Multi-stage Dockerfile: Vite build -> backend/static, uvicorn on :7860 - Serve SPA + /assets from FastAPI; CORS_ORIGINS=* supported - README Space card metadata (sdk: docker) and deploy notes9995f1a -
UI: align header with reference layout; icon-only toolbar with titles
NeonClary- Single-row grid header (brand | title | actions), gradient title - Icon-only header buttons with hover titles (New chat, Refresh, Stop, Options) - Original lucide icons; Stop shown only while streaming - Remove decorative shape dots on user messages; keep purple dots on AI - Restore Vite dev server to localhost:3006 proxying API :80061feab6b
-
docs: README for standalone NeonClary/AskJerry repo
NeonClary011d189 -
Add context window handling, summarize endpoint, and UI polish
NeonClary3bb9213 -
Document sk_klatchat2024 for BrainForge Security (4090 x1-3) vLLM Bearer
NeonClary1ca4754 -
Initial Ask Jerry demo: React UI on 3006, FastAPI proxy to vLLM on 8006
NeonClaryMade-with: Cursor3b28871