- Video pipeline: video file OCR via VLM plus audio track ASR, integrated
into job_handlers with progress emit per phase. New media_utils.py for
audio extraction from video files.
- Document export: richExport.js replaces inline docx builder; DOCX and PDF
export buttons are now enabled in MilkdownEditor. File size limit raised to
100 MB.
- Input block: new InputBlockCrepe.vue component with inputBlockPlugin.ts
and inputBlock.js for custom user-input nodes in the editor.
- Risk config: added Vite dev server ports (5173) to CORS allowlist and
increased OCR max input from 10 MB to 100 MB.
- TTS/ASR refactor: simplified tts_asr.py model loading and warmup logic.
- Test coverage: updated tests for llm, main endpoints, pro completions and
web search modules.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
- Introduced a new web search block plugin to handle web search queries and results.
- Updated copilot, doc block, and pro block plugins to include web search context in AI completions.
- Implemented utility functions for parsing and building web search markdown.
- Enhanced API to support web search requests and responses.
- Added configuration for web search URL and timeout settings.
- Updated size limit checks to account for web search content.
- Introduced `requirements.docker.txt` for Docker-specific dependencies.
- Updated `requirements.txt` to include `psycopg[binary]` and `python-multipart`.
- Enhanced test suite in `test_main_endpoints.py` to cover document CRUD operations.
- Modified `docker-compose.yml` to include PostgreSQL and frontend services.
- Added Nginx configuration for reverse proxying API requests.
- Refactored file handling in Vue components to support new document storage backend.
- Created new utility functions in `docsApi.js` for document management.
- Updated configuration to support new API endpoints for document operations.
- Adjusted Vite configuration to proxy API requests to the local backend.
- Added support for PRO mode in LLM with specific instruction handling and context awareness.
- Updated prompt building functions to include prefill options for better context management.
- Introduced new inline examples for PRO mode in JSON format.
- Enhanced system prompts to reflect PRO mode capabilities and rules.
- Modified API endpoints to accommodate new parameters and ensure backward compatibility.
- Improved test cases to validate new functionality and ensure comprehensive coverage.
The workspace now carries the pro editing flow, streaming completion path, and lighter Office preview state as one checkpoint so the remote has the current runnable project shape.
Constraint: Preserve the current workspace as a single reviewable project commit while excluding local agent state and verification artifacts. Removed stale Univer runtime dependencies from the lockfile so installs match package.json.
Rejected: Commit runtime screenshots, .omx state, and coverage files | they are local artifacts rather than source state.
Confidence: medium
Scope-risk: broad
Directive: Keep package.json and package-lock.json synchronized when changing frontend dependencies.
Tested: npm run build; C:\Users\ydy\.conda\envs\llmwebsite\python.exe -m pytest backend/tests/test_main_endpoints.py backend/tests/test_main_cancel.py backend/tests/test_llm.py backend/tests/test_llm_extended.py -v -o addopts= (44 passed).
Not-tested: Full pytest with repository coverage addopts currently reports 0% coverage because pytest-cov watches backend.* module names while tests import top-level backend modules.
Co-authored-by: OmX <omx@oh-my-codex.dev>
- Normalize line endings in Markdown export for DOCX files.
- Improve selection serialization to Markdown with better handling of empty documents.
- Add a new `updateFile` function to the file system for updating file properties.
- Introduce video transcoding capabilities using FFmpeg, supporting various video formats.
- Update AGENTS.md for clearer plugin structure and responsibilities.
- Add scoped styles for TreeNodeItem component to improve UI consistency.
- Implement cross-origin isolation headers in Vite configuration for enhanced security.
- Remove obsolete test_cross.py file.
- Add AGENTS.md knowledge base with project documentation
- Move UserPreferences model to separate models.py file
- Extract API_KEY to environment variable for security
- Enhance Univer Editor with PPTX support and improved UI
- Improve file system handling with binary file detection
- Add HF_ENDPOINT mirror for better China connectivity
- Clean up unused imports and code structure
Introduce a comprehensive TTS/ASR module that:
- Adds /v1/tts-asr/config, /status, /warmup, /tts, /asr endpoints with detailed JSON responses
- Implements Apple‑Silicon detection, device selection (MPS/CUDA/CPU), and memory limiting logic
- Supports selectable model size, quantization, and offline mode via environment variables
- Adds robust audio validation and multi‑path resampling fallback
- Provides new README sections for API usage, device detection, and performance benchmarking
- Includes a full testing suite: unit tests, integration tests, macOS simulation and performance reports
- Updates backend dependencies and CI scripts
- Adds new front‑end views and components for Univer editor integration
All changes are backward compatible; new features are exposed through environment variables and new API routes.
Adds a DocBlock component that renders embedded documents, new export buttons for DOCX
and PDF, and updates the file‑upload picker to accept *.txt, *.docx, *.pptx, and *.pdf.
Introduces a DOCX→PDF conversion bridge in the backend and new /tts and /asr
endpoints that expose TTS and speech‑recognition functionality. The README is
rewritten to describe the new features and clean up legacy documentation. All
changes are backward‑compatible and do not introduce breaking API changes.
Add support for cancelling in-progress LLM completion requests via new /v1/completions/cancel endpoint with task tracking. Implement mermaid diagram rendering in the Milkdown editor with a new mermaidPlugin. Update copilotPlugin to properly abort requests with descriptive reasons. Refactor settings panel to handle system theme changes reactively. Add camera capture support for image uploads.
Separate prompt generation into system and user prompts for better LLM instruction following. Backend now builds a detailed system prompt with constraints for math formatting, code block handling, boundary newlines, and OCR safety, while user prompt contains context and completion state flags. Added corresponding tests for both modules.
Add API key security using fastapi.security.APIKeyHeader to protect /v1/completions and /v1/ocr endpoints. Updated frontend to include X-API-Key header in API requests. Also changed default API base URL from http://149.104.29.239:8001 to https://api.learnteach.tech:8002.
Add optional thinking parameter to the call_ollama function and pass it from the request. Also enhance timezone handling in prompt generation to support configurable timezone preferences.
- Add privacy mode to hide IP and user preferences from AI requests
- Add model thinking levels (low/medium/high) for context analysis depth
- Add PWA support with service worker, manifest, and app icons
- Add SettingsPanel for user preferences (theme, background, language)
- Add i18n translations for en/zh/ja/ko/de/fr
- Add Pinia store for centralized settings management
- Update backend to support user preferences and thinking levels
- Update config to use absolute API URLs
- Add GeoLite2-City.mmdb database for IP lookup
- Create geoip.py module for IP location services
- Extract client IP from requests and log location info
- Pass location context to LLM prompts for enhanced responses
- Configure Vite dev server to proxy /v1 requests to backend
- Change backend port from 8000 to 8001
- Simplify API URLs to relative paths instead of absolute localhost URLs
- Increase debounce delay from 500ms to 1000ms for better stability
- Update README documentation to reflect all changes
- Implemented a new composable `useTheme` for managing theme state.
- Added functions to read and write theme preference to local storage.
- Applied theme styles to the DOM based on user preference.
- Introduced a toggle function to switch between light and dark themes.
refactor: enhance copilot plugin functionality
- Improved request handling with sequence and document versioning.
- Refactored ghost text handling to improve clarity and efficiency.
- Updated markdown insertion logic to handle parsed content more robustly.
- Enhanced error handling and logging for better debugging.
style: update global styles for light and dark themes
- Defined CSS variables for light and dark themes to streamline styling.
- Improved overall styling consistency and responsiveness.
- Added transitions for smoother theme changes and interactions.
Add image button with dropdown menu for uploading local images or inserting from URL.
Integrate VLM-based OCR to extract text context from images and include in AI suggestions.
Implement document size limits to disable AI when exceeding threshold.
Refactor copilot plugin with per-view runtime state and OCR context injection.
Add OCR cache utility for managing image metadata.
Add code splitting configuration for optimized bundle size.
- Rewrite prompt builder with comprehensive rules for seamless text completion
- Implement Markdown parsing for ghost text with proper mark handling
- Update LLM parameters (temperature 0.7, repeat_penalty, think mode)
- Add CSS styles for formatted ghost text elements
- Add planning documentation for Copilot prompt system analysis
- Implement Markdown parsing for ghost text using Milkdown parser
- Add support for both block and inline content in ghost text
- Refactor prompt system with comprehensive rules and examples
- Adjust LLM parameters: increase temperature to 0.7, add repeat_penalty
- Add CSS styles for formatted ghost text (bold, italic, code, links)
- Add documentation for Copilot prompt system and ghost text rendering
- Replace overlay-based GhostTextOverlay.vue with ProseMirror Mark system
- Add AI toggle button with enable/disable functionality
- Implement new copilotPlugin.ts using copilotGhostMark for inline suggestions
- Fix cursor position offset in prompt.py by moving first suffix char to prefix
- Improve API error handling with abort signal support and debug logging
- Update model configuration from gpt-oss:120b to gpt-oss:20b
- Add button tooltips and improve editor styling
- Remove deprecated inlineSuggestionPlugin.ts
- Update README with new architecture diagram and feature documentation
This commit refactors the LLM integration to use Ollama's native Python client instead of OpenAI-compatible API, while fixing critical template syntax errors and improving project structure.
Key changes:
- Replace openai package with ollama package in backend requirements
- Rewrite llm.py to use ollama.AsyncClient for direct Ollama API calls
- Update main.py to use non-streaming Ollama responses with thinking extraction
- Fix template syntax error in MilkdownEditor.vue (GhostTextOverlay component tags)
- Fix string截取错误 by using slice() instead of substring()
- Add src/utils/api.js and src/utils/config.js for shared configuration
- Add CORS middleware to FastAPI backend
- Update prompt.py with clearer instructions for continuation generation
- Add comprehensive README.md documentation
BREAKING CHANGE: Environment variables OLLAMA_BASE_URL changed to OLLAMA_HOST (remove /v1/ suffix)
“ydy0615”
ydy0615