Release v0.229.098 · microsoft/simplechat

Out of cycle release today for Simple Chat

Scoping issue when selecting All in chat - Search included personal and public documents but was missing groups
- #489
Video indexer logic improvements
- #527
- The API key did not work with paid service, only works with trial
- Removed API key authentication
- Updated config guidance to show how to setup managed identity permissions of the app service to the video indexer service

Added support for .xml, .yaml/.yml, .doc, .docm, and .log file types

Implemented comprehensive image upload support with AI-powered analysis:

Features:

Base64 Conversion & Inline Display: Uploaded images are converted to base64 and displayed inline in the chat (like AI-generated images) instead of as file links
Automatic Chunking: Large images (>1.5MB) are automatically split across multiple Cosmos DB documents to avoid the 2MB document limit, then seamlessly reassembled on retrieval
Dual Text Extraction:
- Document Intelligence OCR: Extracts all visible text from the image
- GPT-4o Vision Analysis: Provides AI-generated description, object detection, contextual analysis, and text interpretation
Info Button: User-uploaded images display an info button that reveals extracted text and vision analysis in a formatted, scrollable drawer
Token-Efficient Chat History: Image context (OCR + Vision analysis) is included in the chat history as system messages so the AI can answer questions about uploaded images, but base64 image data is explicitly excluded to prevent token waste
- OCR + Vision analysis: ~625 tokens per image ✅
- Full base64 data: ~350K tokens per 1MB image ❌ (prevented)
Settings Control: Multi-modal vision can be enabled/disabled in admin settings with model selection (GPT-4o, GPT-4o-mini, o-series, GPT-5, etc.)

Technical Implementation:

Images stored with role: 'image' and metadata.is_user_upload: true flag
Stores extracted_text (OCR), vision_analysis (AI insights), and filename metadata
Backend automatically includes image context in conversation history for AI reasoning
Runtime safety checks prevent base64 data leakage into chat history
Debug logging tracks image context addition and character counts

User Experience:

Uses RecursiveCharacterTextSplitter with XML-aware separators
Structure-preserving chunking:
- Separators prioritized: \n\n → \n → > (end of XML tags) → space → character
- Splits at logical boundaries to maintain tag integrity
Chunked by 4000 characters
Goal: Preserve XML structure by splitting at tag boundaries rather than mid-element, ensuring chunks are more semantically meaningful for LLM processing
See process_xml

Uses RecursiveCharacterTextSplitter with YAML-aware separators
Structure-preserving chunking:
- Separators prioritized: \n\n → \n → - (YAML list items) → space → character
- Splits at logical boundaries to maintain YAML structure
Chunked by 4000 characters
Goal: Preserve YAML hierarchy and list structures by splitting at section boundaries and list items rather than mid-key or mid-value
See process_yaml

Processed using line-based chunking to maintain log record integrity
Never splits mid-line to preserve complete log entries
Line-Level Chunking:
1. Split file by lines using splitlines(keepends=True) to preserve line endings
2. Accumulate complete lines until reaching target word count ≈1000 words
3. When adding next line would exceed target AND chunk already has content:
  - Finalize current chunk
  - Start new chunk with current line
4. If single line exceeds target, it gets its own chunk to prevent infinite loops
5. Emit chunks with complete log records
Goal: Provide substantial log context (1000 words) while ensuring no log entry is split across chunks
See process_log