Kuk1/gaia-agent-evaluation

docs: sync technical-architecture with latest tools, multimodal strategy, and dependencies

feat(multimodal): add Whisper-based audio_transcriber and 4o-mini image_question_answer; wire tools; keep file-first strategy

feat(agent): add youtube_transcript_reader tool; improve Excel/XLSB handling in file_reader; expose tool; docs link fixes already pushed; deps: add pyxlsb

docs: remove ADR-0001 on HF Spaces authentication strategy; decision to use Classic Token documented previously

docs: fix broken/uppercase doc links to existing lowercase files

docs(adr): record decision to use Classic (write) token for HF Spaces pushes; rationale & steps

Feat(agent): improve file access robustness (HEAD 405 -> GET fallback, Range), add table_reader, reverse-text detection; UX: numbered logs; Docs: lowercase headings, roadmap clarified; Deps: yt-dlp, youtube-transcript-api, opencv-python-headless

Feat(docs): Implement ADR system for architectural decisions

Docs: Refactor documentation to professional standard

Move the README.md file into the main folder

Refactor: Professionalize project structure

Emergency fixes for GAIA score recovery

Critical fixes for GAIA evaluation

feat: Revise agent configurations and enhance answer extraction methods to improve response management and accuracy

feat: Integrate professional answer extraction system to enhance answer accuracy and confidence reporting in GAIAAgent

feat: Update agent and tool configurations to enhance response capabilities and improve tool call management

feat: Introduce comprehensive agent and tool configurations for enhanced behavior and API management

feat: Enhance GAIAAgent with improved tool call management and answer formatting based on question type

feat: Revise image_analyzer function to improve user prompts and file detail presentation for enhanced image analysis experience

feat: Enhance file_reader to support image and audio file handling, and add image_analyzer and audio_transcriber functions for content analysis

refactor: Replace OpenAI client with GAIAAgent for improved modular architecture and maintain backward compatibility

feat: Introduce cost-efficient model and implement token management for context handling

feat: Upgrade to GPT-4 model and enhance answer extraction with improved cleaning and validation methods

feat: Add question classification, clear logging, and answer extraction from conversation history

feat: Add anti-loop mechanism and improve system prompt for better reasoning

feat: Add python_interpreter tool for executing Python code and enhance agent instructions

refactor: Simplify to native OpenAI tools API for better stability

fix: Correctly serialize Pydantic models for OpenAI tools API

Update

fix: Implement discriminated unions and correct tool call format

feat: Refactor Code Structure

feat: Refactor to structured output with Pydantic and Instructor

feat: Add webpage_reader tool and improve agent logic

feat: Implement ReAct agent with web search tool

feat: Integrate GPT-4o-mini, add .gitignore, and comment out old code

Merge remote changes and keep local README

initial commit

Update app.py

Update app.py

Update app.py

Update app.py

Update requirements.txt

Update app.py

Update app.py

Update app.py

Update app.py

Update app.py

Update app.py

Update app.py

Update app.py