Spaces:
Sleeping
Sleeping
Commit History
-
UPGRADE: Use langchain-ollama package with OllamaLLM (removes Ollama deprecation warning)
cb48d61 -
FIX: Use langchain_classic.memory for ConversationBufferWindowMemory (langchain 0.3.27 has memory under langchain_classic, not langchain)
ac70217 -
Fix langchain imports: use langchain.memory for ConversationBufferWindowMemory, use langchain_core.tools for BaseTool (langchain 0.3.27 compatibility)
a68baeb -
Fix: Resolve all import errors and dependency issues
- Fix langchain imports: memory and tools now from langchain_community - Add missing 'requests' package to requirements - Correct google-generativeai package name in pyproject.toml - Sync requirements.txt and pyproject.toml dependencies - Remove unused packages (gradio, streamlit, trafilatura) from pyproject.toml1691b3a
-
🚀 CRITICAL FIX: Infinite processing timeout issue
Priyanshi Saxena- Added 30s timeout to Ollama/Gemini tool analysis (was hanging indefinitely) - Added fallback tool selection when analysis times out - Simplified prompts for faster LLM processing - Optimized AI safety prompts to reduce context length - Fixed memory manager integration - Now system will respond within 60-90s maximum instead of infinite hang FIXES: - Tool analysis infinite hang ⚡ - Context cancellation errors ⚡ - Memory integration working ⚡ - Faster response times ⚡d44409c -
timeout changed
Priyanshi Saxena68d7321 -
fix: memory manager
Priyanshi Saxena65703d9 -
fix: Add comprehensive response cleaning validation
Priyanshi Saxena- Added version tracking for deployment debugging - Enhanced logging to track content extraction process - Added critical safety check for LangChain metadata in final responses - Improved debugging for Hugging Face Spaces deployment issues - Added explicit validation to prevent raw metadata from being returned923b4b3 -
fix: Clean LLM responses and fix variable scope errors
Priyanshi Saxena- Fixed Gemini response cleaning to extract content properly from LangChain objects - Fixed Ollama response cleaning to handle string conversion - Fixed variable scope error in timeout handler (btc_data -> summary_data) - Fixed DeFi TVL sorting error by filtering out None values - Enhanced error handling in fallback response mechanism - Improved response content extraction for both LLMsa4b05e9 -
feat: Replace CoinGecko with CryptoCompare/Etherscan APIs
Priyanshi Saxena- Updated tool analysis prompts to prioritize available APIs (CryptoCompare, Etherscan) - Removed CoinGecko from suggested tools since no API key available - Modified ChartDataTool to use CryptoCompare instead of CoinGecko for price data - Added conditional CoinGecko tool initialization (only if API key present) - Updated market overview to use CryptoCompare data source - Enhanced tool selection logic to exclude unavailable APIs - Added comprehensive testing for tool selection functionalityb091b2c -
fix: Improve session cleanup and management
Priyanshi Saxena- Enhanced ChartDataTool cleanup method with proper session management - Added try/finally blocks to ensure CoinGecko tool cleanup - Improved garbage collection for better memory management - Fixed aiohttp session closure issues that were causing warnings18efaca -
fix: Add missing validate_gemini_response method and cleanup method
Priyanshi Saxena- Added validate_gemini_response method to AISafetyGuard - Added cleanup method to ChartDataTool to prevent AttributeError - Improved Gemini tool parsing logic to handle all suggested tools - Updated Gemini model to use 'gemini-2.0-flash-lite'c785b3f -
new
Priyanshi Saxena2fe0e75 -
feat: Add Gemini/Ollama toggle functionality with API error fixes
Priyanshi Saxena✨ New Features: - Add interactive toggle switch in UI to choose between Ollama (Local) and Gemini (Cloud) - Dynamic LLM selection with preference persistence in localStorage - Support both Ollama and Gemini in research agent with automatic fallbacks - Real-time status updates showing which AI model is being used �� Backend Changes: - Update research agent to support both LLMs with use_gemini parameter - Add _research_with_gemini_tools method for Gemini-powered research - Update query request model to include use_gemini field - Automatic LLM initialization and availability detection 🎨 Frontend Changes: - Beautiful toggle switch with animated slider and labels - Persistent user preference storage - Dynamic status messages based on selected LLM - Responsive design for header controls 🛠️ Bug Fixes: - Fix CoinGecko API 401 errors by adding proper fallback to mock data - Improved error handling for chart_data_tool API failures - Better session cleanup to prevent connection warnings 🔒 …6bf47a1 -
fix: Resolve config attribute error and chart_data_tool parameter handling
Priyanshi Saxena- Fix 'Web3ResearchAgent' object has no attribute 'config' error - Change self.config.OLLAMA_MODEL to config.OLLAMA_MODEL in metadata - Add proper parameter handling for chart_data_provider tool - Extract chart_type and symbol from query for structured tool calls - Prevent chart_data_tool from receiving raw query strings Resolves: - AttributeError: 'Web3ResearchAgent' object has no attribute 'config' - chart_data_provider error: 'Unknown chart type: Analyze Bitcoin...' - Proper tool parameter extraction and execution0d94513 -
project complete
Priyanshi Saxenac02fe07
-
final commit with chart generation
Priyanshi Saxena4498d8e -
feat: more features
Priyanshi Saxenac52b367 -
Add comprehensive logging and debug endpoints for tool calling
Priyanshi Saxena2c26464 -
fix: etherscan_tool error fix
Priyanshi Saxenaea46ec8 -
feat: testing done
Priyanshi Saxena9b006e9 -
feat: added api tools
Priyanshi Saxenaf104fee -
feat: Initial commit with project structure and initial files
“Transcendental-Programmer”20eee66