convez.ai
PluginFinanceiroIntermediário
adiiityasiingh

Ferramenta de Pesquisa de Notícias com Langchain

adiiityasiingh/Research-Tool-Langchain

Análise de notícias financeiras com RAG multi-LLM e persistência de conhecimento

Ver no GitHub
InicialPluginScore · de 0 a 100

Plugin que integra múltiplos provedores de LLM (OpenAI, Claude, Gemini, Ollama) para análise de artigos de notícias e pesquisa financeira. Permite carregar URLs, processar conteúdo com RAG (Retrieval-Augmented Generation) via ChromaDB e fazer perguntas específicas sobre o material indexado.

Instalação

$npx convez-ai adiiityasiingh/Research-Tool-Langchain

1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.

Métricas

Estrelas
0
Forks
0
Acompanhando
1
Saúde
42%
Último commit
há 1 ano
Versão
v1.0.0
Issues abertas
0
Contribuidores
1

Casos de uso

  • Pesquisa de notícias do mercado financeiro com QA contextual
  • Análise comparativa de múltiplos artigos sobre um mesmo tema
  • Extração de insights de conteúdo web sem dependência de API única

Veredito

Solução robusta e modular para RAG em domínio financeiro, com suporte a múltiplos provedores e persistência local. A arquitetura é bem pensada, mas requer configuração manual de APIs e é mais adequada para usuários técnicos que precisam de análise profunda de notícias.

README

A tradução para o português deste README ainda está na fila.

News Research Tool 📈

This is a powerful, user-friendly news research tool designed for effortless information retrieval and analysis. Users can input article URLs and ask questions to receive relevant insights from news articles and web content.

✨ Features

🔧 Multiple LLM Providers

  • OpenAI GPT - Fast and reliable responses
  • Anthropic Claude - Advanced reasoning capabilities
  • Google Gemini - Powerful multimodal understanding
  • Ollama (Local) - Run models locally for privacy
  • Local HuggingFace - Offline model support

🧠 Flexible Embedding Options

  • HuggingFace Embeddings - Free, high-quality embeddings
  • OpenAI Embeddings - Premium embeddings with fallback support
  • Automatic fallback between providers

📊 Advanced Vector Storage

  • ChromaDB - Modern, efficient vector database
  • Persistent storage across sessions
  • Automatic knowledge base management
  • Document chunking and indexing

🔍 Smart Content Processing

  • WebBaseLoader - Robust web content extraction
  • RecursiveCharacterTextSplitter - Intelligent text chunking
  • Support for multiple URL processing
  • Automatic content validation

🎯 Enhanced User Experience

  • Real-time processing feedback with loading spinners
  • Knowledge base persistence - no need to reprocess URLs
  • Clear error messages and helpful guidance
  • Debug information for troubleshooting
  • Source attribution for answers

🛠️ Developer Friendly

  • Environment-based configuration (.env support)
  • Comprehensive error handling
  • Deployment-ready with compatibility checks
  • Modular architecture for easy customization

🚀 Installation

1. Clone the Repository

git clone https://github.com/adiiityasiingh/Research-Tool-Langchain.git
cd Research-Tool-Langchain

2. Install Dependencies

pip install -r requirements.txt

3. Environment Setup

Create a .env file in the project root with your API keys:

# Required for OpenAI
OPENAI_API_KEY=your_openai_api_key_here

# Optional - for Anthropic Claude
ANTHROPIC_API_KEY=your_anthropic_api_key_here

# Optional - for Google Gemini  
GOOGLE_API_KEY=your_google_api_key_here

🎮 Usage

1. Start the Application

streamlit run main.py

2. Configure Your Setup

  • Select LLM Provider: Choose from OpenAI, Claude, Gemini, Ollama, or Local HuggingFace
  • Choose Embeddings: Pick between HuggingFace (free) or OpenAI embeddings
  • Add URLs: Input up to 3 news article URLs in the sidebar

3. Process Content

  • Click "Process URLs" to load and index the articles
  • Watch real-time progress indicators
  • Verify document count in the success message

4. Ask Questions

  • Type your question in the main interface
  • Click "🔍 Search for Answer"
  • View the answer with source attribution
  • Expand "Retrieved Document Chunks" for debugging

5. Manage Knowledge Base

  • Clear Knowledge Base: Use the trash button to reset
  • Add More URLs: Process additional articles to expand knowledge
  • Persistent Storage: Your knowledge base persists between sessions

🔧 Configuration Options

LLM Providers

ProviderRequirementsBest For
OpenAIAPI KeyGeneral use, fast responses
ClaudeAPI KeyComplex reasoning, analysis
GeminiAPI KeyMultimodal content
OllamaLocal installPrivacy, offline use
HuggingFaceLocal modelsCustom models, offline

Embedding Providers

ProviderCostQualitySpeed
HuggingFaceFreeHighFast
OpenAIPaidVery HighVery Fast

🚀 Deployment

Streamlit Cloud

  1. Push your code to GitHub
  2. Connect your repository to Streamlit Cloud
  3. Set environment variables in the Streamlit Cloud dashboard
  4. Deploy!

Platform Considerations

  • SQLite Version: Requires SQLite ≥ 3.35.0 for ChromaDB
  • Memory: Local models require additional RAM
  • API Limits: Monitor usage for paid providers

Recommended Platforms

  • ✅ Railway - Modern SQLite, easy deployment
  • ✅ Render - Good compatibility, free tier
  • ✅ Heroku - Reliable, paid service
  • ⚠️ Streamlit Cloud - May have SQLite limitations

📁 Project Structure

Research-Tool-Langchain/
├── main.py                 # Main Streamlit application
├── requirements.txt        # Python dependencies
├── README.md              # This file
├── .env                   # Environment variables (create this)
├── .gitignore            # Git ignore rules
├── chroma_db/            # ChromaDB storage (auto-created)
│   └── ...               # Vector database files
└── rockybot.jpg          # App logo

🛠️ Troubleshooting

Common Issues

"SQLite version too old"

  • Deploy to a platform with SQLite ≥ 3.35.0
  • Consider Railway, Render, or Heroku

"No working embeddings provider found"

  • Check your API keys in .env
  • Try switching between embedding providers

"I don't know" responses

  • Check the "Retrieved Document Chunks" section
  • Verify URLs are accessible and contain text
  • Try rephrasing your question

Local model issues

  • Install required packages: pip install torch transformers sentence-transformers
  • Ensure sufficient disk space for model downloads

Error Messages

  • Red error messages: Critical issues requiring immediate attention
  • Yellow warnings: Issues that may affect functionality
  • Blue info messages: Helpful guidance and status updates

🔄 Recent Updates

  • ✅ Multiple LLM Support - Added Claude, Gemini, Ollama, and local models
  • ✅ ChromaDB Integration - Replaced FAISS with modern vector database
  • ✅ Enhanced Error Handling - Better user feedback and debugging
  • ✅ Persistent Storage - Knowledge base persists between sessions
  • ✅ Deployment Optimizations - Better compatibility and fallbacks
  • ✅ Improved UX - Loading spinners, progress indicators, and clear messaging

🤝 Contributing

  1. Fork the repository
  2. Create a feature branch
  3. Make your changes
  4. Test thoroughly
  5. Submit a pull request

📄 License

This project is licensed under the MIT License - see the LICENSE file for details.

🙏 Acknowledgments

  • LangChain for the powerful framework
  • Streamlit for the web interface
  • ChromaDB for vector storage
  • HuggingFace for open-source models and embeddings
Ver o original no GitHub

Bom para

Relacionados