Spaces:
Paused
Paused
Commit History
-
Transform CyberScraper 2077 into a fully autonomous agentic framework
google-labs-jules[bot] +1: Greene-ctrlChanges: - Re-architected `WebExtractor.process_query` to automatically chain fetching and autonomous investigation. - Enhanced the agentic loop in `_call_model_with_tools` to be truly autonomous, persisting for up to 12 iterations until the task is judged complete. - Implemented a `task_complete` tool that agents MUST use to signal the final answer, ensuring high-quality, continuous investigation. - Added `get_html_source`, `get_page_info`, and `wait_for_element` to the browser toolset for better state perception. - Improved async stability in Streamlit by properly managing the event loop and ensuring resource cleanup. - Maintained full compatibility with existing Hugging Face Space deployment and Blablador LLM provider. - Verified functional autonomy on the live Space. Co-authored-by: Greene-ctrl <192867433+Greene-ctrl@users.noreply.github.com>b2acf55 -
Implement iterative agentic framework for CyberScraper 2077
google-labs-jules[bot] +1: Greene-ctrlSummary of changes: - Enhanced `WebExtractor` with a deep iterative agentic loop (up to 10 iterations). - Added `get_html_source`, `get_page_info`, and `wait_for_element` to browser tools for better state perception. - Updated system prompt to guide the AI through continuous investigation, persistence, and verification. - Improved async reliability in Streamlit by explicitly managing the event loop to avoid `RuntimeError`. - Refined browser fallback logic to be more robust on Linux/Docker environments. - Maintained full compatibility with Hugging Face Space deployment (Nginx proxy, FastAPI, Blablador LLM). - Verified functionality on live Space and addressed code review feedback. Co-authored-by: Greene-ctrl <192867433+Greene-ctrl@users.noreply.github.com>7dd8e08
-
Add conversational chatbot mode with on-demand data export
itsOwen- Transform from JSON-only extractor to intelligent conversational assistant - Add conversation history support for context-aware responses - Only return structured data (CSV/JSON/Excel) when explicitly requested - Add Rebecca (Cyberpunk 2077) personality to the AI - Let LLM handle greetings and chat naturally when no URL provided - Improve JSON extraction with multiple fallback methods47cb030 -
Update scraper with multiple improvements and fixes
itsOwen- Fix deprecated google.generativeai import, use langchain directly - Fix CAPTCHA handling: use 'load' instead of 'networkidle', reuse same page - Fix error detection for HTML content containing 'Error:' string - Increase Tor timeout to 120s for slow onion sites - Add Tor install/start instructions to error messages - Simplify Tor detection to check if port 9050 is open - Fix README URL anchors for Tor documentation - Add centralized error handler with user-friendly messages - Update models to gpt-4.1-mini and gpt-4o-mini - Truncate long website names (onion URLs) in chat history - Update Google Sheets detection to check for client_secret.json file - Reduce gap in Setup Help text - Various code refactoring and improvementse555dc3
-
current browser + gemini support added
itsOwen24f6114
-
launch default browser instance
itsOwen63fcb91
-
ui fix
itsOwena57b699
-
several fixes, removed logging
itsOwen65cb2b8
-
Google Sheets Upload + Ollama Fix
itsOwen4e6c165
-
UI changes + code refactor
itsOwen4acff5f
-
working perfectly, Optimizing filters
itsOwenac47ad6 -
ollama experimental
itsOwenab80480
-
changes
itsOwen827d181 -
few fixes
itsOwena8d4530 -
CyberScraper
itsOwenc2a036b