Spaces:
Running
Running
Commit History
-
Add comprehensive perturbation testing system with E2E testsβ
Features: - New config.py with Pydantic models for perturbation test configuration - New concurrent_executor.py for parallel test execution with retry logic - Enhanced jailbreak testing with configurable num_techniques parameter - Enhanced counterfactual bias testing with comparison modes (vs_baseline, all_pairs, both) - New testing.py router with API endpoints for presets, demographics, and jailbreak prompts - Frontend PerturbationTestConfig.tsx component for configuration UI Tests: - 130 unit and integration tests for the perturbation testing system - Tests cover config validation, concurrent execution, jailbreak testing, bias detection - API integration tests for all new endpoints π€ Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>795b72e
-
Fix: Add langchain-text-splitters dependency for LangChain 0.3+ compatibilityβ
In LangChain 0.3+, the text_splitter module was moved to a separate package langchain-text-splitters. This fixes the "No module named 'langchain.text_splitter'" error when deploying to Hugging Face Spaces. Changes: - Added langchain-text-splitters>=0.3.0 to pyproject.toml dependencies - Updated import in text_chunking_strategies.py to use langchain_text_splitters - Explicitly install langchain-text-splitters in Dockerfile π€ Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude <noreply@anthropic.com>0b205c4
-
Add environment configuration template, enhance .gitignore, and refactor configuration loading. Removed deprecated scripts and improved setup process with a new setup script.
18efc55
-
add
52c4f6d -
add
ef5205a -
add
290f82e -
add
f62e2cd -
add
8e68b5a
-
ad
b0f37e6 -
ad
99618f1 -
ad
faafe4f -
ad
ce68073 -
ad
126ee52 -
ad
f779b69 -
ad
911b955 -
ad
9048e9b -
ad
91c2a08
-
ad
f0fc928 -
add
7bc750c -
add
53562c9 -
add
6b3524e -
add
0d2b318 -
add
f468e8b -
add
f4d0036 -
add
ba6c703 -
add
7bd46cb
-
add
c469fc9 -
add
1d4dcf3
-
add
b7aa617 -
add
0b23e94 -
add
ab39c06 -
add
5632471 -
add
5d07fc9 -
add
fd965f0 -
add
ea56a51 -
add
232e016 -
add
edf083b -
Add OpenAI Structured Outputs extraction methodβ
- Implement simple OpenAI structured outputs extractor using Pydantic models - Register as new production method 'openai_structured' in method registry - Support direct extraction without complex multi-agent workflow - Generate more complex knowledge graphs with better NEXT relationships - Include factory integration for seamless system integration - Build frontend with updated method selection capability7e807a3 -
add
09d46fc
-
add
f231a40 -
add
f29226f -
add
e20ad3b
-
add
13d4394
-
π Add real AI knowledge graph integration (API keys redacted)
5aa050e -
π― Add Automatic Sample Data System for New Usersβ
β Features implemented: β’ Automatic sample data insertion for empty databases β’ 3 curated examples showing increasing complexity levels β’ Pre-generated knowledge graphs for each sample trace β’ Non-destructive insertion (skips if data exists) π Sample data includes: β’ Basic Q&A: Simple arithmetic calculation (6 entities, 5 relations) β’ Research Task: Location/hours research (6 entities, 6 relations) β’ Multi-Agent: Algorithm analysis (10 entities, 16 relations) π§ Technical implementation: β’ backend/database/sample_data.py - Core data and insertion logic β’ backend/database/init_db.py - Modified to trigger insertion β’ Full transaction support with error handling β’ Comprehensive logging and status reporting π― User experience: β’ New users see working examples immediately β’ Examples serve as templates and learning materials β’ Demonstrates different trace types and complexity levels β’ No more empty system on first launch π Documentation: β’ Complete READMβ¦9e60d50 -
π― Add Smart Sample Data Preloading Systemβ
β¨ New Features: β’ Intelligent sample data preloader for better UX β’ Auto-preloads 6 diverse examples on first startup β’ Smart selection algorithm ensures variety in agent types, complexity, domains β’ Non-blocking background preload doesn't delay startup π§ Implementation: β’ backend/scripts/preload_sample_data.py - Core preloading logic β’ Modified backend/app.py with startup preload check β’ Enhanced trace metadata with rich categorization tags β’ Handles database deduplication and error recovery π Benefits: β’ New users get immediate examples to explore β’ No more empty 'My Traces' on first visit β’ Diverse samples showcase different agent interaction patterns β’ Knowledge graphs can be generated on-demand from preloaded traces π Bug Fixes: β’ Fixed 'str expected, not NoneType' errors in multiple modules β’ Added null checks for OPENAI_API_KEY environment variable β’ Resolved circular import issues in knowledge graph components π User Experience: β’ Immediaβ¦95b5fc0 -
π§Ή Code Cleanup and Optimizationβ
β Improvements: β’ Removed debug print statements from backend/app.py β’ Cleaned up unnecessary imports in backend/database/__init__.py β’ Optimized knowledge graph processing modules β’ Removed temporary debugging code across multiple files π― Focus areas: β’ Backend startup process optimization β’ Database initialization cleanup β’ Knowledge graph processing efficiency β’ Multi-agent extractor refinements π Impact: β’ Cleaner console output in production β’ Improved code maintainability β’ Better performance with fewer debug calls β’ Professional deployment-ready codebasefa515bb -
π Complete Fix for HF Spaces Sample Data Loadingβ
β Fixed All Critical Issues: 1. OPENAI_API_KEY None TypeError (4 files): - multi_agent_knowledge_extractor.py - knowledge_graph_merger.py - knowledge_graph_tester.py - knowledge_graph_processor.py β Added null checks: if OPENAI_API_KEY: os.environ[...] = OPENAI_API_KEY 2. SQLAlchemy ObjectNotExecutableError: - Added text() import and wrapped all SQL strings β All execute() calls now use text("SQL") format 3. SQLAlchemy ArgumentError - Parameter Format: - Converted positional parameters to named parameters - Changed ?, ?, ? β :param1, :param2, :param3 - Changed tuple params β dict params β All queries now use {"param1": value1, "param2": value2} format π― Local Test Results: β Application starts successfully β Sample data loads completely: β’ 1 trace inserted β’ 1 knowledge graph inserted β’ 5 entities inserted β’ 4 relations inserted β All database operations successful β Transaction committed successfully β Final counts match expected valuesβ¦de48716
-
Restore all AgentGraph high-level featuresβ
- Fixed evaluation module dependency in methods/__init__.py - Restored SlidingWindowMonitor, PromptReconstructor, testing, causal analysis - Fixed shared models import in agentgraph/shared/__init__.py - All advanced features now work without evaluation module dependency7d48c14 -
π« Fix trace upload failure by removing evaluation module dependencyβ
β Problem solved: β’ Fixed 'No module named evaluation' error in trace upload β’ Removed problematic imports causing circular dependency chain οΏ½οΏ½ Root cause: β’ agentgraph/__init__.py imported SlidingWindowMonitor β’ This triggered import chain: extraction β methods β baseline β evaluation β’ evaluation module doesn't exist in HF Spaces deployment β‘ Fix applied: β’ Commented out SlidingWindowMonitor and related imports in agentgraph/__init__.py β’ Removed problematic __all__ exports β’ Kept essential trace_management functionality intact π§ͺ Local testing: β All passed β’ trace_management import: β β’ analyze_trace_characteristics: β (structured_data type detected) β’ Database save_trace: β (ID: 1, trace_id: 8bdedf4b...) π Result: Trace file upload should work in HF Spaces now π Impact: Only affects advanced features, core trace upload remains functional9f8c703