Release v0.3.0

📦 Semiconductor Process Time-Series Sample Data (v1.0)

This release contains sample time-series datasets collected from multiple key semiconductor manufacturing processes, including lithography, etching, CVD, ion implantation, and CMP. These datasets are intended for benchmarking AI/ML models for anomaly detection, predictive maintenance, and process optimization.

---

🏭 Included Process Datasets

Each dataset contains multivariate time-series sensor readings for a specific semiconductor process step. All datasets are structured in relational table format (CSV) with timestamped records per wafer/lot.

Table Name	Process Name	Description
SEMI_PHOTO_SENSORS	Lithography (Photolitho)	Exposure dose, focus, temperature, alignment errors, etc.
SEMI_ETCH_SENSORS	Dry Etching	RF power, gas flow rates, chamber temp/pressure, etc.
SEMI_CVD_SENSORS	Chemical Vapor Deposition	Precursor gas flows, susceptor temp, deposition rate, etc.
SEMI_IMPLANT_SENSORS	Ion Implantation	Beam current/energy, dose, wafer rotation, vacuum pressure, etc.
SEMI_CMP_SENSORS	Chemical Mechanical Polishing (CMP)	Pressure, rotation, slurry flow/temp, endpoint signal, etc.

---

🧬 Common Schema Fields

All process sensor tables include:

• PNO: Unique measurement ID (Primary Key)
• EQUIPMENT_ID: Equipment identifier
• LOT_NO, WAFER_ID: Lot and wafer tracking
• TIMESTAMP: Measurement timestamp
• Sensor-specific parameters (See table-level documentation)

---

🧪 Anomaly Annotations

This release includes ground-truth anomaly annotations for the SEMI_CMP_SENSORS dataset only. These annotations were generated to support research in time-series anomaly detection, particularly for evaluating point anomalies and collective trend shifts.

✅ Annotation Summary (ex. CMP Dataset)

• EQUIPMENT_ID: CMP_001
• LOT_NO: LOT250001A
• Total Anomalies: 125 (approx. 4.17% of records)
• Annotated Variables:
  • RETAINER_PRESSURE: 90 anomalies (point anomalies)
  • REMOVAL_RATE: 40 anomalies (collective trend anomalies)

📊 Variable-wise Details

Variable	Anomaly Type	Count	Parameters	Notes
RETAINER_PRESSURE	point_global_outliers_latter_half	90	ratio=0.03, factor=4.0, radius=8	Isolated point anomalies mostly in the second half of the series
REMOVAL_RATE	collective_trend_outliers_latter_half	40	ratio=0.015, factor=0.8, radius=10	Consecutive trend-based anomalies (e.g., index range 2569–2588, 2934–2953)

ℹ️ All anomaly positions are provided as row indices. Users can map them to actual timestamps in the CSV to retrieve the precise anomaly time.

---

📈 Example Use Cases

• Time-series anomaly detection using multivariate sensor readings
• Threshold-based health monitoring of process equipment
• AI model development for fault prediction in semiconductor fabs
• Rule-based alert generation with configurable thresholds

---

🔒 Notes

• All sensor values are synthetic or anonymized for safe sharing
• Normal operating ranges are included for select parameters
• Detailed data dictionary is available in each CSV header or README

---

🏷 Version

v1.0 – Initial release with representative samples from 5 process steps

---

Please cite this dataset if used in publications or model training. For questions or collaboration inquiries, contact the maintainer.

Release v0.2.9

- feat: improve error handling and fallback mechanisms in RAGSearchTool for document search and vectorization
- feat: enhance tool mapping and registration in PrismLLMService
- fix: enhance error handling in PostgreSQLDataStore execution methods
- feat: refactor WorkflowManager to support asynchronous execution of workflows
- feat: add document upload functionality to RAGSearchTool
- feat: add document upload and existence check methods to RAGSearchTool

Release v0.2.8

Release v0.2.8 - Enhanced RAGSearchTool with document upload

Release v0.2.7

fix: correct settings attribute names for consistency (v0.2.7)

- Fix settings.vllm_openai_base_url -> settings.VLLM_OPENAI_BASE_URL
- Fix settings.openai_api_key -> settings.OPENAI_API_KEY
- Add missing WEAVIATE_URL and WEAVIATE_API_KEY to config
- Ensure all settings references use correct uppercase naming
- Bump version to 0.2.7

Release v0.2.6

feat: optimize tools module imports with lazy loading (v0.2.6)

- Remove direct imports of specific tool implementations from __init__.py
- Implement lazy loading for RAGSearchTool, ComplianceTool, MemorySearchTool
- Add factory functions for tool creation with lazy imports
- Update main.py to use lazy imports for better performance
- Maintain backward compatibility for all existing usage patterns
- Bump version to 0.2.6

Release v0.2.5

fix: Update version to 0.2.5 in all __init__.py files

Release v0.2.4

feat: Add mem0ai dependency and specify all dependency versions - Add…

Release v0.2.3

refactor: Remove Weaviate from prism-core Docker Compose - Each agent…

Release v0.2.2

fix: Change package name from prism-core to prism_core for clarity

- Update package name in pyproject.toml
- Update version to 0.2.2
- Update PRISM-Orch dependencies to use prism_core
- Improve package naming clarity

Release v0.2.1

feat: Add AgentManager and WorkflowManager to core.agents module

- Add AgentManager for agent lifecycle management
- Add WorkflowManager for orchestration workflows
- Move tools from PRISM-Orch to prism-core for reusability
- Update version to 0.2.1
- Improve configuration structure
- Add factory functions for tool creation