https://github.com/ziatdinovmax/scilinkllm
https://github.com/ziatdinovmax/scilinkllm
Last synced: 2 months ago
JSON representation
- Host: GitHub
- URL: https://github.com/ziatdinovmax/scilinkllm
- Owner: ziatdinovmax
- Created: 2025-04-10T18:04:29.000Z (6 months ago)
- Default Branch: main
- Last Pushed: 2025-05-29T02:25:58.000Z (4 months ago)
- Last Synced: 2025-05-29T02:35:41.294Z (4 months ago)
- Language: Python
- Size: 10.1 MB
- Stars: 3
- Watchers: 1
- Forks: 1
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# SciLinkLLM
A framework aimed at bridging experimental observations with computational materials modeling and literature analysis using large language models. For now it is limited to microscopy images.
## Workflows
### Experiment to DFT Workflow
1. **Image Analysis**: The Microscopy Agent analyzes the uploaded image with contextual metadata and recommends structures for DFT simulations.
2. **Recommendation Selection**: User selects from structural recommendations
3. **Structure Generation**: Structure Generator Agent generates and executes ASE-based scripts to create the selected atomic structures
### Experiment to Literature Workflow
1. **Image Analysis for Claims**: The Microscopy Agent analyzes the uploaded image with contextual metadata and generates scientific claims based on observed features.
2. **Claim Selection**: User selects which claims to validate against existing literature
3. **Literature Search**: The OWL Literature Agent queries scientific databases to determine if similar observations have been reported and provides evidence from relevant publications.
4. **Novelty Assessment**: The system identifies potentially novel findings versus those already reported in literature.
### Experiment to Claims Workflow
1. **Image Analysis for Claims**: The Microscopy Agent analyzes the uploaded image with contextual metadata and generates scientific claims based on observed features.
2. **Claims Output**: The system creates a structured report of the claims for further use.
## Requirements
- Python 3.11+
- Google Generative AI API access (see [here](https://ai.google.dev/gemini-api/docs/api-key))
- ASE (Atomic Simulation Environment)
- OpenCV and Pillow for image processing
- FutureHouse API key for literature search workflow (only needed for exp2lit.py)
## How to use
### For Experiment to DFT Workflow:
1. Edit ```config.py``` to set:
- Image and experimental metadata file paths
- Model selection for analysis and generation
- Additional generation instructions
2. Run ```python exp2dft.py```
**Note**: In `exp2dft.py`, LLM generates and executes ASE code to build atomic structures. As a good general practice, it's recommended to run LLM-generated code in an isolated environment such as a Docker container or a virtual machine like Google Colab. The `example.ipynb` notebook is provided specifically for running the workflow in Google Colab's isolated environment.
### For Experiment to Literature Workflow:
1. Edit ```config.py``` to set:
- Image and experimental metadata file paths
- Model selection for analysis
- FutureHouse API key for OWL literature agent
2. Run ```python exp2lit.py```
### For Experiment to Claims Workflow:
1. Edit ```config.py``` as above for image and model settings
2. Run ```python exp2claims.py```