deepmodeling / deepmodeling/build-your-agent

Issue Template of Build-Your-Agent Project

Open
#3 0 comments 0 reactions 0 assignees View on GitHub
Dominant language
TypeScript
Stars
38
Forks
12
PR merge metrics
No merged PRs in 30d

Description

# 1. If you want to propose your idea

### **\[Title of requirement/idea]**

#### **Background and objectives**

1. **Research scenario**: Describe your research scenario or problem area (such as material chemistry, biomedicine, computational simulation, etc.).

* Example: In protein molecular dynamics simulation, it is necessary to quickly extract hydrogen bond distribution from trajectory files and compare it with literature results.

2. **Current pain points**: Repetitive manual operations/poor cross-system data sharing/high threshold of professional knowledge, etc.

* Example: Manual analysis is time-consuming and relies on experience. It is hoped that the Agent will automatically call VMD/OVITO scripts and generate batch analysis and visualization reports.

3. **Expected results**: How do you want the Agent to solve this problem? We encourage decomposing tasks as much as possible

* **Example 1: Rapid screening of catalyst surface adsorption energy**

```plain text
Input: catalyst crystal surface structure (POSCAR format) + gas molecule library (xyz format)
→ Step 1 Structural preprocessing: call ASE to automatically build slab model, use pymatgen for symmetry optimization
→ Step 2 Adsorption site enumeration: use CatKit to generate possible adsorption sites (TOP/HCP/FCC)
→ Step 3 Automatically submit calculations: schedule DFT calculations through the Bohrium platform (ABACUS input template)
→ Step 4 Result analysis: extract adsorption energy and write it into a structured table (CSV), and automatically resubmit abnormal values
Output: adsorption energy ranking with confidence rating + CIF file of optimal structure
```

* **Example 2: Calculation of detonation properties of energetic materials**

```plain text
Input: molecular structure + potential function + impact direction and impact velocity range
→ Step 1 Structure prediction: call OMC25 for molecular crystal structure prediction (multi-configuration)
→ Step 2 Stability ranking: select the most stable structure based on density/energy
→ Step 3 Detonation velocity simulation: use LAMMPS program to perform multi-scale shock technique (MSST) calculation from large to small impact velocity
→ Step 4 Post-processing: automatically find the points before and after the decomposition on the P-V curve, and use the Hugoniot equation to fit the detonation pressure and detonation velocity of the Chapman-Jouget point
Output: detonation pressure and detonation velocity of the Chapman-Jouget point
```

#### **Suggested solution (optional)**

* Is there existing MCP tools or agents but not enough to meet the function?

* Do you have a preliminary technical path (such as tool combination, algorithm ideas)?

* Do you want to collaborate on development?

#### **Additional information**

* Related tools/data sources: software involved (such as LAMMPS, ASE), data formats (such as HDF5), and literature.

* Urgency/priority: ⭐️⭐️⭐️ (high, medium, low)

***

# 2. If you want to contribute your code

### **\[Tool contribution title]**

#### **Tool basic information**

* **Function description**: The core function of the tool (1\~2 sentences).

* Example: Data cleaning and format conversion tool based on `dpdata`

* **Input/output specification**: Input data type, output format.

* Example: Input `xyz` → Output `deepmd/npy`; `fmax = 50`

* **Dependencies**: Dependent frameworks/libraries (such as dpdata, PyTorch, RDKit).

#### **Adaptation status**

* \[ ] Already encapsulated in accordance with the MCP protocol

* \[ ] Provide standardized API

* \[ ] Existing call examples (with bohrium notebook or colab link)

* \[ ] Need community assistance in adaptation

## **General supplementary notes**

1. **Community interaction**: Before submitting an issue, please check the \\\[Tool Index]\\(link) and \\\[Project Roadmap]\\(link) to avoid duplication.

2. **Collaboration invitation**: If you need to form a team for development, you can @ members of related fields in the issue or add the `💬 Discussion/Collaboration` tag.

3) **Reward mechanism**: High-quality tools and requirements will be marked as `🏆 Featured` and displayed on the community homepage.

***

Contributor guide

No contributing guide indexed for this repository

Assessment

This issue has not been assessed yet.

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.