feat: Mr-Dice_v0.82 add Mofdb agent

Author: NingWang-art

agents/matmaster_agent/MrDice_agent/agent.py

@@ -11,6 +11,9 @@
 from agents.matmaster_agent.MrDice_agent.optimade_agent.agent import (
     init_optimade_database_agent,
 )
+from agents.matmaster_agent.MrDice_agent.mofdb_agent.agent import (
+    init_mofdb_database_agent,
+)
 from agents.matmaster_agent.MrDice_agent.prompt import (
     MrDiceAgentDescription,
     MrDiceAgentInstruction,
@@ -23,6 +26,7 @@ def __init__(self, llm_config):
         optimade_agent = init_optimade_database_agent(llm_config)
         openlam_agent = init_openlam_database_agent(llm_config)
         bohriumpublic_agent = init_bohriumpublic_database_agent(llm_config)
+        mofdb_agent = init_mofdb_database_agent(llm_config)
         super().__init__(
             model=llm_config.gpt_5_chat,
             name=MrDiceAgentName,
@@ -32,6 +36,7 @@ def __init__(self, llm_config):
                 optimade_agent,
                 openlam_agent,
                 bohriumpublic_agent,
+                mofdb_agent,
             ],
             supervisor_agent=MATMASTER_AGENT_NAME,
         )

agents/matmaster_agent/MrDice_agent/bohriumpublic_agent/constant.py

@@ -3,7 +3,7 @@
 BOHRIUMPUBLIC_DATABASE_AGENT_NAME = 'bohriumpublic_agent'
 
 if CURRENT_ENV in ['test', 'uat']:
-    # OPTIMADE_URL="http://bekc1366122.bohrium.tech:50001/sse"
-    BOHRIUMPUBLIC_URL = 'http://jjxr1366132.bohrium.tech:50003/sse'
+    BOHRIUMPUBLIC_URL="http://bekc1366122.bohrium.tech:50003/sse"
+    # BOHRIUMPUBLIC_URL = 'http://jjxr1366132.bohrium.tech:50003/sse'
 else:
     BOHRIUMPUBLIC_URL = 'https://mr-dice-bohriumpublic-uuid1758096978.app-space.dplink.cc/sse?token=f395e31f5d5d48bc9d1d7018989e12bd'

agents/matmaster_agent/MrDice_agent/mofdb_agent/agent.py

@@ -0,0 +1,43 @@
+from dotenv import load_dotenv
+from dp.agent.adapter.adk import CalculationMCPToolset
+from google.adk.agents import BaseAgent
+from google.adk.tools.mcp_tool.mcp_session_manager import SseServerParams
+
+from agents.matmaster_agent.base_agents.job_agent import CalculationMCPLlmAgent
+from agents.matmaster_agent.constant import LOCAL_EXECUTOR, BohriumStorge
+from agents.matmaster_agent.MrDice_agent.mofdb_agent.constant import (
+    MOFDB_URL,
+)
+from agents.matmaster_agent.MrDice_agent.mofdb_agent.prompt import (
+    MofdbAgentDescription,
+    MofdbAgentInstruction,
+    MofdbAgentName,
+)
+from agents.matmaster_agent.MrDice_agent.constant import MrDice_Agent_Name
+
+load_dotenv()
+
+# Initialize MCP tools and agent
+mcp_tools = CalculationMCPToolset(
+    connection_params=SseServerParams(url=MOFDB_URL),
+    storage=BohriumStorge,
+    executor=LOCAL_EXECUTOR,
+)
+
+
+class Mofdb_Agent(CalculationMCPLlmAgent):
+    def __init__(self, llm_config):
+        super().__init__(
+            # model=llm_config.deepseek_chat,
+            model=llm_config.gpt_5_chat,
+            name=MofdbAgentName,
+            description=MofdbAgentDescription,
+            instruction=MofdbAgentInstruction,
+            tools=[mcp_tools],
+            render_tool_response=True,
+            supervisor_agent=MrDice_Agent_Name,
+        )
+
+
+def init_mofdb_database_agent(llm_config) -> BaseAgent:
+    return Mofdb_Agent(llm_config)

agents/matmaster_agent/MrDice_agent/mofdb_agent/constant.py

@@ -0,0 +1,9 @@
+from agents.matmaster_agent.constant import CURRENT_ENV
+
+MOFDB_DATABASE_AGENT_NAME = 'mofdb_agent'
+
+if CURRENT_ENV in ['test', 'uat']:
+    MOFDB_URL="http://bekc1366122.bohrium.tech:50004/sse"
+    # MOFDB_URL = 'http://jjxr1366132.bohrium.tech:50003/sse'
+else:
+    MOFDB_URL = 'https://mr-dice-mofdb-uuid1758704787.app-space.dplink.cc/sse?token=c15a46f6440d41b086aca38446d959bf'

agents/matmaster_agent/MrDice_agent/mofdb_agent/prompt.py

@@ -0,0 +1,105 @@
+MofdbAgentName = "mofdb_agent"
+
+MofdbAgentDescription = (
+    "An agent specialized in retrieving MOF (Metal-Organic Framework) structures from the MOFdb database. "
+    "Supports flexible queries by MOFid, MOFkey, name, database source, void fraction, pore sizes, and surface area. "
+    "Results can be exported in CIF or JSON format for structural visualization or metadata analysis."
+)
+
+MofdbAgentInstruction = """
+You are a MOF (Metal-Organic Framework) retrieval assistant with access to MCP tools powered by the **MOFdb database**.
+
+## WHAT YOU CAN DO
+You can call **one MCP tool**:
+
+1) fetch_mofs(
+       mofid: str | None = None,
+       mofkey: str | None = None,
+       name: str | None = None,
+       database: str | None = None,   # one of: "CoREMOF 2014", "CoREMOF 2019", "CSD", "hMOF", "IZA", "PCOD-syn", "Tobacco"
+       vf_min: float | None = None,
+       vf_max: float | None = None,
+       lcd_min: float | None = None,
+       lcd_max: float | None = None,
+       pld_min: float | None = None,
+       pld_max: float | None = None,
+       sa_m2g_min: float | None = None,
+       sa_m2g_max: float | None = None,
+       sa_m2cm3_min: float | None = None,
+       sa_m2cm3_max: float | None = None,
+       n_results: int = 10,
+       output_formats: list['cif'|'json'] = ['cif']
+   )
+   - Queries the MOFdb database.
+   - All parameters are optional; combine them for precise filtering.
+
+## FILTER OPTIONS
+- **mofid**: unique identifier string for a MOF (long chemical signature + tag).
+- **mofkey**: hashed key (unique code for each MOF entry).
+- **name**: short MOF name (e.g., `"tobmof-27"`).
+- **database**: select source (CoREMOF, CSD, hMOF, IZA, PCOD-syn, Tobacco, etc.).
+- **vf_min/vf_max**: void fraction range (unitless, from 0.0 to 1.0).
+- **lcd_min/lcd_max**: largest cavity diameter (Å).
+- **pld_min/pld_max**: pore limiting diameter (Å).
+- **sa_m2g_min/sa_m2g_max**: surface area per gram (m²/g).
+- **sa_m2cm3_min/sa_m2cm3_max**: surface area per volume (m²/cm³).
+- **n_results**: maximum number of MOFs to return.
+- **output_formats**:
+  - `"cif"` → crystallographic structure files
+  - `"json"` → complete metadata
+
+## HOW TO CHOOSE PARAMETERS
+- If user specifies a **name** → set `name` and optional `database`.
+- If user specifies a **MOFid** → set `mofid` directly.
+- If user specifies a **MOFkey** → set `mofkey`.
+- If user specifies **database** → set `database`.
+- If user specifies **pore sizes, void fraction, surface area** → set corresponding ranges.
+- If the user requests **metadata only** → use `output_formats=['json']`.
+- If the user requests **downloadable structure files** → use `output_formats=['cif']`.
+- If the user requests **both** → set `output_formats=['json','cif']`.
+
+## RESPONSE FORMAT
+The response must always include:
+1. ✅ A brief explanation of the filters applied.
+2. 📊 A Markdown table of the retrieved MOFs with columns (fixed order):
+   (1) Name  
+   (2) MOFid  
+   (3) MOFkey  
+   (4) Database  
+   (5) Void Fraction  
+   (6) LCD (Å)  
+   (7) PLD (Å)  
+   (8) Surface Area (m²/g)  
+   (9) Surface Area (m²/cm³)  
+   (10) Download link (CIF/JSON, based on `output_formats`)  
+   - All missing values must be shown as **Not Provided**.  
+   - The number of rows must exactly equal `n_found`.
+3. 📦 The `output_dir` path returned by the tool (for download/archive).
+If `n_found = 0`, clearly state that no matches were found, repeat the applied filters, and suggest loosening criteria. Do **not** generate an empty table.
+
+## DEMOS (用户问题 → 工具与参数)
+1) 用户：我想查 tobmof-27  
+   → Tool: fetch_mofs  
+     name: "tobmof-27"  
+     database: "Tobacco"  
+
+2) 用户：我想要比表面积 500–1000 m²/g 且 LCD 在 6–8 Å 之间的 MOF  
+   → Tool: fetch_mofs  
+     sa_m2g_min: 500  
+     sa_m2g_max: 1000  
+     lcd_min: 6.0  
+     lcd_max: 8.0  
+
+3) 用户：我有一个 MOFid：[O-]C(=O)c1cc(F)c(c(c1F)F)C(=O)[O-].[O-]C(=O)c1cc(F)c(cc1F)C(=O)[O-].[O-]C(=O)c1ccc(c(c1)F)C(=O)[O-].[Zn][O]([Zn])([Zn])[Zn] MOFid-v1.pcu.cat1，能帮我查一下吗？  
+   → Tool: fetch_mofs  
+     mofid: "[O-]C(=O)c1cc(F)c(c(c1F)F)C(=O)[O-].[O-]C(=O)c1cc(F)c(cc1F)C(=O)[O-].[O-]C(=O)c1ccc(c(c1)F)C(=O)[O-].[Zn][O]([Zn])([Zn])[Zn] MOFid-v1.pcu.cat1"  
+
+4) 用户：我知道一个 MOFkey：Cu.QMKYBPDZANOJGF.MOFkey-v1.tbo，帮我找一下对应的结构  
+   → Tool: fetch_mofs  
+     mofkey: "Cu.QMKYBPDZANOJGF.MOFkey-v1.tbo"  
+
+5) 用户：我想查找名叫 ABAYIO_clean 的 MOF，并导出所有信息  
+   → Tool: fetch_mofs  
+     name: "ABAYIO_clean"  
+     output_formats: ["cif","json"]  
+"""

agents/matmaster_agent/MrDice_agent/openlam_agent/constant.py

@@ -3,7 +3,7 @@
 OPENLAM_DATABASE_AGENT_NAME = 'openlam_agent'
 
 if CURRENT_ENV in ['test', 'uat']:
-    # OPTIMADE_URL="http://bekc1366122.bohrium.tech:50001/sse"
-    OPENLAM_URL = 'http://jjxr1366132.bohrium.tech:50002/sse'
+    OPENLAM_URL="http://bekc1366122.bohrium.tech:50002/sse"
+    # OPENLAM_URL = 'http://jjxr1366132.bohrium.tech:50002/sse'
 else:
     OPENLAM_URL = 'https://mr-dice-openlam-uuid1758096918.app-space.dplink.cc/sse?token=c2c0edb9c99f454896c4965ef1a5cf28'

agents/matmaster_agent/MrDice_agent/optimade_agent/constant.py

@@ -3,7 +3,7 @@
 OPTIMADE_DATABASE_AGENT_NAME = 'optimade_agent'
 
 if CURRENT_ENV in ['test', 'uat']:
-    # OPTIMADE_URL="http://bekc1366122.bohrium.tech:50001/sse"
-    OPTIMADE_URL = 'http://jjxr1366132.bohrium.tech:50001/sse'
+    OPTIMADE_URL="http://bekc1366122.bohrium.tech:50001/sse"
+    # OPTIMADE_URL = 'http://jjxr1366132.bohrium.tech:50001/sse'
 else:
     OPTIMADE_URL = 'https://material-data-retriever-uuid1754467958.app-space.dplink.cc/sse?token=16a1dd29dbb54cf9bf75748ba8df282d'

agents/matmaster_agent/MrDice_agent/prompt.py

@@ -22,10 +22,11 @@
   4. After execution, **collect their results, verify them, and merge into one unified Markdown table**.
 
 ## WHAT YOU CAN DO
-You have access to three sub-agents:
-- **bohrium_public_agent** → retrieves data from the Bohrium Public database (formula, elements, space group, atom counts, band gap, formation energy).
-- **optimade_agent** → retrieves data from OPTIMADE-compatible providers (multiple external materials databases, wide coverage).
-- **openlam_agent** → retrieves data from the OpenLAM internal database (formula, energy range, submission time filters).
+You have access to four sub-agents:
+- **bohrium_public_agent**
+- **optimade_agent**
+- **openlam_agent**
+- **mofdb_agent**
 
 ## HOW TO CHOOSE SUB-AGENTS
 - Default: select the **single most suitable sub-agent** that fully supports the query.
@@ -48,6 +49,9 @@
   - ✅ Supports: `formula`, `min_energy`, `max_energy`, `min_submission_time`, `max_submission_time`.
   - ❌ No support for space group, band gap, elements list, or logical filters.
   - support **energy window searches** and **time-based filters**.
+- **MOFdb**
+  - ✅ Supports all MOF-related queries: by **MOFid, MOFkey, name, database source**, or by **void fraction, pore sizes, surface area**.
+  - 🎯 Any request clearly about **MOFs** should be handled by MOFdb.
 
 💡 **Decision logic examples**:
 - If query is about **submission time** → use `openlam_agent`.
@@ -86,7 +90,7 @@
   - ✅ Just pass the retrieval requirements, and let each sub-agent handle its own parameters.
 
 ## EXECUTION RULES
-- User or higher-level agent instructions are always **clear and detailed**. Do not ask for confirmation; begin retrieval immediately.
+- User or higher-level agent instructions are always **clear and detailed**. Do not ask for confirmation or more details; begin retrieval immediately.
 - Always call the tool for a **real retrieval**; never simulate results or fabricate outputs.
 - If multiple agents are required, run them **sequentially**, not in parallel.
 - Each sub-agent works independently; never pass results from one to another.
@@ -101,8 +105,8 @@
 ## RESPONSE FORMAT
 The response must always include:
 1. ✅ A short explanation of which sub-agents were used and which filters were applied.
-2. 📊 A unified Markdown table with results from **all queried sources**.
-   - Columns (fixed order):
+2. 📊 Results presentation:
+   - For **crystal-structure agents** (BohriumPublic, OPTIMADE, OpenLAM), results **must** be shown in a unified Markdown table with columns (fixed order):
      (1) Formula
      (2) Elements
      (3) Atom count (if available; else **Not Provided**)
@@ -112,8 +116,9 @@
      (7) Download link (CIF/JSON)
      (8) Source database (`BohriumPublic`,`OPTIMADE` provider name, or `OpenLAM`)
      (9) ID
-   - Fill missing values with exactly **Not Provided**.
-   - Number of rows must equal the total `n_found`.
+     - Fill missing values with exactly **Not Provided**.
+     - Number of rows must equal the total `n_found`.
+   - For **non-crystal agents** (e.g., MOFdb), you do **not** need to force this schema.
 3. 📦 If multiple agents provide downloadable archives (`output_dir`), list all paths at the end.
 4. If the user explicitly requests additional attributes (e.g., lattice constants, density, symmetry operations):
    - **Before retrieval**: include these attributes in the query plan and instruct each sub-agent to provide them if available.