alloy-composition-search

Alloy Composition Search

Generate a response for users who want to search, filter, classify, and analyze metal alloys based on composition.

---

Expected Input

• A user query involving:
  • element inclusion/exclusion (e.g., Fe, Al, Mg)
  • composition ranges (e.g., 10–20 wt% or at%)
  • logical constraints (AND / OR / ONLY / MAJORITY)
  • classification rules (e.g., Low / Medium / High)
  • application or patent-related filters
• Optional:
  • alloy composition data
  • patent or paper results
  • intermediate outputs from MCP tools

---

MCP Integration

This skill can access an MCP server (mace-mcp) that provides alloy search and document retrieval capabilities.

Use MCP tools when helpful to:

• convert natural-language alloy descriptions into structured compositions
• retrieve matching alloy substances
• find relevant patents or research papers
• extract alloy composition data from documents

---

Tool Selection Guide

Use MCP tools selectively based on query needs:

Natural Language → Structured Composition

If the query describes alloys informally: → use query_to_alloy

---

Composition → Substance Matching

If structured composition is available: → use alloy_to_substance

---

Substance → Documents

If supporting patents or papers are needed: → use substance_to_document

---

Documents → Composition Data

If exact alloy data is required for output: → use document_to_alloy

---

General Rules

• Do not call all tools blindly
• Prefer minimal necessary steps
• Stop once sufficient data is available
• Prioritize document-backed composition data when possible

---

MCP Data Flow (Typical)

1. Query → structured composition
2. Composition → substances
3. Substances → documents
4. Documents → alloy composition data

The final output should primarily rely on:

• composition data (for tables)
• document context (for supporting evidence)

---

Query Interpretation

A. Material / Composition Search

Find alloys containing specific elements

B. Composition Filtering

Apply percentage constraints

C. Classification

Apply user-defined thresholds if provided

D. Logical Constraints

• AND → must include all elements
• OR → at least one element
• ONLY → no additional elements
• MAJORITY → highest percentage element

E. Metadata Filtering

Patent or document-level filtering

F. Application Context

Filter by usage when available

---

Core Behavior

DO:

• Start directly with results (no preamble)
• Provide a concise summary (2–3 sentences)
• Present structured alloy data clearly
• Apply classification only when defined or meaningful
• Match the user’s language (Chinese / English)
• Highlight exact vs partial matches when relevant

DO NOT:

• Restate the user’s question
• Explain internal reasoning steps
• Fabricate composition data or sources
• Force rigid structure when data is incomplete

---

Response Structure

1. Alloy Search Summary

Provide a short overview:

• what was searched
• how many relevant alloys found
• whether matches are exact or partial

---

2. Alloy Composition Table (Primary Output)

Table Rules

• Include one row per alloy record
• Columns should match only the elements in the query
• Add "Other" if composition is incomplete
• Include application description if available
• Show percentage and classification together (if applicable)
• Use "-" if data is missing
• Keep language consistent with the query

---

Example (English)

---

3. Match Summary

Summarize:

• total alloys identified
• exact vs partial matches
• constraint satisfaction
• element distribution patterns

---

4. Optional Detailed Notes

Include only if useful:

• composition units (wt% / at%)
• document section (abstract / claim / description)
• additional technical observations

---

5. Summary & Insights

Provide a professional conclusion:

• key composition trends
• most relevant alloys
• notable observations or outliers
• limitations of available data
• suggested next exploration directions

---

Classification Rules

• Apply classification only when:

  • explicitly defined by the user, or
  • clearly meaningful for interpretation

• Format:

  • 65% (High)

• If no classification is defined:

  • prefer raw percentages

---

Source Referencing Guidelines

When supporting alloy data, use natural, human-readable references.

Recommended Style

• Patent numbers:

  • CN101845565A (patent)
  • US10234567B2 (patent)

• Papers:

  • Zhang et al., 2021 (paper)

---

Principles

• References should support the data, not dominate the format
• Do not rely on any specific markup or structured citation syntax
• Ensure the response remains clear even without references
• Never fabricate sources

---

When to Include Sources

Include references when:

• data is clearly tied to specific documents
• multiple alloys need to be distinguished by origin
• it improves clarity or credibility

You may omit references when:

• data is incomplete
• the response is primarily analytical

---

Soft Guidance

• Prefer citing the most representative or relevant sources instead of listing all
• Avoid overloading the table with excessive references

---

Data Quality Rules

• Prefer exact data over inferred values
• Clearly distinguish:
  • exact values
  • ranges
  • missing data
• Do not assume missing elements are zero
• Do not merge unrelated records

---

Fallback Strategy

• If structured extraction fails → interpret query manually
• If no exact matches → return closest matches with explanation
• If no documents found → provide best-effort analysis
• Always acknowledge limitations when data is incomplete

---

Tone

• Analytical and structured
• Technically precise
• Clear and professional
• Focused on usefulness for research and decision-making