nsip-skills 1.4.7

Breeding decision support tools for sheep genetics using NSIP EBV data

NSIP Search API Client

A Python client for the NSIP (National Sheep Improvement Program) Search API, reverse-engineered from http://nsipsearch.nsip.org

---

🐑 New to NSIP Tools? Check out the Getting Started Guide for step-by-step instructions on using these tools with Claude Desktop or other AI assistants. No programming experience required!

📚 Example Prompts: Browse the Examples Library for copy-paste prompts covering tools, workflows, resources, and skills.

---

Overview

This client provides programmatic access to sheep breeding data including:

• Animal details and genetic traits (EBVs)
• Pedigree/lineage information
• Progeny records
• Breed groups and trait ranges
• Search and filtering capabilities

Installation

Important: This package is NOT published to PyPI. It is only available through GitHub Releases. See DISTRIBUTION.md for our distribution policy.

From GitHub Release (Recommended)

Download the latest release from the releases page:

# Install from wheel (recommended)
pip install https://github.com/epicpast/nsip-api-client/releases/latest/download/nsip_client-1.3.5-py3-none-any.whl

# Or install from source distribution
pip install https://github.com/epicpast/nsip-api-client/releases/download/v1.3.5/nsip-client-1.3.5.tar.gz

From Git

# Install latest from main branch
pip install git+https://github.com/epicpast/nsip-api-client.git

# Install specific version tag
pip install git+https://github.com/epicpast/nsip-api-client.git@v1.3.5

From Source

git clone https://github.com/epicpast/nsip-api-client.git
cd nsip-api-client
pip install -e .

---

NSIP Skills: Breeding Decision Support Tools

The NSIP Skills module provides a suite of breeding decision support tools designed to help sheep breeders make data-driven genetic improvement decisions. These tools are available as both Python modules and Claude Code slash commands.

Why Use NSIP Skills?

Genetic improvement in sheep breeding requires balancing multiple factors: growth traits, maternal ability, carcass quality, and inbreeding risk. The NSIP Skills module automates complex calculations that would otherwise require extensive spreadsheet work or specialized software.

Key Benefits:

• Data-driven decisions: Replace guesswork with EBV-based analysis
• Inbreeding management: Automatically calculate and flag high-risk matings
• Multi-trait optimization: Balance competing breeding goals
• Time savings: Automate repetitive analysis tasks
• Accessible genetics: Make NSIP data actionable for all breeders

Quick Start

All tools can be run as Python modules or through Claude Code slash commands:

# Python module usage
uv run python -m nsip_skills.ebv_analysis LPN_ID1 LPN_ID2 LPN_ID3

# Claude Code slash command (when using Claude Code)
/nsip:ebv-analyzer LPN_ID1 LPN_ID2 LPN_ID3

The 10 NSIP Skills

Command	Purpose	Key Use Case
/nsip:flock-import	Import spreadsheet data and enrich with NSIP EBVs	Getting started with your flock data
/nsip:ebv-analyzer	Compare EBVs across animals	Evaluating purchase candidates
/nsip:inbreeding	Calculate inbreeding coefficients	Pre-mating risk assessment
/nsip:mating-plan	Optimize ram-ewe pairings	Breeding season planning
/nsip:progeny-report	Evaluate sires by offspring	Sire selection and evaluation
/nsip:trait-improvement	Multi-generation selection planning	Long-term genetic goals
/nsip:ancestry	Generate pedigree trees	Understanding genetic background
/nsip:flock-dashboard	Aggregate flock statistics	Flock performance overview
/nsip:selection-index	Calculate breeding indexes	Ranking animals by composite merit
/nsip:breeding-recs	AI-powered recommendations	Comprehensive breeding advice

Example Workflows

Workflow 1: Evaluating Ram Purchase Candidates

# Compare EBVs of rams you're considering
uv run python -m nsip_skills.ebv_analysis RAM_LPN_1 RAM_LPN_2 RAM_LPN_3

# Check progeny performance for proven sires
uv run python -m nsip_skills.progeny_analysis RAM_LPN_1 --compare RAM_LPN_2,RAM_LPN_3

# Calculate inbreeding risk with your ewes
uv run python -m nsip_skills.inbreeding --mating RAM_LPN_1,EWE_LPN_1

Workflow 2: Breeding Season Planning

# Import your flock data
uv run python -m nsip_skills.flock_import my_flock.csv

# Generate optimized mating plan
uv run python -m nsip_skills.mating_optimizer --rams rams.csv --ewes ewes.csv --goal terminal

# Get AI-powered recommendations
uv run python -m nsip_skills.recommendation_engine my_flock.csv --goal terminal

Workflow 3: Long-term Genetic Planning

# Analyze current flock performance
uv run python -m nsip_skills.flock_stats my_flock.csv --name "Valley Farm"

# Plan multi-generation improvement
uv run python -m nsip_skills.trait_planner my_flock.csv --targets '{"WWT": 5.0, "NLW": 0.20}'

Understanding EBVs

Estimated Breeding Values (EBVs) predict an animal's genetic merit for specific traits. Key traits include:

Trait	Description	Breeding Impact
BWT	Birth Weight	Lower values = easier lambing
WWT	Weaning Weight	Growth to weaning age
PWWT	Post-Weaning Weight	Growth potential
MWWT	Maternal Weaning Weight	Dam's milk/mothering effect
NLW	Number of Lambs Weaned	Fertility and lamb survival
YEMD	Yearling Eye Muscle Depth	Carcass muscling
YFAT	Yearling Fat Depth	Carcass fat cover

Selection Indexes

Selection indexes combine multiple traits into a single score based on economic weights:

Index	Focus	Best For
Terminal	Growth + carcass	Market lamb production
Maternal	Reproduction + milk	Replacement ewe selection
Range	Balanced traits	Extensive operations
Hair	Growth without wool	Hair sheep breeds

Detailed Documentation

For comprehensive documentation including:

• Detailed tool descriptions
• Input/output formats
• Real-world use cases
• Genetic concepts explained

See docs/nsip-skills.md

---

API Endpoints Discovered

Base URL

http://nsipsearch.nsip.org/api

Authentication

No authentication required - this is a public API.

Endpoints

Search & Metadata

• GET /search/getDateLastUpdated - Get database last update date
• GET /search/getAvailableBreedGroups - List all breed groups
• GET /search/getStatusesByBreedGroup - List animal statuses
• GET /search/getTraitRangesByBreed?breedId={id} - Get trait ranges for a breed
• POST /search/getPageOfSearchResults?page={p}&pageSize={size}&breedId={id} - Search animals

Animal Details

• GET /details/getAnimalDetails?searchString={lpnId} - Get animal details
• GET /details/getLineage?lpnId={lpnId} - Get pedigree/lineage
• GET /details/getPageOfProgeny?lpnId={lpnId}&page={p}&pageSize={size} - Get offspring

Usage

Basic Example

from nsip_client import NSIPClient

# Initialize client
client = NSIPClient()

# Get animal details
animal = client.get_animal_details("6####92020###249")
print(f"Breed: {animal['Breed']}")
print(f"DOB: {animal['DateOfBirth']}")

# Get progeny
progeny = client.get_progeny("6####92020###249")
print(f"Total offspring: {progeny['TotalCount']}")

# Get lineage
lineage = client.get_lineage("6####92020###249")

Search Animals

# Search by breed
results = client.search_animals(
    breed_id=486,  # South African Meat Merino
    page=0,
    page_size=15
)

for animal in results['Results']:
    print(animal['LpnId'])

Get Breed Information

# Get all breed groups
breed_groups = client.get_available_breed_groups()
for group in breed_groups:
    print(f"{group['Id']}: {group['Name']}")

# Get trait ranges for a breed
trait_ranges = client.get_trait_ranges_by_breed(486)

Complete Animal Profile

# Get all information at once
profile = client.search_by_lpn("6####92020###249")
print(profile['details'])
print(profile['lineage'])
print(profile['progeny'])

MCP Server

The NSIP API Client includes an MCP (Model Context Protocol) server that exposes NSIP sheep breeding data to LLM applications. The server provides intelligent context management with automatic response summarization to prevent LLM context window overflow.

Features

• 15 MCP Tools: 10 NSIP API tools + 5 Shepherd consultation tools
• Context Management: Automatic summarization of large responses (>2000 tokens)
• Smart Caching: 1-hour TTL cache with 40%+ hit rate target
• Multiple Transports: stdio (CLI), HTTP SSE (web), WebSocket (real-time)
• Performance Optimized: <3s startup, <5s tool discovery, 70% token reduction

Installation

The MCP server is included with the package and accessible via the nsip-mcp-server command:

# Install from GitHub release
pip install git+https://github.com/epicpast/nsip-api-client@v1.0.0

# Or using uvx (no installation required)
uvx --from git+https://github.com/epicpast/nsip-api-client@v1.0.0 nsip-mcp-server

Quick Start

Start the server (stdio mode - default):

nsip-mcp-server

Configure with Claude Desktop (claude_desktop_config.json):

{
  "mcpServers": {
    "nsip": {
      "command": "nsip-mcp-server",
      "env": {
        "MCP_TRANSPORT": "stdio"
      }
    }
  }
}

Available MCP Tools (15 Total)

NSIP API Tools (10)

Tool Name	Description	May Be Summarized
nsip_get_last_update	Get database last updated timestamp	No
nsip_list_breeds	List breed groups with individual breeds	No
nsip_get_statuses	Get status options for breed group	No
nsip_get_trait_ranges	Get trait value ranges for breed	No
nsip_search_animals	Search animals by breed and traits	Yes
nsip_get_animal	Get detailed animal information	Yes
nsip_get_lineage	Get animal lineage/pedigree	Yes
nsip_get_progeny	Get animal progeny list	Yes
nsip_search_by_lpn	Comprehensive animal lookup	Optional
get_server_health	Get server metrics and health status	No

Shepherd Consultation Tools (5)

Tool Name	Description	Parameters
shepherd_consult	General sheep husbandry advice	question, region
shepherd_breeding	EBV interpretation, mating advice	question, region, production_goal
shepherd_health	Disease prevention, nutrition, parasites	question, region, life_stage
shepherd_calendar	Seasonal planning, breeding schedules	question, region, task_type
shepherd_economics	Cost analysis, ROI, marketing	question, flock_size, market_focus

Transport Options

stdio (Default - for MCP clients):

nsip-mcp-server
# Or explicitly:
MCP_TRANSPORT=stdio nsip-mcp-server

Streamable HTTP (for web applications):

MCP_TRANSPORT=streamable-http MCP_PORT=8000 nsip-mcp-server

WebSocket (for real-time bidirectional communication):

MCP_TRANSPORT=websocket MCP_PORT=9000 nsip-mcp-server

Environment Variables

Variable	Description	Default	Example
MCP_TRANSPORT	Transport mechanism	stdio	streamable-http, websocket
MCP_PORT	Port for HTTP/WebSocket	None (required for non-stdio)	8000, 9000
MCP_HOST	Host address to bind to	0.0.0.0	127.0.0.1, localhost
MCP_PATH	Path for HTTP endpoints	/mcp	/api/mcp
TIKTOKEN_CACHE_DIR	Token cache directory	OS default	/tmp/tiktoken
LOG_LEVEL	Logging verbosity	INFO	DEBUG, WARNING

Context Management

The server automatically manages response sizes to prevent LLM context overflow:

Responses <=2000 tokens: Pass through unmodified Responses >2000 tokens: Automatically summarized to ~70% reduction

What's preserved in summaries:

• Identity: LPN ID, breed, sire, dam
• Progeny: Total count (not full list)
• Contact: All breeder information
• Top 3 traits by accuracy (>=50% accuracy only)

What's omitted:

• Low-accuracy traits (<50%)
• Verbose metadata
• Full progeny lists (count only)

Example summarized response:

{
  "lpn_id": "6####92020###249",
  "breed": "Katahdin",
  "sire": "6####92019###124",
  "dam": "6####92018###035",
  "total_progeny": 6,
  "contact": {
    "farm_name": "[FARM_NAME]",
    "phone": "[PHONE_NUMBER]"
  },
  "top_traits": [
    {"trait": "WWT", "value": 3.051, "accuracy": 0.71},
    {"trait": "BWT", "value": 0.246, "accuracy": 0.74},
    {"trait": "YWT", "value": 5.123, "accuracy": 0.68}
  ],
  "_summarized": true,
  "_original_token_count": 3500,
  "_summary_token_count": 1050,
  "_reduction_percent": 70.0
}

Caching

The server caches NSIP API responses with a 1-hour TTL to reduce API load:

• Cache Key: method_name:sorted_json_params
• Target Hit Rate: 40%+ in typical usage
• Max Size: 1000 entries
• Eviction: FIFO (oldest removed first)

Example:

# First call - API fetch
nsip_get_animal(search_string="6####92020###249")  # Cache miss

# Within 1 hour - cached response
nsip_get_animal(search_string="6####92020###249")  # Cache hit (fast)

# After 1 hour - fresh data
nsip_get_animal(search_string="6####92020###249")  # Cache expired

Error Handling

All MCP tools return structured errors following JSON-RPC 2.0 with LLM-friendly suggestions:

{
  "error": {
    "code": -32602,
    "message": "Invalid parameter: lpn_id",
    "data": {
      "parameter": "lpn_id",
      "value": "123",
      "expected": "Non-empty string with at least 5 characters",
      "suggestion": "Provide a valid LPN ID like '6####92020###249'"
    }
  }
}

Common error codes:

• -32602: Invalid parameters (check suggestion for fix)
• -32000: NSIP API error (verify API availability)
• -32001: Cache error (retry without cache)
• -32002: Summarization error (request smaller dataset)

Health Monitoring

Get server health and performance metrics (HTTP SSE/WebSocket only):

curl http://localhost:8000/health

Response includes:

• Discovery times (target: <5s)
• Summarization reduction (target: >=70%)
• Validation success rate (target: >=95%)
• Cache hit rate (target: >=40%)
• Concurrent connections (support: 50+)
• Startup time (target: <3s)

Performance

The MCP server is designed to meet these targets:

Metric	Target
Tool Discovery	<5 seconds
Summarization Reduction	>=70% token reduction
Validation Success	>=95% caught before API
Cache Hit Rate	>=40%
Concurrent Connections	50+ without degradation
Startup Time	<3 seconds

Advanced Usage

Example MCP tool call:

{
  "jsonrpc": "2.0",
  "method": "tools/call",
  "params": {
    "name": "nsip_get_animal",
    "arguments": {
      "search_string": "6####92020###249"
    }
  },
  "id": 1
}

Example search with traits:

{
  "jsonrpc": "2.0",
  "method": "tools/call",
  "params": {
    "name": "nsip_search_animals",
    "arguments": {
      "breed_id": 486,
      "page": 0,
      "page_size": 15
    }
  },
  "id": 2
}

For complete MCP server documentation, examples, and troubleshooting, see specs/001-create-an-mcp/quickstart.md.

API Response Structure

Animal Details Response

{
  "LpnId": "6####92020###249",
  "Breed": "Katahdin",
  "BreedGroup": "Hair",
  "DateOfBirth": "2/5/2020",
  "Gender": "Female",
  "Status": "CURRENT",
  "Sire": "6####92019###124",
  "Dam": "6####92018###035",
  "TotalProgeny": 6,
  "Genotyped": "No",
  "Traits": {
    "BWT": {"Value": 0.246, "Accuracy": 74},
    "WWT": {"Value": 3.051, "Accuracy": 71},
    ...
  },
  "ContactInfo": {
    "FarmName": "[FARM_NAME]",
    "ContactName": "[OWNER_NAME]",
    "Phone": "[PHONE_NUMBER]",
    "Email": "[EMAIL_ADDRESS]"
  }
}

Breed Groups

• 61 - Range
• 62 - Maternal Wool
• 64 - Hair
• 69 - Terminal

Common Traits (EBVs)

Growth Traits

• BWT - Birth Weight
• WWT - Weaning Weight
• MWWT - Maternal Weaning Weight
• PWWT - Post Weaning Weight
• YWT - Yearling Weight

Carcass Traits

• YEMD - Yearling Eye Muscle Depth
• YFAT - Yearling Fat
• PEMD - Post Weaning Eye Muscle Depth
• PFAT - Post Weaning Fat

Wool Traits (for wool breeds)

• YGFW - Yearling Greasy Fleece Weight
• YFD - Yearling Fibre Diameter
• YSL - Yearling Staple Length

Reproduction

• NLB - Number of Lambs Born
• NLW - Number of Lambs Weaned

Parasite Resistance

• WFEC - Weaning Fecal Egg Count
• PFEC - Post Weaning Fecal Egg Count

Indexes

• US Range Index
• US Hair Index
• SRC$ Index

Network Traffic Analysis

The API was reverse-engineered using Chrome DevTools by:

1. Loading http://nsipsearch.nsip.org/#!/search
2. Performing searches and viewing animal details
3. Capturing network requests to identify API endpoints
4. Analyzing request/response headers and payloads

Headers Used

Accept: application/json, text/plain, */*
User-Agent: Mozilla/5.0...

No authentication headers, cookies, or API keys required.

Limitations

• This is a reverse-engineered client based on public web traffic
• The API is not officially documented
• Endpoints may change without notice
• Rate limiting behavior is unknown
• Some endpoints (like breed list by group) may need additional discovery

Example: Your Sheep

client = NSIPClient()

# Get details for your sheep
my_sheep = client.search_by_lpn("6####92020###249")

print(f"Name/ID: {my_sheep['details']['LpnId']}")
print(f"Breed: {my_sheep['details']['Breed']}")
print(f"Farm: {my_sheep['details']['ContactInfo']['FarmName']}")
print(f"Total Progeny: {my_sheep['details']['TotalProgeny']}")
print(f"US Hair Index: {my_sheep['details']['Traits']['USHairIndex']['Value']}")

License

MIT License - see LICENSE for details.

Note: The NSIP data accessed through this client remains property of the National Sheep Improvement Program.