pytriz 0.2.0

A Python package for TRIZ (Theory of Inventive Problem Solving) tools and techniques.

PyTRIZ

A Python library for applying TRIZ (Theory of Inventive Problem Solving) — look up parameters, principles, and contradiction matrix results, or use LLM-powered agents to analyze engineering trade-offs.

Installation

pip install pytriz

Requires Python 3.12+.

Quick start

from pytriz import TRIZStore

store = TRIZStore()  # uses default embedding model, reads config from env

# Find TRIZ principles for a contradiction
principles = store.get_principles_from_matrix(
    improving_parameters=[1, 3],
    preserving_parameters=[17, 23],
)

# Search parameters and principles by description
params = store.search_parameters("improves durability", top_k=5)
principles = store.search_principles("segmentation", top_k=5)

TRIZStore builds the indexed corpus on instantiation — create it once and reuse it across your application. Semantic search uses a local embedding model by default, no API key needed.

LLM-powered analysis

import asyncio
from pytriz import TRIZStore
from pytriz import contradictions

store = TRIZStore()

result = asyncio.run(
    contradictions.analyze_contradiction(
        "Increasing blade thickness improves durability but increases weight.",
        store=store,
    )
)

print(result.contradiction)
print(result.improving_parameter)
print(result.preserving_parameter)

Set your LLM provider in a .env file:

DEFAULT_PROVIDER=openrouter    # openai | anthropic | mistral | openrouter | ollama | together
DEFAULT_MODEL=qwen/qwen3.6-35b-a3b
OPENROUTER_API_KEY=your-key-here

Explicit configuration

For full control — useful when building FastAPI apps, MCP servers, or any long-running service:

from pytriz import TRIZStore, get_embedder, get_model, ModelSettings

store = TRIZStore(
    embed_model=get_embedder(provider="ollama", model="nomic-embed-text", url="http://my-server:11434"),
)
llm = get_model(
    provider="anthropic",
    model_name="claude-sonnet-4-6",
    settings=ModelSettings(temperature=0.2),
)

result = await contradictions.analyze_contradiction(
    "Increasing blade thickness improves durability but increases weight.",
    store=store,
    llm=llm,
)

store and llm are independent — configure each separately and pass them where needed.

Embedding providers

PyTRIZ supports three embedding backends, configured via EMBEDDING_PROVIDER:

HuggingFace (default)

Downloads and runs the model locally using sentence-transformers. No API key needed for public models.

EMBEDDING_PROVIDER=huggingface
EMBEDDING_MODEL=sentence-transformers/all-MiniLM-L6-v2

For gated models (e.g. google/embeddinggemma-300m), set your HuggingFace token:

HF_TOKEN=your-token-here
EMBEDDING_MODEL=google/embeddinggemma-300m

Ollama

Runs embeddings locally via a running Ollama instance.

EMBEDDING_PROVIDER=ollama
EMBEDDING_MODEL=nomic-embed-text
OLLAMA_BASE_URL=http://localhost:11434   # optional, this is the default

OpenAI

EMBEDDING_PROVIDER=openai
EMBEDDING_MODEL=text-embedding-3-small
OPENAI_API_KEY=your-key-here

LLM providers

Provider	DEFAULT_PROVIDER	Required env var
OpenAI	openai	OPENAI_API_KEY
Anthropic	anthropic	ANTHROPIC_API_KEY
Mistral	mistral	MISTRAL_API_KEY
OpenRouter	openrouter	OPENROUTER_API_KEY
Together	together	TOGETHER_API_KEY
Ollama	ollama	— (uses OLLAMA_BASE_URL)

Environment variables reference

Variable	Default	Description
DEFAULT_PROVIDER	openrouter	LLM provider
DEFAULT_MODEL	qwen/qwen3.6-35b-a3b	LLM model name
EMBEDDING_PROVIDER	huggingface	Embedding backend
EMBEDDING_MODEL	sentence-transformers/all-MiniLM-L6-v2	Embedding model name
OLLAMA_BASE_URL	http://localhost:11434	Ollama host (used for both LLM and embeddings)
OPENAI_API_KEY	—	OpenAI API key
ANTHROPIC_API_KEY	—	Anthropic API key
MISTRAL_API_KEY	—	Mistral API key
OPENROUTER_API_KEY	—	OpenRouter API key
TOGETHER_API_KEY	—	Together AI API key
HF_TOKEN	—	HuggingFace token (for gated models)