langchain-konstruct 0.2.0

LangChain integration for Konstruct - Typed Knowledge Graph for AI Agents

Konstruct × LangChain

LangChain retriever classes powered by Konstruct's typed knowledge graph engine.

---

What This Does

Standard RAG gives your LLM chunks of text. Konstruct gives it structured knowledge — concepts connected by typed relationships like causes, requires, contradicts, and enables.

Your LLM stops guessing and starts reasoning.

Installation

pip install langchain-konstruct

Get an API key at aerware.ai.

Quick Start

from langchain_konstruct import KonstructRetriever
from langchain.chains import RetrievalQA
from langchain_anthropic import ChatAnthropic

retriever = KonstructRetriever(
    api_key="your_konstruct_api_key",
    pack_ids=["causal-startup-failure", "market-dynamics"],
)

qa = RetrievalQA.from_chain_type(
    llm=ChatAnthropic(model="claude-sonnet-4-20250514"),
    retriever=retriever,
)

response = qa.invoke({"query": "What causes startups to fail?"})
print(response["result"])

Using the Hosted Corpus

Konstruct ships with 2,800+ pre-built knowledge packs across business, science, health, law, and engineering. Reference them by ID:

retriever = KonstructRetriever(
    api_key="your_api_key",
    pack_ids=["causal-startup-failure", "market-dynamics", "monetary-policy"],
)

Browse available packs at aerware.ai/library.

Using Your Own Knowledge Packs

Load custom packs from JSON files or dicts:

retriever = KonstructRetriever(api_key="your_api_key")

# From file
retriever.load_pack_from_file("packs/my-domain.json")

# From dict
retriever.load_pack({
    "id": "my_pack",
    "name": "My Domain Knowledge",
    "concepts": [
        {
            "id": "concept_a",
            "name": "Concept A",
            "category": "strategy",
            "importance": 0.9,
            "description": "What this concept means"
        }
    ],
    "edges": [
        {
            "source": "concept_a",
            "target": "concept_b",
            "relation": "causes",
            "weight": 0.85
        }
    ]
})

Retriever Options

retriever = KonstructRetriever(
    api_key="your_api_key",      # From aerware.ai
    pack_ids=["pack-1"],         # Hosted packs to load
    max_concepts=10,             # How many concepts to retrieve
    max_depth=2,                 # Edge traversal depth (1 or 2)
    search_type="advisory",      # "advisory" or "concepts"
    category_filter="strategy",  # Optional: filter by category
)

search_type:

• "advisory" — Returns structured guidance (key concepts + relationships + frameworks). Best for chat/QA.
• "concepts" — Returns raw concepts as separate documents. Best for custom pipelines.

Causal Reasoning

Use KonstructCausalRetriever when you need answers about prerequisites, dependencies, or causal chains:

from langchain_konstruct import KonstructCausalRetriever

retriever = KonstructCausalRetriever(
    api_key="your_api_key",
    pack_ids=["fintech-compliance"],
)

docs = retriever.get_relevant_documents("What must be in place before AML compliance?")
# Prioritizes "requires", "causes", "enables", "prevents" edges

Knowledge Pack Format

Packs are validated JSON with concepts, typed edges, and optional frameworks:

{
  "id": "my_pack",
  "name": "My Knowledge Pack",
  "concepts": [
    {
      "id": "product_market_fit",
      "name": "Product-Market Fit",
      "category": "strategy",
      "importance": 0.95,
      "description": "When a product satisfies strong market demand"
    }
  ],
  "edges": [
    {
      "source": "customer_discovery",
      "target": "product_market_fit",
      "relation": "enables",
      "weight": 0.9
    }
  ],
  "frameworks": [
    {
      "id": "pmf_framework",
      "name": "PMF Checklist",
      "trigger_concepts": ["product_market_fit"],
      "questions": ["Are users actively requesting access?"],
      "red_flags": ["Building without user feedback"],
      "green_lights": ["Users paying before launch"]
    }
  ]
}

Supported relation types: causes, requires, enables, prevents, conflicts_with, supports, contradicts, is_a, part_of, precedes, amplifies, diminishes, implements, depends_on, correlates_with, replaces, triggers, validates, similar_to

Reverse edges are inferred automatically — add causes and caused_by is created for you.

Explore Concept Relationships

# Explore outward from a specific concept
docs = retriever.explore_concept("product_market_fit", depth=2)

# List all loaded concepts
docs = retriever.get_all_concepts(category="strategy")

# Graph statistics
stats = retriever.get_stats()
print(f"Concepts: {stats['concept_count']}, Edges: {stats['edge_count']}")

Environment Variables

ANTHROPIC_API_KEY=your_anthropic_key
KONSTRUCT_API_KEY=your_konstruct_key   # from aerware.ai

Links

• Get API Key: aerware.ai
• Browse Packs: aerware.ai/library
• Docs: docs.aerware.ai/konstruct
• Issues: github.com/AerwareAI/konstruct-v2/issues

---

© 2026 AerwareAI — Proprietary