kt-rag
Knowledge-graph-powered RAG sidecar for infrastructure querying
Turn your infrastructure graph into a searchable knowledge base. kt-rag feeds discovered infrastructure data — services, databases, Kubernetes clusters, DNS zones, CI/CD pipelines — into a LightRAG knowledge graph, then lets you ask questions in plain English.
What you can do:
- "Which services depend on Postgres?" — get a structured answer with confidence scoring
- "What's the blast radius if us-east-1 goes down?" — trace dependencies across providers
- "Show me all EC2 instances tagged
productionin eu-west-1" — filter by metadata - "Generate a runbook for the auth service" — auto-document from live graph data
- "What changed in my infrastructure since last week?" — incremental sync awareness
How It Works
┌─────────────┐ ┌──────────┐ ┌──────────┐ ┌──────┐
│ Discovery │───▶│ kt-rag │───▶│ LightRAG │───▶│ LLM │
│ Plugins │ │ seed/sync│ │ KG Store │ │ │
└─────────────┘ └──────────┘ └──────────┘ └──────┘
│
┌────┴────┐ ┌─────┴──────┐
│ API │ │ Query CLI │
│ Server │ │ / Library │
└─────────┘ └─────────────┘
- Plugins discover your infrastructure (AWS, GitHub, K8s, etc.)
- kt-rag seeds entities and relationships into LightRAG's graph store
- LightRAG indexes everything for hybrid search (graph + vector)
- Your question triggers retrieval-augmented generation through an LLM
Quick Start
Install
pip install knowledgetree-rag
Or from source:
git clone https://github.com/knowledgetree-dev/kt-rag.git
cd kt-rag
pip install -e .
Configure
Copy the template and set your LLM API key:
cp .env.example .env
# Edit .env with your LLM provider and API key
Minimal .env:
LLM_BINDING_API_KEY=sk-...
Seed the graph
Populate LightRAG from your infrastructure data:
kt-rag-seed
Query
kt-rag-query "which services depend on postgres?"
Example output:
Response: The following services depend on Postgres:
- auth-service (production, us-east-1)
- user-api (staging, eu-west-1)
- analytics-backend (production, us-west-2)
- docs-service (production, us-east-1)
Confidence: 0.82
Start the API server
kt-rag-server
# Listening on http://0.0.0.0:8085
curl -s http://localhost:8085/api/v1/rag/health
CLI Commands
| Command | Description |
|---|---|
kt-rag-seed |
Full seed from Knowledge Tree into LightRAG |
kt-rag-query "..." |
Natural-language query (supports --filter, --min-score, --profile) |
kt-rag-server |
REST API on port 8085 |
API
Query
POST /api/v1/rag/query
{
"question": "which services depend on postgres?",
"mode": "hybrid",
"filter": {"provider": "aws", "region": "us-east-1"},
"profile_name": "claude"
}
Response:
{
"response": "The following services depend on Postgres...",
"confidence": 0.82,
"metadata": {
"mode": "hybrid",
"profile": "claude",
"tokens_used": 1247
}
}
Profile Management
# List profiles
GET /api/v1/rag/profiles
# Create profile
POST /api/v1/rag/profiles
{
"name": "claude",
"provider": "anthropic",
"api_key": "sk-ant-...",
"model": "claude-sonnet-4-20250514"
}
# Delete profile
DELETE /api/v1/rag/profiles/{name}
Multi-LLM Profiles
Switch between LLM providers at query time:
# Named profiles (persisted)
kt-rag-query "what's my blast radius?" --profile claude
# Ad-hoc via headers (server mode)
curl -H "X-KT-LLM-Provider: anthropic" \
-H "X-KT-LLM-Api-Key: sk-ant-..." \
-d '{"question": "..."}' \
http://localhost:8085/api/v1/rag/query
Configuration
All config via environment variables or .env:
| Variable | Default | Description |
|---|---|---|
LLM_BINDING |
openai |
LLM provider (openai, anthropic, ollama, azure) |
LLM_BINDING_API_KEY |
— | API key |
LLM_BINDING_MODEL |
gpt-4o |
Model name |
EMBEDDING_BINDING |
openai |
Embedding provider |
EMBEDDING_BINDING_MODEL |
text-embedding-3-small |
Embedding model |
LIGHTRAG_DIR |
./output |
LightRAG working directory |
LIGHTRAG_PORT |
8085 |
API server port |
KT_API_URL |
http://localhost:8080 |
Knowledge Tree API URL |
KT_API_TOKEN |
— | Knowledge Tree auth token |
PG_URI |
postgresql://localhost:5432/knowledgetree |
Database (read-only seed) |
SYNC_INTERVAL |
300 |
Sync interval in seconds |
Confidence Scoring
Every query response includes a 0.0–1.0 confidence score based on:
- Retrieval quality (30%) — how well the query matched stored entities
- Context coverage (30%) — how completely the retrieved data covers the question
- Factual support (40%) — whether the LLM's answer is grounded in retrieved facts
Set a minimum threshold:
kt-rag-query "what repos are production?" --min-score 0.6
Contributing
PRs are welcome. For feature requests or major changes, open an issue first.
# Dev setup
python -m venv .venv && source .venv/bin/activate
pip install -e ".[dev]"
# Run tests
pytest tests/ -v --tb=short
# Type check
mypy .
Project Structure
├── seed.py # Full graph seed
├── sync.py # Incremental sync
├── query.py # CLI query entry point
├── server.py # FastAPI server
├── config.py # Environment config loader
├── profile_store.py # LLM provider profile persistence
├── scorer.py # Confidence scoring
├── kt_client.py # Knowledge Tree API client
├── tests/ # Test suite
└── .env.example # Config template
License
MIT
Metadata
Release files for knowledgetree-rag 0.1.0
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | Provenance | Actions |
|---|---|---|---|---|
| knowledgetree_rag-0.1.0.tar.gz | 10.6 kB | With provenance | Details for knowledgetree_rag-0.1.0.tar.gz |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Actions Reset |
|---|---|---|---|---|
| knowledgetree_rag-0.1.0-py3-none-any.whl | Python 3 | none | any | Details for knowledgetree_rag-0.1.0-py3-none-any.whl |
Total release size: 25.1 kB
Release files / knowledgetree_rag-0.1.0.tar.gz
| Download URL | knowledgetree_rag-0.1.0.tar.gz |
|---|---|
| Size | 10.6 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
62d97a7d0c3012eb40c35ba09975f6dc2ec5700a4fc91a70f7e61cad0679eaff
|
|
BLAKE2b-256 checksum How to use checksums |
98cfa9bcc034952178fb995f9e9cacd0fd13d96f8e0a31c28cd5def0e48b49b1
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/6.1.0 CPython/3.13.12
|
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Jun 9, 2026.
Transparency logRelease files / knowledgetree_rag-0.1.0-py3-none-any.whl
| Download URL | knowledgetree_rag-0.1.0-py3-none-any.whl |
|---|---|
| Size | 14.5 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
95ad310aeddf0fe63591eee8671e86f8f2ed3f702607a29243b5e4c560fd431d
|
|
BLAKE2b-256 checksum How to use checksums |
e5737ce5d634a30f5dfee61335d18a327b868cda1475a0e49a479eef20da898b
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/6.1.0 CPython/3.13.12
|
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Jun 9, 2026.
Transparency logSecurity for knowledgetree-rag 0.1.0
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
Full provenance
Both files in this release have been uploaded with provenance metadata.
Attestation for knowledgetree_rag-0.1.0.tar.gz
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Jun 9, 2026.
Transparency logA note on package security
PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.
To learn more about protecting yourself from malicious actors, visit our package security documentation.