qortex
Knowledge that learns
Your AI assistant forgets everything between conversations. qortex adds a knowledge graph that learns from every interaction. One command to install. Zero config.
Install
Claude Code
claude mcp add qortex -- uvx qortex mcp-serve
Cursor / Windsurf
uvx qortex mcp-serve # add as stdio MCP server in settings
Any MCP client
pip install qortex[all] && qortex mcp-serve
What happens next
Once installed, your assistant automatically:
- Searches the knowledge graph before answering architecture questions
- Retrieves relevant concepts, relationships, and rules (not just similar text)
- Learns from your feedback: accepted results get boosted, rejected ones get suppressed
- Persists everything to SQLite so knowledge survives restarts
No config files. No API keys for the knowledge layer. Just start asking questions.
The difference
| Vanilla RAG | qortex | |
|---|---|---|
| Retrieval | Cosine similarity (what's textually similar) | Graph-enhanced (what's structurally relevant) |
| Context | Flat chunks | Concepts + typed edges + rules |
| Learning | Static | Adapts from every accept/reject signal |
| Cross-references | None | Traverses REQUIRES, REFINES, USES edges |
Prove it
Call qortex_compare to see the difference on your own data:
{
"summary": "Graph-enhanced retrieval found 2 item(s) that cosine missed, surfaced 1 rule(s), replaced 1 distractor(s).",
"diff": {
"graph_found_that_cosine_missed": [
{"rank": 3, "id": "security:JWTValidation", "score": 0.72}
],
"cosine_found_that_graph_dropped": [
{"rank": 4, "id": "security:PasswordHashing", "score": 0.68}
],
"rank_changes": [
{"id": "security:AuthMiddleware", "vec_rank": 3, "graph_rank": 1, "delta": 2}
]
}
}
Graph retrieval promotes structurally connected concepts (AuthMiddleware depends on JWTValidation) and demotes textually similar but unrelated results.
How it works
- Auto-ingest: Feed it docs, specs, or code. LLM extraction builds concepts, typed edges, and rules automatically.
- Graph retrieval: Queries combine vector similarity with structural graph traversal. Related concepts get promoted even if they don't share keywords.
- Adaptive learning: Every
qortex_feedbackcall updates retrieval weights. The system gets smarter the more you use it. - Persistent by default: SQLite stores the knowledge graph, vector index, and learning state across restarts.
For framework authors
agno KnowledgeProtocol
from qortex.adapters.agno import QortexKnowledgeSource
knowledge = QortexKnowledgeSource(domains=["security"])
agent = Agent(knowledge=knowledge)
LangChain VectorStore
from langchain_qortex import QortexVectorStore
vs = QortexVectorStore.from_texts(texts, embedding, domain="security")
retriever = vs.as_retriever()
See langchain-qortex for the standalone package.
Mastra MastraVector
import { QortexVector } from "@peleke.s/mastra-qortex";
const qortex = new QortexVector({ id: "qortex" });
await qortex.createIndex({ indexName: "docs", dimension: 384 });
const results = await qortex.query({ indexName: "docs", queryVector: q, topK: 10 });
See @peleke.s/mastra-qortex for the standalone package.
| Framework | Package | Language | Interface |
|---|---|---|---|
| agno | Built-in adapter | Python | KnowledgeProtocol |
| LangChain | langchain-qortex |
Python | VectorStore ABC |
| Mastra | @peleke.s/mastra-qortex |
TypeScript | MastraVector abstract class |
| Any MCP client | Built-in MCP server | Any | MCP tools (JSON-RPC) |
Install extras
| Capability | Install | What you get |
|---|---|---|
| Core + MCP tools | pip install qortex |
Knowledge graph, MCP server, vector-level tools. Consumers provide embeddings. |
| Text-level search | pip install qortex[vec] |
qortex embeds text with sentence-transformers. Adds ~2GB for PyTorch + model weights. |
| Persistent vectors | pip install qortex[vec-sqlite] |
SQLite-backed vector index. Without this, vectors are in-memory only. |
| Production graph | pip install qortex[memgraph] |
Memgraph backend for production-scale graph operations. |
| Everything | pip install qortex[all] |
All of the above. |
License
MIT
1 maintainer
Metadata
1 maintainer
Release files for qortex 0.8.1
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | |
|---|---|---|---|
| qortex-0.8.1.tar.gz | 9.3 MB | Details |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Reset |
|---|---|---|---|---|
| qortex-0.8.1-py3-none-any.whl | Python 3 | none | any | Details |
Total release size: 9.6 MB
Release files / qortex-0.8.1.tar.gz
| Download URL | qortex-0.8.1.tar.gz |
|---|---|
| Size | 9.3 MB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
1f14aaa5002aa3d7ea1cf9aec49b7e4765fa1c10aba1ee18146cac64e0f20825
|
|
BLAKE2b-256 checksum How to use checksums |
fc6fb9ef4b281d1f4cd014d66bd177b92ede1ceca3afed1b63b4def1a6b5f9fc
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/6.1.0 CPython/3.13.7
|
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 Feb 24, 2026.
Transparency logRelease files / qortex-0.8.1-py3-none-any.whl
| Download URL | qortex-0.8.1-py3-none-any.whl |
|---|---|
| Size | 231.4 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
18582f2afad75e7082f03bc1442fce20130534f4054bc2c03b525879c93be2fb
|
|
BLAKE2b-256 checksum How to use checksums |
4f44339859e457cf95102dd1fd94356ab3fdd88f2e87d560f4d6c875e73b9c67
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/6.1.0 CPython/3.13.7
|
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 Feb 24, 2026.
Transparency log
