Cortrix turns documents, code, workflow events, and retrieval evidence into semantic records built for agentic systems.
The diagram is a system overview of the semantic storage layer: ingestion, hybrid retrieval, memory, source context, and agent-facing connector paths.
Docker or PG Extension Path
Choose docker-compose, standalone server, or pgcortrix PostgreSQL extension paths based on the stack you already operate.
Feedback Signals for Review
Observability and feedback signals help teams inspect what supported an agent answer. Automatic learning from those signals is a coming roadmap topic.
Scoped Connector Paths
Designed to connect with MCP, CLI workflows, HTTP upload, filesystem watchers, and framework adapters as each public integration matures.
Defining a New Category
Agent data workflows need more than isolated retrieval calls. They need a storage layer that can keep semantic context, memory, and traceability together.
Agent-Native
Designed around agent data workflows: semantic processing, memory, and traceability are treated as shared storage concerns rather than scattered integration code.
Semantic Storage
Documents are parsed, chunked, embedded, and indexed in a shared semantic layer so agent workflows can query meaning and source context together.
Built to Fit Existing Stacks
Designed to work with existing databases, tools, and agent workflows while giving retrieval, memory, and audit a shared semantic layer.
One Engine. Agent Data Foundations.
Cortrix organizes semantic ingestion, hybrid retrieval with reranking, interaction memory, and source-level context into a shared engine design for REST, MCP, SDK, and PostgreSQL-backed workflows.
Semantic Processing Chain (SPC)
Prepare common document types such as PDF, Word, and Markdown through the v1.0 SPC path: Docling parsing with OCR fallback, parent-child chunking, NER + summary enrichment, embedding, and indexing under launch verification.
Core
Hybrid Query + Reranker
Vector similarity (P-HNSW with WAL persistence) and BM25 keyword search are fused via RRF, then precision-reranked with bge-reranker-v2-m3 for agent-facing retrieval.
Session, trace, and agent headers help teams inspect what happened inside an agent workflow. Retrieval feedback learning remains a coming roadmap area rather than an automatic self-learning claim.
Core
Source-Level Traceability
Cortrix targets inspectable provenance beyond call indexes. Answers can link back to contributing chunks, documents, and conversation turns through citation tracing and retrieval attribution.
Core
Namespaces + Cross-NS Query
Use namespaces to isolate data by project, team, or tenant. Cross-namespace query keeps scatter-gather and unified ranking semantics explicit across agent workflows.
Core
Retrieval Quality Foundations
Parent-child chunking is part of the release quality foundation. Contextual retrieval, multi-query patterns, retrieval grading, and hypothetical-question indexing remain coming evaluation paths.
Core
pgCortrix Extension
For PostgreSQL-backed applications, pgCortrix is the path for bringing semantic storage closer to existing PostgreSQL infrastructure while preserving the option to run Cortrix as a separate service.
Integration
MCP Server
The Cortrix MCP server exposes 29 core tools plus 2 admin tools over stdio, so IDE agents can search, upload, manage memory, check task status, and trigger controlled admin imports.
Integration
Local Embedding Path (bge-m3)
ONNX Runtime integration with bge-m3 is the local embedding path for teams that want semantic search without making every query depend on an external embedding service.
Local
Connector Paths
Cortrix meets agent tools through MCP, REST, Python SDK, framework adapters, HTTP upload, directory watch, DB import, and PostgreSQL extension paths.
Paths
Python SDK + REST + Web UI
pip install cortrix is the Python client path, alongside HTTP APIs and a local dashboard for document management, search, and AI chat.
Interface
Run the Local Path
Start the server, verify readiness, then query the demo namespace through HTTP or the Python SDK.
1
Start Cortrix
Terminal
docker run -d -p 8420:8420 \
-v cortrix-data:/data \
cortrix/cortrix-demo:v1.0
curl -X POST http://localhost:8420/api/v1/query \
-H "Content-Type: application/json" \
-d '{"namespaces": ["demo"], "query": "What is Cortrix?", "top_k": 10}'
Or use the Python SDK:
Python
pip install cortrix
from cortrix import Cortrix
client = Cortrix(base_url="http://localhost:8420")
results = client.search("demo", "What is Cortrix?", top_k=10)
Built for Real Agent Data Workflows
Cortrix focuses on the shared data layer behind agent workflows: documents, retrieval context, memory, feedback signals, and traceability that are often stitched together with custom glue code.
Agent Workflow Storage
Agent workflows need persistent semantic memory and inspectable source context. Cortrix gives builders a shared storage foundation to evaluate those public release paths.
Reviewable Agent Context
Source-level traceability helps teams inspect which documents, chunks, and turns contributed to an answer or workflow step.
Workflow Orchestration
Use Cortrix as a shared semantic layer that MCP, REST, Python SDK, and framework-based workflows can call without each agent owning a separate retrieval stack.
Developer Knowledge Workflows
Use Cortrix as a semantic layer around documents, workflow context, and agent memory while keeping existing databases and tools in place.
How Cortrix Compares
A scope-aware view of a shared semantic storage layer versus maintaining many glue-code paths.
