Show HN: Instant GraphRAG over any Postgres database

Evokoa

Introducing Polygres

Structured records. Connected relationships. Semantic meaning. Polygres combines PostgreSQL hosting with native graph traversal, HNSW vector search, and hybrid retrieval APIs. Integrate using Evokoa/polygres-sdk.

Query everything from one place. #

Combine SQL, graph relationships, and vector similarity search inside a single query.

from polygres import Polygres

client = Polygres(
    api_key="<POLYGRES_API_KEY>",
    runtime_url="<POLYGRES_RUNTIME_URL>",
)
project = client.project()

page = project.hybrid.joint(
    [0.12, 0.34, 0.56],
    {"schema": "public", "table": "users", "id": "<USER_ID>"},
    graph_weight=0.40,
    vector_weight=0.60,
    limit=2
)

for doc in page.results:
    print(f"[{doc.score:.4f}] {doc.properties.get('title')}")
{
  "results": [
    {
      "schema": "public",
      "table": "documents",
      "id": "doc_architecture",
      "score": 0.9634,
      "graph_score": 0.92,
      "vector_score": 0.99,
      "properties": {
        "title": "Polygres Distributed Spec"
      }
    },
    {
      "schema": "public",
      "table": "documents",
      "id": "doc_pitch_deck",
      "score": 0.8415,
      "graph_score": 0.69,
      "vector_score": 0.94,
      "properties": {
        "title": "Evokoa V2 Pitch Deck"
      }
    }
  ]
}

Everything your agent needs. Wired together. #

All core components are open source. Self-host with pgGraph and pgVector, or use our managed cloud to skip the setup.

Postgres

A full PostgreSQL 17 instance. Use the ORMs, clients, and tools you already know. ACID-compliant, battle-tested, and exactly what you'd expect.

View on GitHub

pgGraph

Model relationships between your data natively. Traverse connections, discover neighbors, and map entity graphs without wrestling with recursive SQL joins.

View on GitHub

pgVector

Store and search embeddings directly in your database. HNSW indexing for fast similarity lookups at scale, with scalar filtering to narrow results precisely.

View on GitHub

Hybrid Search & Cloud

Hybrid search and Polygres scalable cloud combine Postgres, pgVector, and pgGraph together. Built with native Kubernetes and Docker scaling support, plus personalized support for pgGraph.

Create Account

Your stack. Your choice. #

Connect with any Postgres-compatible client, use our Python SDK Evokoa/polygres-sdk for agent workflows, or talk directly to the REST API. Get running in minutes, not hours.

from polygres import Polygres

client = Polygres(
    api_key="<POLYGRES_API_KEY>",
    runtime_url="<POLYGRES_RUNTIME_URL>",
)
project = client.project()

page = project.hybrid.joint(
    [0.12, 0.34, 0.56],  # Replace with your embedding vector
    {"schema": "public", "table": "users", "id": "<USER_ID>"},
    graph_weight=0.40,
    vector_weight=0.60,
    limit=5
)

for item in page.results:
    print(f"[{item.score:.4f}] {item.properties.get('title')}")

Postgres for the Agent era. #

Deploy your managed database in seconds, or integrate with your existing schema using the open-source SDK.

Start Building with Polygres

Get full access to relational table definitions, pgGraph traversals, and pgVector similarity search instantly.