Skip to main content

​
Knowledge Graph API

The Knowledge Graph API provides endpoints for retrieving and managing the knowledge graph structure, including entities, relationships, and graph visualization data.

​
Quick Start

import requests

# No authentication needed!
response = requests.get("https://vedaya-kge.fly.dev/graph/label/list")
labels = response.json()
print(f"Available labels: {labels}")

# Get knowledge graph for a specific label
response = requests.get(
    "https://vedaya-kge.fly.dev/graphs",
    params={"label": "machine_learning", "max_depth": 2}
)
graph_data = response.json()

​
Get Graph Labels

Get all available graph labels in the knowledge graph.

​
Response

Returns a list of all graph labels as an array of strings. 
curl -X GET "https://vedaya-kge.fly.dev/graph/label/list"

[
  "machine_learning",
  "neural_network",
  "artificial_intelligence",
  "deep_learning",
  "data_science",
  "computer_vision",
  "natural_language_processing"
]

​
Get Knowledge Graph

Retrieve a connected subgraph of nodes where the label includes the specified label.

​
Query Parameters

​
label
string
required
Label to get knowledge graph for
​
max_depth
integer
default:"3"
Maximum depth of graph traversal (min: 1)
​
max_nodes
integer
default:"1000"
Maximum nodes to return (min: 1)

​
Response

Returns a knowledge graph for the specified label. When reducing the number of nodes, prioritization is based on:
  1. Hops (path) to the starting node take precedence
  2. Followed by the degree of the nodes
curl -X GET "https://vedaya-kge.fly.dev/graphs?label=machine_learning&max_depth=2&max_nodes=100"

{
  "nodes": [
    {
      "id": "machine_learning",
      "label": "Machine Learning",
      "properties": {
        "type": "concept",
        "importance": "high"
      }
    },
    {
      "id": "neural_network",
      "label": "Neural Network",
      "properties": {
        "type": "technology"
      }
    }
  ],
  "edges": [
    {
      "source": "machine_learning",
      "target": "neural_network",
      "relationship": "includes",
      "weight": 0.9
    }
  ]
}

​
Check Entity Exists

Check if an entity with the given name exists in the knowledge graph.

​
Query Parameters

​
name
string
required
Entity name to check

​
Response

Returns a dictionary with an ‘exists’ key indicating if the entity exists. 
{
  "exists": true
}

curl -X GET "https://vedaya-kge.fly.dev/graph/entity/exists?name=neural_network"

{
  "exists": true
}

​
Update Entity

Update an entity’s properties in the knowledge graph.

​
Request Body

​
entity_name
string
required
Name of the entity to update
​
updated_data
object
required
Updated properties for the entity
​
allow_rename
boolean
default:"false"
Whether to allow renaming the entity

​
Response

Returns the updated entity information. 
curl -X POST "https://vedaya-kge.fly.dev/graph/entity/edit" \
  -H "Content-Type: application/json" \
  -d '{
    "entity_name": "machine_learning",
    "updated_data": {
      "description": "Advanced pattern recognition techniques",
      "category": "AI Technology",
      "importance": "high"
    },
    "allow_rename": false
  }'

{
  "entity_name": "machine_learning",
  "updated": true,
  "properties": {
    "description": "Advanced pattern recognition techniques",
    "category": "AI Technology",
    "importance": "high"
  }
}

​
Update Relation

Update a relation’s properties in the knowledge graph.

​
Request Body

​
source_id
string
required
ID of the source entity
​
target_id
string
required
ID of the target entity
​
updated_data
object
required
Updated properties for the relation

​
Response

Returns the updated relation information. 
curl -X POST "https://vedaya-kge.fly.dev/graph/relation/edit" \
  -H "Content-Type: application/json" \
  -d '{
    "source_id": "entity_123",
    "target_id": "entity_456",
    "updated_data": {
      "relationship_type": "influences",
      "strength": 0.8,
      "confidence": 0.95
    }
  }'

{
  "source_id": "entity_123",
  "target_id": "entity_456",
  "updated": true,
  "properties": {
    "relationship_type": "influences",
    "strength": 0.8,
    "confidence": 0.95
  }
}

​
Complete Knowledge Graph Workflow

Here’s how to work with the knowledge graph after uploading documents: 
import requests
import time
from openai import OpenAI

# 1. Upload documents (creates knowledge graph automatically)
docs = ["Document about AI and machine learning..."]
requests.post(
    "https://vedaya-kge.fly.dev/documents/texts",
    json={"texts": docs}
)

# 2. Wait for processing
time.sleep(5)

# 3. Explore the knowledge graph
labels = requests.get("https://vedaya-kge.fly.dev/graph/label/list").json()
print(f"Discovered entities: {labels[:10]}")

# 4. Get graph for specific concept
graph = requests.get(
    "https://vedaya-kge.fly.dev/graphs",
    params={"label": labels[0], "max_depth": 2}
).json()

# 5. Query using graph-aware RAG
client = OpenAI(api_key="sk-dummy", base_url="https://vedaya-kge.fly.dev/v1")
response = client.chat.completions.create(
    model="vedaya-global",  # Use global mode for relationship queries
    messages=[{"role": "user", "content": f"What is related to {labels[0]}?"}]
)
print(response.choices[0].message.content)

​
Knowledge Graph RAG Modes

Different RAG modes leverage the knowledge graph differently:
ModeModel NameGraph Usage
Naivevedaya-naiveNo graph usage, keyword only
Localvedaya-localEntity-focused retrieval
Globalvedaya-globalRelationship traversal
Hybridvedaya-hybridCombined approach
Use vedaya-global for best results when querying relationships and connections in the knowledge graph.