Skip to content

Latest commit

 

History

History
132 lines (89 loc) · 3.32 KB

File metadata and controls

132 lines (89 loc) · 3.32 KB

Deepthought Technical Documentation Overview

Deepthought is a graph-based reasoning engine that ingests arbitrary JSON, models each atomic value as a datum, links them with edges, and continuously refines the strength of those links through a reinforcement-learning loop. There is no required schema and no domain-specific assumptions—you send JSON; Deepthought turns it into a living, weighted knowledge graph.

Getting Started

  1. API Access

You’ll need:

Base API URL

Authentication token or key (if your deployment requires one)

  1. Prepare Your Data

Send any valid JSON. Deepthought will:

Parse every primitive or composite value into datums.

Infer relationships (edges) from JSON structure (nesting, arrays, key/value pairs).

Example payload:

{ "user": { "name": "Ada Lovelace", "age": 36, "interests": ["mathematics", "computing", "poetry"] }, "metadata": { "source": "imported-profile" } }

  1. Send Your Request curl -X POST https://api.deepthought.io/v1/analyze
    -H "Content-Type: application/json"
    -H "Authorization: Bearer YOUR_TOKEN"
    -d @your-data.json

Making API Requests POST /v1/analyze

Submit JSON for ingestion and graph updates.

Field Description Body Any JSON object Auth header Authorization: Bearer (if required) Example Response { "graphId": "run-20250610-ab12", "summary": { "datumsProcessed": 25, "newDatums": 7, "edgesCreated": 42, "reinforcedEdges": 18 } }

GET /v1/health

Simple liveness check:

curl https://api.deepthought.io/v1/health

Core Concepts Concept Description Datum Atomic unit derived from any JSON key, value, or array element. Stored with a unique ID. Edge Directed relationship between two datums, inferred from JSON structure. Carries a weight ∈ [0, 1]. Graph The evolving store of all datums and edges. Each ingest run mutates this graph. Graph ID Returned per request for traceability (graphId). Edge Weights

New edges → initialized with a random weight 0 – 1.

Existing edges → adjusted by a reinforcement-learning algorithm that strengthens frequently co-occurring relationships.

How It Works (Pipeline)

Parsing – Recursively walk JSON, emit datums, infer edges.

Graph Lookup – Match datums against the graph; create any that are unknown.

Weight Adjustment –

Reinforce existing edges.

Add new edges with random weights.

Persistence – Store new/updated nodes and edges in the backing graph database.

Typical Use Cases Use Case What Deepthought Provides Knowledge-graph construction Feed semi-structured data to build a continuously evolving graph. Relationship inference Surface emergent links for recommendations, clustering, or analytics. ML feature engineering Export weighted edges as high-signal relational features. Data-quality gates Detect disconnected or weakly connected datums before ETL loads. Exploratory analysis Traverse the graph to uncover patterns or anomalies in large datasets. Submitting Bug Reports

Where

Email support@deepthought.io

Enterprise users: internal ticketing portal.

Include

Clear title

Full input JSON

graphId, timestamp, and response body

Expected vs. actual behavior

Logs or request IDs (if available)

Template

Title: New datums not reported in summary

Input: { "test": { "value": "unseen-datum" } }

Expected: summary.newDatums == 1 Actual: summary.newDatums == 0

GraphId: run-20250611-cdef