Skip to main content

Reference

This section is the reference for OKT's two process flows and its core concepts.

Processes​

Open Knowledge Tree runs two complementary processes. The Knowledge Flow turns a single source into a graph of facts, concepts, and syntheses. The Agentic Flow is what an agent does on top of that graph — research, query, and author reports — looping back to fetch more sources whenever the graph is thin.

Knowledge Flow (per-source, automatic)​

Given a URL or DOI, OKT fetches the content, decomposes it into atomic self-contained facts, embeds and deduplicates them by meaning, extracts concepts and aliases to build the graph, and incrementally accumulates summaries and an authoritative synthesis per concept. Seven stages, each a background job that enqueues the next. The moment a source is fetched the pipeline runs to completion without further input.

See Knowledge Flow for the stage-by-stage walkthrough.

Agentic Flow (agent-driven, iterative)​

An agent works on top of the graph in three phases — research (gather sources, open investigations, fetch them), query (search facts and concepts, read syntheses, walk related concepts), and reports (author markdown and get back per-sentence fact annotations). The phases are not a pipeline; the agent loops back to research whenever a query surfaces a thin area, and only authors a report once the graph covers the topic.

See Agentic Flow for the phase-by-phase walkthrough.

How they relate​

The Knowledge Flow is the substrate. The Agentic Flow is the loop an agent runs against that substrate. The Agentic Flow feeds the Knowledge Flow (every fetchAndProcessSource enqueues the 7-stage pipeline) and reads from it (every searchFacts, getConcept, getRelatedConcepts reads what the pipeline produced). The two processes together form the engine: sources in, annotated reports out, with a growing knowledge graph as the durable artifact in between.

   Agentic Flow (loop)
   +---------------------------------------------------+
   |  research -> query -> (thin?) -> research -> ...  |
   |       |         |                                  |
   |       v         v                                  |
   |  fetchAndProcessSource   searchFacts/getConcept    |
   |       |                  read syntheses             |
   +-------|---------|----------------------------------+
           v         ^
   Knowledge Flow (per source)
   +-----------------+
   |  source -> facts -> embed -> dedup                |
   |             -> concepts -> summaries -> synthesis  |
   +-----------------+

The Knowledge Flow runs unattended once a source is fetched. The Agentic Flow is where judgement lives — which sources to fetch, when the graph is enough, what to write in the report. The MCP tools are the agent's verbs across both processes.

Concepts​

The Concepts pages define the core artifacts OKT produces and stores: facts, concepts and contexts, summaries and synthesis, and reports and auto-annotation. See Concepts for the overview.

Examples​

The Examples subsection ships static, self-contained snapshots of real OKT meta-synthesis reports. Every sentence that has supporting facts is clickable — click it to see the underlying facts and their source URLs, the same interaction the in-app report view provides. See Example Meta-Syntheses.