v1.0.0 · backend rev 2c4a · 8819

The canonical knowledge layer for reliable decisions.

Gnostikon turns information into validated, revisable units of knowledge — gnoses — so people, teams, and agents can think, decide, and act on what they actually know. Offline-first by construction. Auditable by design.

Get started API reference Star on GitHub

scope personal · project·ethos mandatory · injected·retrieval offline-first

curlpythontypescript

op_2026_05_10_T7c2a

# Retrieve gnoses validated under the active ethos.
$ curl "https://api.gnostikon.dev/v1/retrieval" \
  -H "Authorization: Bearer $GN_KEY" \
  -H "Gn-Scope: project_atlas" \
  -d '{"query":"why did Q1 ingestion stall?","top_k":8}'

→ 200 OK · 4 gnoses · confidence 0.78 · revision 2c4a
  ethos enforced: scientific_rigor
  conflict surfaced: gn_8214 ↔ gn_7991

Trusted by teams who refuse to guess

◆ Aletheia Labs▲ Tessera Health∎ Boréal Research○ Forum Legal⌬ Helios Field Ops▮ Codex Studio

Offline-first by construction

WatermelonDB on the client, FastAPI on the server. Capture, validate, and retrieve gnoses without a connection — sync resolves on its own terms when you reconnect.

module · sync · revisions

One canonical contract

The OpenAPI spec is the source of truth. Backend, client, and SDKs are generated from it — schemas, types, validators stay aligned by construction, not by convention.

spec · openapi 3.1 · 47 endpoints

Intelligence built in

Selection, retrieval, and Aletheia validation ship with the platform. Confidence and conflict are first-class fields. The system reports what it knows — and what it doesn't.

aletheia · ethos · selection

II · Architecture

Two runtimes, one contract.

A Python/FastAPI backend that owns the canonical model. An Expo/React Native client that consumes it offline-first. Both shaped by the same OpenAPI spec — so the surfaces cannot drift.

BackendFastAPI · Python 3.12

The canonical layer.

Selection, retrieval, ethos, indexing jobs, telemetry. The server arbitrates conflicts and stores revisions; agents never decide alone.

POST /v1/gnoses
→ creates a gnosis under the active scope
  injects ethos, returns confidence + provenance

ClientExpo · React Native · WatermelonDB

The offline-first surface.

Local-first writes, eventual sync, fully indexed for retrieval. Field engineers, voice notes, citation work — all without depending on the network.

codex.gnoses.observe('project_atlas')
  .pipe(filter(by_confidence(0.6)))
→ live, validated stream

Aletheiavalidation layer

Doubt is a first-class field.

Every gnosis carries confidence, references, and conflict markers. Aletheia surfaces gaps and paradoxes before they become bad decisions.

# validation result
confidence: 0.62
conflict_with: ["gn_7991"]
gaps: ["missing_origin"]

Ethosgovernance · mandatory

Agents operate inside their mandate.

Ethos is injected before retrieval. Define principles, scopes, and limits per agent — the system refuses to operate outside them, and tells you why.

ethos.scientific_rigor:
  forbid: ["speculation", "undated_sources"]
  require: ["provenance", "confidence"]

III · Decide from what you actually know

Start with one gnosis.

The smallest canonical unit of validated knowledge. Capture it, ground it, watch the system challenge it. Everything else follows.

Read the docs Browse use cases