How It Works

Deterministic Intelligence Pipeline

CTK follows one explicit operational reasoning pipeline. Every output is evidence-backed, contradiction-aware, replayable, and safe-by-policy before action.

Evidence FirstContradiction AwareReplayable MemorySafety Gate

Explore Product Surface Compare Plans

Pipeline State

Deterministic chain execution for current workspace memory.

replay: activemode: deterministic

1. Capture

Evidence intake

Observe runtime status, logs, endpoints, containers, and repository metadata as raw evidence input.

2. Normalize

Entity contracts

Convert heterogeneous runtime findings into deterministic entities, scopes, and evidence contracts.

3. Infer

Graph reasoning

Infer relationships, confidence, and topology links with explicit contradiction-aware rule sets.

4. Adjudicate

Truth maintenance

Resolve conflicting signals and emit truth posture updates for each operational target deterministically.

5. Narrate

Operational memory

Write replayable timeline memory moments using Evidence → Contradiction → Adjudication → Final chain.

6. Act Safely

Policy-gated action

Issue or block recommendations through safety gates based on confidence, freshness, and contradiction state.

Replay Chain

Evidence to Final, Without Hidden Logic

1. Evidence

Signals, incidents, and runtime observations enter as traceable evidence records.

2. Contradiction

Inconsistencies are grouped, scored, and linked to affected entities and relationships.

3. Adjudication

Deterministic truth maintenance applies confidence penalties, freshness shifts, and final posture.

4. Final

Timeline memory emits replayable operational truth and policy-safe action guidance.

Runtime Paths

Source Mode and Provider selection define what CTK can observe

In desktop onboarding, users choose a Source Mode first, then map a Provider when needed. This pairing controls runtime probe coverage, log ingestion quality, and topology confidence depth.

Project Source Mode

Already Running

Attach to services that are already up and stream runtime signals without restart.

Runtime statusHealth/port checksLog signals

Docker

Use container metadata and service topology as the primary runtime observation surface.

Container inventoryService relationsRuntime and log visibility

Start with Toolkit

Boot project processes through Toolkit and keep deterministic observation from startup onward.

Managed process controlBoot-time evidenceReplay-ready timeline

Local Connected

Observe remote/cloud workloads from local desktop through configured endpoints and provider mappings.

Provider log streamRuntime checksCloud-oriented topology hints

Provider Mapping Options

AWS

Use profile/context and map CloudWatch + runtime scope to CTK evidence contracts.

Azure

Map Azure Monitor signals and runtime targets into the same deterministic intelligence chain.

GCP

Attach Cloud Logging and runtime mappings to keep replayable operational memory intact.

Kubernetes

Use cluster/namespace/resource mappings for runtime + inventory evidence ingestion.

Generic

Use custom endpoint/log/runtime mappings when platform-specific adapters are not required.

Operational Guarantees

Same Input, Same Output

Rules are explicit and deterministic. Same evidence state produces the same intelligence result.

Auditable Causal Replay

Every moment retains causal lineage so teams can replay why a conclusion was emitted.

Safety Before Action

Low confidence, stale evidence, or active contradictions can block unsafe recommendations.