Release navigatorLOCAL

Current run 01

release

not run

NOT RUN

no receipt returned

Run a manifest to populate the causal evidence graph.

policy boundaryno automatic deploy

Run input manifest

baseline: not selected
candidate: not selected
dataset: not selected
context: not selected
owner: not returned

Request lifecycle

truthful state

READY FOR A NEW RUN—

READY
The graph will remain neutral until an actual receipt is returned.

Live runner logidle

Live runner events appear when a release gate starts.

Scenario templates

catalog / API required

Start the local API to load executable manifests.

build a custom run manifest-backed

Start from an executable API template, edit release identities, then submit the same guarded release endpoint. These fields are configured input, not runtime evidence.

base template release id baseline agent URN candidate agent URN changed dataset URN changed context URN request owner / email candidate reference

optional AI agent boundary real response required

Choose a provider transport only when a real local endpoint is running. The returned SQL must pass Themis’ structured contract, digest check, DuckDB execution, and oracle comparison.

agent transport agent endpoint engine name required for DataHub Analytics Agent

No custom manifest changes applied.

Replay cases

not run

No replay cases are selected until a receipt returns.

advanced manifest / API access

Release manifest JSON

Scenario templates populate this editor from GET /v1/scenarios. Raw manifest editing remains an advanced local escape hatch.

API bearer token / memory only Load saved receipt

DETERMINISTIC RELEASE POLICY / RECEIPT-DERIVED

NOT RUNNOT RUN

Run a real release manifest to build the evidence graph.

cases—

MCP—

traces—

write-back—

AI AGENT / RECEIPT PROVENANCE

Artifact supplier, not decision authority

Run a real release to populate provider, artifact, retrieval, tool, and trace provenance.

Receipt-backed states: NOT_CONFIGURED · USED · ERROR · SKIPPED_ABLATION. Live status never claims an invocation before the receipt returns.

NOT RUN

No receipt-backed agent invocation records.

Decision boundary

The agent can supply an artifact, but Themis has not calculated a release decision without a receipt.

REPLAYPLAN / RECEIPT EVIDENCE

Why these cases ran

ReplayPlan appears only in an API receipt; no selection, order, or confidence is claimed.

NOT RUN

Run a release to inspect the receipt-backed replay selection.

DATAHUB / WRITE-BACK RECEIPT

Durable status boundary

DataHub write-back appears only in an API receipt; no status is claimed before a run.

NOT RUN

API status: —
Verified: —
Receipt URN: —

Expanded evidenceselect a graph filter

DATAHUB / GATE INPUT

Metadata turns SQL into evidence.

provider-independent

01schemawhat exists

02lineagewhat is affected

03trustquality + owner

04replaydoes SQL behave

05receiptdurable decision

DataHub is the evidence source: metadata resolves the path, replay tests behavior, and the receipt records the decision.

verified evidence blocking / incomplete review / disabled select a node or edge to inspect its source