Features
Review this build’s configuration and prepare a change proposal.
Reading feature configuration…
Feature guide
Configuration and evidence requirements
Features shows configuration compiled into this build and prepares a reviewed change request. Operator access is required. It does not activate services from the browser.
Read the four states
- This build: the checked or unchecked flag in the authored site configuration used to build this app.
- Proposed setting: a local checkbox value, saved in this browser while operator access remains available.
- Readiness: an audited blocker, owner and required flags. No catalog blocker means a proposal is permitted; it is not proof of production acceptance.
- Runtime observation: retained connector evidence with its original observation and expiration times. Fresh evidence does not establish the deployed revision. Missing evidence remains unknown. Other features are marked not instrumented.
Use Details to inspect the source description, owner, dependencies, blockers and evidence timestamps. Search and category filters narrow the catalog; Blocked and Changed filter its state.
Prepare a proposal
- Change the proposed settings and enter a reason.
- Open Review proposal and check every before and after value.
- Export the Markdown request and JSON plan together.
- Have a repository operator validate and apply the plan through the normal review workflow.
A proposal contains a fingerprint of the configuration and readiness catalog. If either changes, discard the stale draft and review the current catalog again. Unlocking a different operator session clears the local proposal. Browser storage can be disabled; in that case the page keeps the draft only while it is open.
Apply a reviewed plan
The repository operator runs these commands from a current checkout:
bun run cli site apply-flags feature-proposal.json --check --json
bun run cli site apply-flags feature-proposal.json --json
The check validates the complete batch without changing files. Application uses the existing site-edit lock and updates the authored flag task list and generated map together. Unknown fields, duplicate keys, stale fingerprints, mismatched previous values, blocked enables and unsatisfied dependencies reject the whole plan.
Review the diff, run the repository release checks and land the pull request. Build and deploy through the normal release process when deployment is authorized. Exporting or applying a plan does not deploy anything.
Source and runtime limits
The authored flag task list remains the authority. Generated configuration and descriptions are checked for drift. The browser has no feature mutation endpoint or provider credentials.
Read-only Novig observations are implemented; comparison delivery is not connected. Several canonical flags describe cleaning status without switching live acquisition. Partner delivery, settlement and analytics retain their production acceptance blockers. Turning a flag on cannot supply missing transports, provider acceptance or results evidence.
Authored feature configuration
This table comes from the flag task list in docs/site/flags.mdpart. Configured
on or off describes repository configuration; authored intent does not prove
implementation, readiness, deployment or a live observation. Use the catalog's
owner, dependencies and blockers when reviewing a proposal.
| Flag | Configured | Authored intent |
|---|---|---|
canonical.4c | off | The 4C board goes through the cleaning adapter instead of the legacy path |
canonical.f402 | off | F402 ticker rows go through the cleaning adapter instead of the legacy path |
canonical.replay | off | Replay seed bets go through the cleaning adapter instead of the legacy path |
canonical.weblog | off | Weblog rows go through the cleaning adapter instead of the legacy path |
connector.f402.ip | off | The IP tracker runs through the connector runtime |
connector.f402.settle | off | F402 settlement reads run in shadow through the connector runtime; nothing downstream reads them |
connector.f402.ticker | off | The F402 ticker pull lands raw batches, checks the raw schema and writes source health through the connector runtime |
connector.f402.weblog | off | The web log extract lands raw envelopes and writes source health through the connector runtime |
connector.fourc.board | off | 4C board refreshes land raw, record frame and delivery evidence and coverage gaps through the connector runtime |
connector.model.sam | off | The Sam model page runs through the connector runtime with a page fingerprint |
connector.model.sasser | off | The Sasser model page runs through the connector runtime with a page fingerprint |
connector.results.scores | off | Licensed results and scores run through the connector runtime (no vendor yet) |
connector.venue.kalshi | off | Kalshi comparison prices run through the connector runtime |
connector.venue.novig | off | Novig read-only catalog, market and configured private feed observations and the authenticated sidecar view; comparison delivery is not connected |
connector.venue.polymarket | off | Polymarket comparison prices run through the connector runtime |
drift.proposals | off | The DriftStore proposes aliases for quarantined spellings every five minutes, and backfill applies approved aliases from the overlay |
drift.stats | off | Backfill profiles each raw batch, checks it against the seven-day baseline and writes drift findings; a unit shift holds the batch |
identity.events | off | The event review scores retained F402 ticker legs against the latest verified 4C board into the /desk event review queue; nothing links without an operator decision |
identity.resolution | off | The identity pass scores account pairs from the F402 tape into the /security review queue; nothing links without an operator decision |
keys.global | on | The theme key and the shortcut sheet work outside the desk |
keys.pages | on | Number keys jump between pages |
mcp.worker | off | The Worker serves POST /mcp to per-principal bearer tokens: twelve structured canonical tools filtered by scope, pseudonymous ids without desk:identity, a token bucket per principal and an audit row per call |
partners.live | off | Active partners get live Telegram and signed webhook sends, and GET /api/partner/v1/signals answers pulls; stays off until decision D1's gate (provider contract, results feed, counsel) clears |
partners.settlement | off | Partners report wagers and open disputes over /api/partner/v1/wagers and /api/partner/v1/disputes, the settlement pass attributes and settles them, issues statements and posts the ledger, and partner events go onto the daily audit chain; nothing settles until the results feed lands (D15) |
partners.v1 | off | The partner pass plans a delivery intent per partner and channel for each open rated signal its deal entitles, paper-delivers the due ones, and /partners shows partners, deals, intents and latency |
profile.v2 | off | SharpScoreV2 scores accounts and bettors per cell from graded wager features and daily rollups; v1 SharpScore stays the score of record until it is removed |
ratings.v1 | off | Edge and profile signals carry a grade and confidence from their rating cell's history when they first fire, the track pass scores them after their events start, and /health shows live against backtest calibration |
theme.contrast | on | Contrast is in the theme cycle and the switch |
theme.system | on | System is in the theme cycle and the switch |