feat(scoring): upstream time-decay, behind a default-off flag (#703)

[JSONbored]

What

Implements upstream gittensor's sigmoid time-decay in gittensory's score model (#703), default-off and owner-gated. The roadmap defers #703 "gated on owner review of a before/after diff" — so this changes nothing live until you set SCORING_TIME_DECAY_ENABLED.

Faithfulness

Ports the validator's calculate_time_decay (gittensor/validator/utils/datetime_utils.py) verbatim:

• Hard 12h grace → multiplier exactly 1.0.
• Then a logistic on days-since-merge: 1 / (1 + exp(0.4·(days − 10))) → 50% at 10 days.
• Floored at the 0.05 minimum.

The 4 constants are now modeled in DEFAULT_SCORING_CONSTANTS (so they stop surfacing as unmodeled drift via #690).

Before / after (what enabling would do — for your review)

PR age	multiplier
fresh (< 12h)	1.00
5 days	~0.88
10 days	0.50
20 days	0.05 (floor)
30 days	0.05 (floor)

A fresh PR is unaffected even when enabled, so a normal new-PR preview never changes — only an aged-PR projection decays.

Architecture note (why per-PR projection, not a historical-aggregate refactor)

I'd initially scoped "per-PR merge-age data spine." On investigation, gittensory's preview scores a single fresh planned PR and surfaces the contributor's upstream-API standing (already decayed upstream) — there's no raw historical per-PR aggregation in gittensory to decay, so re-applying there would double-count. The faithful, double-decay-safe model is per-PR: predicted score × calculateTimeDecay(prAgeHours), with prAgeHours an optional preview input and applyTimeDecay injected server-side from the env flag (not caller-controllable).

How to enable (after you review)

Set SCORING_TIME_DECAY_ENABLED=true. Then a score preview that supplies prAgeHours decays per the curve above.

Tests

calculateTimeDecay vs upstream (grace / 50%@10D / floor / monotonic), constants-now-modeled, the env gate, preview applied-vs-unapplied paths, and a before/after trajectory. Typecheck + UI lint + OpenAPI (regenerated) clean; 97% coverage held.

[ghost]

Note

Gittensory Gate skipped

PR closed before full evaluation. No late first comment was created.

Signal	Result	Evidence	Action
Gate result	⚠️ Skipped	#731 is no longer open.	No action.

---

💰 Earn for open-source contributions like this. Gittensor lets GitHub contributors earn for the work they already do — register to start earning →.

Checked by Gittensory, a quiet PR intelligence layer for OSS maintainers.

[superagent-security]

Superagent didn't find any vulnerabilities or security issues in this PR.

[ghost]

gittensory · advisory review

Reviewed 8 changed file(s) — two independent AI reviewers.

Suggested action: 🔍 Manual review. The changed route's preview could not be captured — review the rendered page manually before merging.

Reviewer A · gpt-oss-120b — recommends ✅ merge
This PR introduces an upstream time‑decay multiplier to scoring previews, adds a default‑off env‑gated flag, updates schemas, constants, and tests. The implementation is clean, type‑safe, and includes comprehensive unit tests.

Suggestions

• Add a brief description for the new timeDecayMultiplier property in the OpenAPI spec for better documentation.
• Consider adding an integration test for the API endpoint to verify the field appears in the JSON response when the flag is enabled.
• Document the new SCORING_TIME_DECAY_ENABLED env var in the project's README or configuration guide.

Worth double-checking

• Consumers of the scoring preview API must handle the additional timeDecayMultiplier field; ensure downstream code tolerates extra properties.
• If the env flag is mistakenly enabled in production, aged PR scores will be reduced – verify operational monitoring reflects this change.

Reviewer B · nemotron-3-120b-a12b — recommends ✅ merge
The PR adds an optional upstream time-decay feature to score previews, controlled by a default-off environment variable (SCORING_TIME_DECAY_ENABLED). It introduces prAgeHours and applyTimeDecay inputs, a timeDecayMultiplier in results, and implements the decay logic via a sigmoid function with grace period. Changes are confined to scoring logic, API routes, MCP server, OpenAPI schemas, and tests. The feature is owner-gated and off by default, preserving the public/private boundary as it does not introduce forbidden terms in public output.

Suggestions

• Consider adding a JSDoc comment to calculateTimeDecay explaining the sigmoid formula's parameters for future maintainers.
• Verify that the OpenAPI spec updates (apps/gittensory-ui/public/openapi.json) are complete and consistent with the schema definitions in src/openapi/schemas.ts.
• Ensure the environment variable documentation (in src/env.d.ts) note that the value is case-insensitive (already handled by regex).

Worth double-checking

• When enabled, time decay will reduce estimated scores for aged PRs; owners must review the before/after impact as noted in the comment.
• The calculateTimeDecay function uses constants from DEFAULT_SCORING_CONSTANTS; ensure these constants are not overridden unexpectedly in specific environments.
• The OpenAPI spec updates are manual and error-prone; consider automating spec generation from schemas to avoid drift.

Warning

Could not capture a preview of the changed route(s) — review the rendered page manually before merging.

Before = production · After = this PR's preview deploy.

Route	Viewport	Before	After
/	desktop
	mobile

[cloudflare-workers-and-pages]

Deploying with    Cloudflare Workers

The latest updates on your project. Learn more about integrating Git with Workers.

Status	Name	Latest Commit	Preview URL	Updated (UTC)
✅ Deployment successful! View logs	gittensory-ui	52dbcd9	Commit Preview URL Branch Preview URL	Jun 14 2026, 04:55 PM