Ranking parameters¶

Query and body fields that control server/search.ts ranking and related API handlers.

`SearchParams` (compare / for-you)¶

Field	Type	Default	Description
`q`	string	`""`	Search query; empty on For You
`mode`	`platform_style` \| `refined`	(per request)	Which score profile
`recencyWeight`	number	`0.78`	Scales recency term (0 = off)
`strictMatch`	boolean	`true`	Enables spam/hashtag penalties
`maxResults`	number	Route-specific	Cap before `rankTweets` (not inside `scoreTweet`)
`omitRetweets`	boolean	`false`	Live search only: append `-is:retweet`

Server defaults (`server/index.ts`)¶

Route	`maxResults` default	Clamp
`GET /api/compare`	30	10–100
`GET /api/for-you` (live/mock)	35	5–100
`GET /api/for-you?xai=1`	12	1–25

Compare endpoint¶

GET /api/compare

Param	Notes
`q`	Required for meaningful Explore results
`source`	`mock` \| `live`
`recencyWeight`, `strictMatch`, `maxResults`, `omitRetweets`	Corpus cap + ranker inputs

Returns platform_style and refined arrays (field names match ComparePayload).

For You endpoint¶

GET /api/for-you

Param	Notes
`home`	`live` \| `xai` \| `seed` (client `homeFeedMode`)
Same slider fields	From Ranking lab state

Paths:

live — X home timeline (session required); query xai omitted
xai — xai=1; Grok batch + optional xaiIngest → Grok ingests
seed — mock=1 / seed path — server/seed-tweets.json

Scoring reference¶

See Ranking system for overlap, phrase, recency, engagement, spam.

Pipetrix snapshot¶

On ingest, UI may pass rankingLab: { recencyWeight, strictMatch, maxResults } into canonical writes → stored in mixerMetaJson. See Ranking lab snapshot.