OmniRoute/docs/i18n/ha/docs/compression/COMPRESSION_GUIDE.md

🗜️ Prompt Compression Guide — OmniRoute (Hausa)

🌐 Languages: 🇺🇸 English · 🇪🇹 am · 🇸🇦 ar · 🇦🇿 az · 🇧🇬 bg · 🇧🇩 bn · 🇨🇿 cs · 🇩🇰 da · 🇩🇪 de · 🇬🇷 el · 🇪🇸 es · 🇪🇪 et · 🇮🇷 fa · 🇫🇮 fi · 🇫🇷 fr · 🇮🇪 ga · 🇮🇳 gu · 🇮🇱 he · 🇮🇳 hi · 🇭🇷 hr · 🇭🇺 hu · 🇦🇲 hy · 🇮🇩 id · 🇳🇬 ig · 🇮🇹 it · 🇯🇵 ja · 🇬🇪 ka · 🇰🇭 km · 🇮🇳 kn · 🇰🇷 ko · 🇱🇹 lt · 🇱🇻 lv · 🇮🇳 ml · 🇮🇳 mr · 🇲🇾 ms · 🇲🇹 mt · 🇲🇲 my · 🇳🇵 ne · 🇳🇱 nl · 🇳🇴 no · 🇮🇳 or · 🇮🇳 pa · 🇵🇭 phi · 🇵🇱 pl · 🇵🇹 pt · 🇧🇷 pt-BR · 🇷🇴 ro · 🇷🇺 ru · 🇱🇰 si · 🇸🇰 sk · 🇸🇮 sl · 🇷🇸 sr · 🇸🇪 sv · 🇰🇪 sw · 🇮🇳 ta · 🇮🇳 te · 🇹🇭 th · 🇹🇷 tr · 🇺🇦 uk-UA · 🇵🇰 ur · 🇺🇿 uz · 🇻🇳 vi · 🇳🇬 yo · 🇨🇳 zh-CN · 🇹🇼 zh-TW

---

Ajiye 15-95% na mahallin da ya cancanta ta atomatik. Don taƙaitaccen bayani, duba sashen Matsawar README.

Bayani Gabaɗaya

OmniRoute yana aiwatar da tsarin matakai na matsa prompt mai sassa daban-daban wanda ke gudana tun da wuri kafin buƙatu su isa ga masu samarwa na upstream. Wannan yana nufin ana adana token ɗinka a ɓoye — ba sai ka canza tsarin aikinka ba.

Buƙatar Abokin Ciniki
  → Mai Zaɓar Dabarar Matsawa
    → Akwai saitin combo na musamman? → Yi amfani da saitin combo
    → An kai ma'aunin kunnawa ta atomatik? → Yi amfani da yanayin auto
    → Yanayin tsoho? → Yi amfani da saitin gama-gari
    → Off? → Tsallake matsawa
  → Yanayin Matsawa da Aka Zaɓa
    → Off: Babu matsawa
    → Lite: Amintaccen tsaftace fararen sarari/tsarawa (~15%)
    → Standard: Cire kalmomin cike na salon Caveman (~30%)
    → Aggressive: Tsufar tarihi + taƙaitawa (~50%)
    → Ultra: Ragewa bisa ƙa'ida + siranta tubalan lamba (~75%)
    → RTK: Tace fitowar terminal/kayan aiki bisa umarni (kewayon upstream na 60-90%)
    → Stacked: Tsarin injuna da yawa bisa jeri, galibi RTK sannan Caveman (kewayon da ya cancanta na 78-95%)
  → Buƙatar da Aka Matse → Mai Samarwa

---

Yanayin Matsawa

Off

Ba a amfani da matsawa. Duk saƙonni suna wucewa ba tare da canji ba.

Yanayin Lite (~15% na tanadi, jinkiri <1ms)

Yanayi mafi aminci — babu wani canjin ma'ana, tsaftace tsarawa kawai:

Dabara	Bayani
collapseWhitespace	Haɗa layukan fanko masu jere da sararin ƙarshen layi
dedupSystemPrompt	Cire saƙonnin tsarin da aka maimaita
compressToolResults	Matse dogayen fitowar kayan aiki/function
removeRedundantContent	Cire umarnin da aka maimaita
replaceImageUrls	Gajarta URI na bayanan hoton base64

Ya fi dacewa da: Amfani na dindindin, tsarin aiki masu matuƙar buƙatar aminci.

Yanayin Standard (~30% na tanadi)

An samo ra'ayinsa daga Caveman — yana cire kalmomin cike da dogon salon magana tare da kiyaye ma'ana:

• Yana cire kalmomin cike ("don Allah", "ina ganin", "a zahiri", "haƙiƙa")
• Yana taƙaita dogayen jimloli ("domin a yi" → "don", "sakamakon" → "saboda")
• Yana cire maganganun ladabi masu sassauta buƙata ("Za ka damu idan...", "Idan za ka iya...")
• Dokokin regex sama da 30 da aka daidaita don prompt na rubuta lamba

Ya fi dacewa da: Ayyukan rubuta lamba na yau da kullum, ƙungiyoyin da ke kula da tsada.

Yanayin Aggressive (~50% na tanadi)

Sarrafawar tarihi mai wayo don dogayen zaman aiki:

• Tsufar Saƙonni — ana ƙara matsa tsofaffin saƙonni a hankali
• Taƙaita Sakamakon Kayan Aiki — ana maye gurbin dogayen fitowar kayan aiki da taƙaitawa
• Masu Kiyaye Ingancin Tsari — suna tabbatar da cewa nau'ikan tool_use + tool_result sun kasance daidai
• Sanin Girman Tagar Mahalli — yana mutunta iyakokin token na kowane model

Ya fi dacewa da: Dogayen zaman gyara kurakurai, manyan rumbunan lamba.

Yanayin Ultra (~75% na tanadi)

Matsawa mafi ƙarfi don yanayin da token yake da matuƙar muhimmanci:

• Ragewa Bisa Ƙa'ida — yana cire saƙonnin da ke ƙasa da ma'aunin dacewa
• Siranta Tubalan Lamba — yana matse misalan lamba masu maimaituwa
• Yankewa ta Binciken Binary — yana gano mafi kyawun wurin yankewa don tagar mahalli
• An haɗa duk fasalolin yanayin Aggressive

Ya fi dacewa da: Lokacin da kake kaiwa iyakar mahalli akai-akai.

Yanayin RTK (kewayon upstream na 60-90%)

An inganta yanayin RTK don dogayen fitowar kayan aiki da ke bayyana a zaman coding-agent:

• Yana gano nau'ikan umarni/fitowa kamar git status, git diff, git log, masu gudanar da gwaji, build na TypeScript/Vite/Webpack, ESLint/Biome/Prettier, npm audit/installs, log na Docker, fitowar infra, da fitowar shell ta gama-gari
• Yana amfani da tarin matatun JSON daga open-sse/services/compression/engines/rtk/filters/
• Yana shigo da matatun RTK TOML schema v1 daga fayilolin filters.toml na project ko na gama-gari, tare da tabbatarwa ta inline-test da takaita amincewa ga fayilolin project
• Ya ƙunshi matatu 49 na ciki tare da samfuran tabbatarwa na inline
• Yana cire jerin sarrafa ANSI, sandunan ci gaba, layukan da aka maimaita, da hayaniyar da ba za a iya ɗaukar mataki a kanta ba
• Yana kiyaye gazawa, kurakurai, gargaɗi, fayilolin da aka canza, taƙaitawa, da ƙarshen doguwar fitowa
• Yana goyon bayan matatun project masu takaita amincewa, matatun gama-gari, da dawo da ɗanyen fitarwa da aka ɓoye bayanansa idan ana so

Ya fi dacewa da: Zaman agent masu ɗauke da bayanan shell, build, test, git, grep, da fitowar fayil.

Yanayin Stacked (kewayon da ya cancanta na 78-95%)

Yanayin Stacked yana gudanar da injunan matsawa da yawa bisa tabbataccen jeri. Tsarin tsoho shi ne:

RTK -> Caveman

Wannan jeri yana fara taƙaita fitowar terminal/kayan aiki, sannan ya yi amfani da taƙaita ma'ana ta Caveman ga sauran prompt na harshe na yau da kullum. Ana iya saita tsarin Stacked a matakin gama-gari ko ta hanyar combo na matsawa da aka keɓe wa combo na routing.

Ya fi dacewa da: Cakudadden mahalli mai manyan log na kayan aiki tare da umarnin ɗan Adam ko taƙaitawar assistant.

---

Lissafin Tanadi Daga Asalin Aiki

OmniRoute yana bayyana tanadin matsawa daga tushe biyu: gwaje-gwajen aikin asalin ayyuka da kuma haɗa injinan OmniRoute da kansa.

Tushe	Adadin README na asalin aiki da aka yi amfani da shi a nan
Caveman	~75% ƙarancin token ɗin fitarwa, matsakaicin tanadin fitarwa na gwajin aiki 65%, kewayon 22-87%, da kayan aikin matsawar shigarwa na ~46%
RTK	Tanadin fitarwar umarni na 60-90%; zaman misali mai token ~118,000 -> ~23,900, ko tanadin 79.7% (~80%)

Ga bayanan kayan aiki/mahalli masu ruɓanya juna, haɗin OmniRoute na asali yana jera injinan kamar haka:

RTK -> Caveman

Haɗaɗɗen tanadin yana ninkuwa ne, ba tarawa ba:

combined = 1 - (1 - RTK savings) * (1 - Caveman input savings)
average  = 1 - (1 - 0.80) * (1 - 0.46) = 89.2%
range    = 1 - (1 - 0.60..0.90) * (1 - 0.46) = 78.4-94.6%

Wannan adadin 78-95% yana aiki ne idan RTK da Caveman duka za su iya rage bayanan shigarwa/mahalli iri ɗaya. Yanayin fitarwar amsar Caveman daban yake: idan an kunna shi, yi amfani da tanadin fitarwa na Caveman (65% a matsakaici, babban adadin ~75%, kewayon 22-87%). Jimillar tanadin kuɗin caji ya dogara da haɗin prompt/fitarwarku.

Abin da "cancanta" yake nufi a zahiri

Kewayon da aka fi nunawa na 15-95% gaskiya ne, amma yana aiki ne kawai ga abun ciki mai maimaituwa ko yawan surutu — layukan kuskure da aka maimaita, log ɗin ginawa da ke yawaita gargaɗi iri ɗaya, ko tarin sakamakon grep/karanta-fayil mai girma fiye da kima. Wannan ba yana nufin kowace buƙata za ta tanadi wannan adadin ba.

An tabbatar a aikace (tests/unit/compression/stacked-compression-tool-result-savings.test.ts): gudanarwar stacked (RTK + Caveman) a kan tubalin tool_result mai tsarin Anthropic wanda ya ƙunshi layukan kuskure iri ɗaya guda 300 ya samar da tanadin token na 95.93% / tanadin haruffa na 96.26% — daidai cikin kewayon da aka tallata. Amma gudanar da wannan jerin aikin a kan fitarwar kayan aiki ta al'ada wadda ba ta maimaituwa (jerin sakamakon grep mai tsabta, ɗan gajeren karatun fayil, rubutun hira na yau da kullum) yana samar da tanadi kusa da sifili yadda ya dace, saboda babu wani abu mai maimaituwa da za a cire, kuma validateCompression() (validation.ts) yana ƙin aika sake-rubutawar da za ta cire ko sauya tubalan lamba, URLs, kanun labarai, nau'ukan siga, ko masu gano constants masu ALL-CAPS.

Wannan hali ne da ake tsammani kuma mai aminci, ba bug ba ne: zaman coding wanda galibi yake karanta/grep na fayiloli masu tsabta zai ga ƙaramin jimillar tanadi ko da an kunna matsawa gaba ɗaya, yayin da zaman da ya haɗu da madauwari mai gazawa ko linter mai yawan fitar da bayanai zai ga cikakken kewayon 78-95% a kan wannan zirga-zirgar. Kada ku yi amfani da ƙaramin kason jimillar tanadi na zama guda a matsayin hujjar cewa an saita matsawa ba daidai ba — da farko ku bincika ko ainihin fitarwar kayan aikin tana maimaituwa.

---

Nuna Tanadin Token

Ba tare da matsawa ba: An aika token 47K zuwa LLM
Tare da Lite:          An aika token 40K          (an tanadi 15% — amintacce, koyaushe a kunne)
Tare da Standard:      An aika token 33K          (an tanadi 30% — ƙa'idodin caveman-speak)
Tare da Aggressive:    An aika token 24K          (an tanadi 50% — tsufar bayanai + taƙaitawa)
Tare da Ultra:         An aika token 12K          (an tanadi 75% — datsewa ta heuristic)
Tare da RTK:           An aika token 19K-5K       (an tanadi 60-90% a fitarwar umarni/kayan aiki)
Tare da Stacked:       An aika token 10K-2.5K     (kewayon RTK+Caveman na 78-95% ga bayanan da suka cancanta)

---

Kanfigareshan

Dashboard

Je zuwa Dashboard → Context & Cache:

• Caveman — zaɓin yanayi, kunshin harsuna, samfoti, da tsoffin saitunan gaba ɗaya
• RTK — samfotin tace-umarni, saitunan tsaron RTK, da kundin matatu
• Compression Combos — jerin injina masu suna waɗanda aka sanya wa haɗaɗɗun hanyoyin turawa
• Auto-Trigger Threshold — kunna matsawa ta atomatik idan adadin token ya wuce iyakar da aka saita

Sauya Saiti Ga Kowane Combo

A cikin Dashboard → Context & Cache → Compression Combos, sanya compression combo ga routing combo:

Combo: "free-tier-fallback"
  Compression Combo: "coding-agent-stack"
  Pipeline: RTK -> Caveman
  Targets:
    1. if/kimi-k2.7-code
    2. if/qwen3.8-max-preview

Wannan yana ba ka damar amfani da matsawa mai matakai a kan masu samar da sabis na kyauta/na coding, yayin da ake riƙe lite mode a kan biyan kuɗi na kuɗi.

Wannan saitin "Per-Combo Override" wani iko ne daban da sauya routing-combo compression mode (Default/Off/Lite/Standard/Aggressive/Ultra) — wannan sauyin ba ya zaɓar jerin compression-combo mai suna; kawai yana saita filin compressionMode da resolveCompressionPlan ke dubawa. Ana iya saita shi ko dai a kan combo card (Dashboard → Combos) ko kuma, tun daga #6760, ga kowane routing combo a cikin jerin "Assign to routing" da ke Dashboard → Context & Cache → Compression Combos, kusa da akwatin zaɓin sanya pipeline da aka bayyana a sama. Duk wuraren biyu suna adana saitin ta endpoint ɗin PUT /api/combos/{id} iri ɗaya.

Sauya saitin kowace buƙata

Aika request header na x-omniroute-compression don sauya tsarin matsawa na buƙata guda ɗaya. Shi ne mafi fifiko — yana rinjayar sauyin routing-combo, active profile, auto-trigger, da Default na panel. Ana yin watsi da ƙimomin da ba a sani ba (ba a taɓa ƙin karɓar buƙatar ba), kuma global master switch har yanzu shi ne ke sarrafa komai: idan an kashe matsawa gaba ɗaya, header ba zai iya kunna ta ba. Ƙimomi:

Ƙima	Tasiri
off	Babu matsawa ga wannan buƙatar.
default	Default profile da aka samo daga panel (yana yin watsi da active profile).
engine:<id>	Inji guda ɗaya idan an kunna shi, misali engine:rtk.
<combo>	Combo mai suna, ana fara daidaita shi ta suna (ba tare da la’akari da girman haruffa ba), sannan ta id.

Ana mayar da tsarin da aka yi amfani da shi a cikin response header na X-OmniRoute-Compression: <mode>; source=<source>, inda <source> yake ɗaya daga cikin request-header, routing-override, active-profile, auto-trigger, default, ko off.

API

# Samo saitunan matsawa
curl http://localhost:20128/api/settings/compression

# Sabunta saitunan matsawa
curl -X PUT http://localhost:20128/api/settings/compression \
  -H "Content-Type: application/json" \
  -d '{"defaultMode":"stacked","autoTriggerMode":"stacked","autoTriggerTokens":32000}'

# Yi samfotin takamaiman payload na RTK/stacked
curl -X POST http://localhost:20128/api/compression/preview \
  -H "Content-Type: application/json" \
  -d '{"mode":"rtk","messages":[{"role":"tool","content":"npm test output here"}]}'

# Jera kunshin matatun RTK
curl http://localhost:20128/api/context/rtk/filters

# Gwada RTK kai tsaye tare da metadata na umarni na zaɓi
curl -X POST http://localhost:20128/api/context/rtk/test \
  -H "Content-Type: application/json" \
  -d '{"command":"npm test","text":"FAIL tests/example.test.ts\nError: boom"}'

---

Abubuwan da Ake Karewa

Injin matsawa koyaushe yana kiyaye:

• ✅ Tubalan lamba (masu shinge da na cikin layi)
• ✅ URLs da hanyoyin fayil
• ✅ Tsarukan JSON da bayanai masu tsari
• ✅ Masu ganewa da keɓaɓɓun alamomin fasaha
• ✅ Bayanan lissafi
• ✅ Ma’anonin kiran kayan aiki/ayyuka
• ✅ Umarnin tsarin (a yanayin lite)

Maido da ɗanyen fitarwa na RTK yana ɓoye sanannun maɓallan API, bearer tokens, Slack tokens, AWS access keys, kalmomin sirri, tokens, da sirrika kafin a adana wani abu.

---

Ƙididdigar Matsawa

Kowace buƙata da aka matsa tana ƙunshe da ƙididdiga a cikin rajistan uwar garke:

{
  "originalTokens": 47200,
  "compressedTokens": 40120,
  "savingsPercent": 15.0,
  "techniquesUsed": ["collapseWhitespace", "dedupSystemPrompt"],
  "mode": "lite",
  "engine": "caveman",
  "compressionComboId": "coding-agent-stack",
  "durationMs": 0.8,
  "rtkRawOutputPointers": []
}

---

Taswirar Matakai

Mataki	Yanayoyi	Matsayi
Mataki na 1	Kashe, Lite	✅ An fitar
Mataki na 2	Standard, Aggressive, Ultra	✅ An fitar
Mataki na 3	RTK, Stacked, Haɗaɗɗun Matsawa	✅ An fitar
Mataki na 4	Salon Fitarwa, Ultra na matakin SLM, tsarin gwaji	✅ An fitar
Mataki na 4C	Kasafin mahallin da ke daidaita kansa ("dial") — injin lissafi + API (contextBudget a kan PUT /api/settings/compression) + sarrafa yanayi/manufofi a allon kulawa	✅ An fitar

---

Godiya

Dokokin matsawa na yanayin Standard sun samo asali ne daga Caveman na JuliusBrussee (⭐ 51K+) — shahararren aikin "me ya sa za a yi amfani da tokens masu yawa alhali kaɗan za su yi aikin". Caveman ya bayar da rahoton ƙarancin tokens na fitarwa da ~75%, matsakaicin tanadin fitarwa na gwajin ƙima da 65%, kewayon fitarwa na 22-87%, da kayan aikin matsa bayanan shigarwa na ~46%.

Yanayin RTK ya samo asali ne daga RTK - Rust Token Killer na RTK AI — aikin matsa fitarwar umarni mai inganci sosai don terminal, build, test, git, da tace fitarwar kayan aiki. RTK ya bayar da rahoton tanadi na 60-90%, inda zaman samfurin README ɗinsa ya nuna an tanadi ~80%.

---

Tsarukan Matsawa na Ci-gaba

Baya ga daidaitattun yanayi 7, OmniRoute ya haɗa da wasu tsarukan matsawa na ci-gaba waɗanda ke aiki ta atomatik bisa ga mahalli.

Matsawa Mai La’akari da Cache

Wasu masu samarwa (kamar Anthropic mai amfani da prompt caching) suna goyon bayan prompt caching, wanda ke ba su damar adana wasu sassan prompt a cache domin rage kuɗi da jinkiri. Idan an kunna caching, matsawa mai tsanani na iya cutar da aiki saboda yana canza tokens da aka adana a cache, wanda ke sa cache ɗin ya daina aiki.

Module ɗin cachingAware.ts yana magance wannan ta hanyar gano mahallin caching da daidaita dabarar matsawa yadda ya dace.

Yadda yake aiki

1. Gano mahallin caching — Yana bincika jikin buƙata don alamomin cache_control
2. Gano masu samarwa masu caching — Yana bincika ko mai samarwa da aka nufa yana goyon bayan caching
3. Daidaita dabara — Yana saukar da aggressive/ultra zuwa standard ga masu samarwa masu caching
4. Tsallake system prompt — Galibi ana adana system prompts a cache, don haka kada a matsa su
5. Yi amfani da sauye-sauye masu tabbataccen sakamako — Yi amfani kawai da sauye-sauyen da ke samar da sakamako iri ɗaya

Misalin lamba

import {
  detectCachingContext,
  getCacheAwareStrategy,
} from "@omniroute/open-sse/services/compression/cachingAware";

const body = {
  model: "anthropic/claude-sonnet-4.5",
  messages: [{ role: "user", content: "Hello" }],
  cache_control: { type: "ephemeral" }, // ← Alamar cache
};

const ctx = detectCachingContext(body, { provider: "anthropic" });
// → { hasCacheControl: true, provider: "anthropic", isCachingProvider: true }

const strategy = getCacheAwareStrategy("aggressive", ctx);
// → { strategy: "standard", skipSystemPrompt: true, deterministicOnly: true }

Lokacin amfani

Matsawa mai la’akari da cache koyaushe a kunne yake — ba a buƙatar saiti. Yana fara aiki ne kawai idan:

• Buƙatar tana da alamomin cache_control
• Mai samarwar da aka nufa yana goyon bayan prompt caching (Anthropic, OpenAI, da sauransu.)

Tsufa a Hankali

Dogayen tattaunawa suna tara zagayen saƙonni masu yawa, amma tsofaffin zagayen suna raguwa a muhimmanci. Module ɗin progressiveAging.ts yana rage cikakken bayanin saƙonni bisa nisan zagaye:

• Zagayen baya-bayan nan (0-3): Ana kiyaye su yadda suke (cikakken bayani)
• Matsakaicin zagaye (4-8): Matsawar Lite (gyaran farin sarari da tsarawa)
• Tsofaffin zagaye (9+): Matsawar Caveman (cire kalmomin cikawa da taƙaitawa)
• Zagayen da suka tsufa sosai (20+): Ana taƙaita su sosai ko a cire su

Misalin lamba

import { applyAging } from "@omniroute/open-sse/services/compression/progressiveAging";

const messages = [
  { role: "system", content: "You are a helpful assistant" },
  { role: "user", content: "What is 2+2?" },
  { role: "assistant", content: "4" },
  // ... ƙarin zagaye 50 ...
];

const { messages: aged, saved } = applyAging(messages, {
  verbatim: 3, // Zagaye 3 na farko: yadda suke
  light: 8, // Zagaye na 4-8: matsawar lite
  moderate: 20, // Zagaye na 9-20: matsawar caveman
  // Zagaye na 21+: taƙaitawa mai yawa
});

// saved = adadin tokens da aka tanada

Lokacin amfani

Tsufa mai ci gaba yana kunnawa koyaushe a yanayin aggressive da ultra. Yana da tasiri musamman ga:

• Zaman rubuta lamba na dogon lokaci
• Tattaunawa masu ɗaukar kwanaki da yawa
• Tsarukan aiki na wakilai masu kiran kayan aiki da yawa

Yanayin Fitarwa na Caveman

Modul ɗin outputMode.ts yana saka umarnin saƙon tsarin domin sa samfurin da kansa ya samar da gajartacciyar fitarwa mai taƙaitaccen salo (salon "caveman").

Yadda yake aiki

Maimakon matsa bayanan shigarwa, wannan yanayin yana ƙara saƙon tsarin kamar:

"Ka ba da amsa da kalmomi kaɗan. Ka tsallake maganganun ladabi. Yi amfani da gajerun jimloli."

Wannan yana aiki sosai musamman ga:

• Samar da lamba (gajeriyar fitarwa = ƙarancin tokens)
• Tambaya da amsa cikin sauri (ba a buƙatar dogayen bayanai)
• Sarrafa bayanai rukuni-rukuni (ƙara yawan abin da ake sarrafawa)

Lokacin amfani

Yanayin fitarwa na Caveman na zaɓi ne — saita shi ta hanyar combo config:

{
  "strategy": "auto",
  "config": {
    "auto": {
      "outputMode": "caveman"
    }
  }
}

Salon Fitarwa (kasida)

Yanayin fitarwa na Caveman da ke sama shi ne tsohuwar hanya mai salo guda ɗaya. Phase 4 ya faɗaɗa shi zuwa kasidar salon fitarwa da za a iya haɗawa: OUTPUT_STYLE_CATALOG a cikin open-sse/services/compression/outputStyles/catalog.ts. Kowane salo umarnin saƙon tsarin ne wanda ke sa samfurin da kansa ya samar da fitarwa mai rahusa; ana iya kunna salo tare kuma ana saka su bisa jerin kasidar.

Salo	id	Abin da yake yi	Harsunan umarni
Taƙaitaccen rubutu	terse-prose	Cire kalmomin cike giɓi/articles/kalamai marasa tabbaci; kiyaye ainihin ma’anar fasaha. Rubutu ɗaya da tsohon yanayin fitarwa na caveman (an yi masa nuni, ba a sake rubuta shi ba).	en, pt-BR, es, de, fr, it, ru, zh, ja, id, vi
Ƙarancin lamba	less-code	Matakan YAGNI: mafi ƙarancin canjin da ke aiki, ba tare da abstractions da ba a nema ba.	en, pt-BR, es, de, fr, it, ru, zh, ja, id, vi
Ponytail (malalacin babban mai haɓakawa)	ponytail	"Mafi kyawun lamba ita ce lambar da ba a taɓa rubutawa ba": sake amfani > sake rubutawa, tushen matsala > alama, mafi gajeren diff mai aiki.	en, pt-BR, es, de, fr, it, ru, zh, ja, id, vi
Ina da ADHD (aiki da farko)	i-have-adhd	Aiki da farko (umarni/path/snippet kafin bayani), matakai masu lamba da iyaka, takamaiman mataki na gaba GUDA ƊAYA, babu gabatarwa/taƙaitawa/kalmomin rufewa. An daidaita daga ayghri/i-have-adhd (MIT).	en, pt-BR, es, de, fr, it, ru, zh, ja, id, vi
Taƙaitaccen CJK (文言)	terse-cjk	Salon Sinanci na gargajiya mai matuƙar taƙaitawa.	zh (an iyakance ta locale: ana bayarwa ne kawai idan harshen da aka tantance shi ne zh)

Kowane salo yana zuwa da matakan ƙarfi uku — lite, full, ultra — kuma kowane mataki yana ƙarewa da sashe ɗaya na iyakoki, wanda ke barin tubalan lamba, hanyoyin fayil, umarni, saƙonnin kuskure, URLs da identifiers yadda suke ba tare da canji ba.

Yadda sakawa ke aiki

applyOutputStyles() (open-sse/services/compression/outputStyles/apply.ts) yana daidaita zaɓin da kasidar (ana watsar da ids da ba a sani ba da salon da ba su dace da locale ba, ba tare da zama kuskure ba), yana haɗa zaɓaɓɓun umarnin bisa jerin kasidar, yana ƙara sashen iyakoki sau ɗaya, sannan yana sanya sakamakon a farkon saƙon tsarin bayan alamar idempotency guda ɗaya ([OmniRoute Output Styles]) — sake aiwatar da shi ba ya yin komai. Idan harshen buƙatar da aka gano yana da fassara, ana saka umarnin da aka fassara maimakon Turanci.

Yadda ake kunnawa

A dashboard: Context → Settings → Compression — layi ɗaya ga kowane salo mai maɓallin kunnawa/kashewa da mai zaɓen mataki. Ta hanyar shirye-shirye, compression config yana adana zaɓin kamar haka:

{
  "outputStyles": [
    { "id": "i-have-adhd", "level": "full" },
    { "id": "less-code", "level": "lite" }
  ]
}

Daidaituwa da tsohon tsari: tsohon saitin combo na outputMode: "caveman" har yanzu yana aiki kuma yana map zuwa terse-prose, daidai da tsohuwar sakawa byte-zuwa-byte a kowane tsohon harshe.

Zaɓin harshe: idan languageConfig.enabled yana kunne, autoDetect yana zaɓar harshen sabon saƙon mai amfani (mai gano harshe ɗaya da na injunan shigarwa); kashe autoDetect yana kulle defaultLanguage. A kashe → Turanci.

An kulle matrix na salo × harshe ta tests/unit/compression/output-styles-i18n-matrix.test.ts: ba za a iya fitar da sabon salo ba tare da aƙalla fassarar pt-BR ba (ko keɓantacciyar izini da aka rubuta), kuma salon da ke akwai ba zai iya rasa locale a ɓoye ba. Don ƙara salo, duba EXTENDING_COMPRESSION.md.

Matsa Sakamakon Kayan Aiki

Modul ɗin toolResultCompressor.ts yana samar da dabarun matsawa na musamman guda 5 don sakamakon kayan aiki (kiran functions, fitarwar wakilai, sakamakon bincike, da sauransu):

1. Matsa sakamakon bincike — Yana cire sakamako masu maimaituwa, yana riƙe top-N
2. Matsa karatun fayil — Yana datse manyan fayiloli, yana kiyaye headers/imports
3. Matsa aiwatar da lamba — Yana riƙe stdout/stderr masu muhimmanci kawai
4. Matsa tambayar database — Yana iyakance layuka, yana cire metadata mai tsawo
5. Matsa martanin API — Yana cire fields masu null, yana taƙaita arrays

Lokacin amfani

Matsa sakamakon kayan aiki yana kunnawa koyaushe idan akwai kiran kayan aiki. Ba a buƙatar saiti.

Bututun Sarrafawa Mai Jere

Yanayin stacked yana gudanar da injuna da yawa a jere — yawanci RTK da farko (ajin 60-90% a fitarwar kayan aiki), sannan Caveman (ƙarin ajin 30% a sauran rubutun). Wannan yana samar da ajin jimilla na 78-95%.

Yadda yake aiki

Shigarwa (tokens 1000)
  → RTK (matacin da ke fahimtar umarni) → tokens 200
    → Caveman (cire kalmomin cike giɓi) → tokens 140
  → Fitarwa (tokens 140, ajin 86%)

Lokacin amfani

Yi amfani da yanayin stacked don:

• Tsarukan aiki masu yawan amfani da kayan aiki (rubuta lamba ta wakilai, bincike)
• Sarrafa bayanai rukuni-rukuni da ke kula da farashi
• Lokacin da kake buƙatar matsakaicin ajin tokens

Saita ta hanyar combo:

{
  "strategy": "auto",
  "config": {
    "auto": {
      "modePack": "stacked"
    }
  }
}

---

Sauya Saitunan Haɗin Matsewa

Za ka iya sauya yanayin matsewa na gama-gari ga kowane haɗi domin daidaita halayya ga lokutan amfani daban-daban:

{
  "id": "coding-combo",
  "strategy": "priority",
  "config": {
    "auto": {
      "weights": { "taskFit": 0.5 },
      "modePack": "quality-first"
    }
  },
  "compressionOverride": {
    "mode": "aggressive",
    "stackedPipelines": ["rtk", "caveman"],
    "preserveToolDefinitions": true
  }
}

Wannan yana da amfani ga:

• Haɗe-haɗen rubuta lamba: Yi amfani da yanayin aggressive don zaman aiki masu tsawo
• Haɗe-haɗen tambaya da amsa cikin sauri: Yi amfani da yanayin lite don amsoshi masu sauri
• Haɗe-haɗe masu yawan amfani da kayan aiki: Yi amfani da yanayin stacked don samun iyakar tanadi
• Haɗe-haɗen samarwa: Yi amfani da yanayin cache-aware ga masu samarwa masu amfani da caching

---

Duba Kuma

• Saitin Muhalli — Sauye-sauyen muhallin matsewa
• Jagorar Tsarin Gine-gine — Bayanan ciki na bututun sarrafa matsewa
• Jagorar Mai Amfani — Yadda ake fara amfani da matsewa
• Matsawar RTK — Matatun RTK, tsarin amincewa, ƙofar tabbatarwa, dawo da fitarwa marar sarrafawa
• Injinan Matsewa — Caveman, RTK, stacked, APIs, MCP, dashboard
• Tsarin Dokokin Matsewa — Tsarin fakitin dokoki na JSON
• Fakitocin Harshen Matsewa — Dokokin Caveman na musamman ga harsuna