Code Explainer

Paste duk wani code (ko upload a fayil har zuwa 500KB) da kuma samu da bayyananne, layin-by-layin bayani a kan ka zaɓi na 4 matsakaici mataki: Junior dev (ya yi imanin da farko syntax sani), Senior dev (ya yi imanin da yaren da yaren, mayar da hankali a kan gini / zane), Non-technical (ba code damuwa - bayyana a cikin kasuwanci sharudda), ko Learn-the-language (ya bayyana yayin da koyar da harshen idioms).

ChatGPT / Cursor / Copilot Chat duk bayyana code kyau - bambanci ne Free.ai ne mai bincike-da aka dogara a kan (ba shigarwa, ba $10-20/mo rajista) da masu sauraro-mataki picker da karfi daidai rajista. Cursor ($20/mo) da Copilot ($10/mo) ne cikin-IDE; amfani da su lokacin da a cikin gudu. Yi amfani da Free.ai ga ad-hoc bayyananniyar kana so ka raba da wani abokan aiki, paste a cikin PR bayani, ko amfani da shi a matsayin taimakon koyarwa.

Junior dev don shigarwa (ya yi tsammanin sun san loops/aiki amma ba alamu masu zurfi ba). Senior dev don shirya nazarin alamun (ya manta da abubuwan da suka shafi, ya zauna a kan ginin, zaɓuɓɓukan zane, gotchas). Ba-teknĩki ga masu amfani da samfur/PM (ƙoɗi, sakamakon kasuwanci). Koyon-harshe don koyon harshe-da-harshe ("I know Python, explain this Rust" — call out the language-specific patterns).

Idan akwai kurakurai masu bayyana (ba-da-wannan-daya, halaye da ba'a bayyana su ba, matsalolin tsaro) mai bayyanawa zai nuna su cikin sashe na "Matsaloli". Don neman kurakurai na tsarin amfani da /code/debug/ (mai gano kurakurai da aka gina bisa manufa) ko /code/review/ (audit da aka yi da darajar tsanani). Modin bayyana yana mai mai da hankali kan fahimta, ba a kan audit ba.

Scales da alamun rubutu da tsawo da masu sauraro. Senior-dev bayyanar da 50-line aiki: ~200 kalmomi + maɓalli annotations. Non-technical bayyanar: 100-150 kalmomi abstracting away mechanics. Learn-the-language: mafi tsawo saboda shi dakata don koyar da idioms - tsammani 300+ kalmomi ga daidai shigarwa.

A wani bangare — sanya lissafin stack a kusa da tushen. Mafi kyawun kayan aiki: /code/debug/ wanda aka gina don "code + error → root cause + fix" kuma yana samar da amsa mai tsari na sashe huɗu.

Qwen 3 Coder (farashin, diff) yana kula da ~95% na bayanin yadda ya kamata. DeepSeek R1 don lambar da take buƙatar jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin jin dadin

Ya - JSON / YAML / TOML / Dockerfile / Kubernetes YAML / SQL / regex / shell scripts duk aiki. The model auto-detects format da picks the right explanation framing. Ga SQL bayani mai dacewa amfani da / code / sql /, ga regex amfani da / code / regex /.

Three sections: **What it does** (one-paragraph summary at the chosen audience level), **Step by step** (numbered walkthrough, each step ties to a code segment), **Why it works the way it does** (design choices, trade-offs, alternatives). Junior level adds vocabulary callouts; learn-the-language adds idiom comparisons.

Up to 12,000 characters (~180 lines) on the self-hosted path. Premium long-context models handle 200K+ characters — switch when explaining a whole module.

An aika da Ƙoɗa zuwa ga sifar, an yi shi, kuma an bar shi. Ba a adana shi ba, ba a amfani da shi don koyarwar. Zaɓi Qwen 3 Coder don mafi girman sirri — wanda aka yi da kansa a kan GPU na Free.ai, tushen ba zai taba barin gininmu ba.

Na'am — POST zuwa /v1/chat/ tare da bayyanar da tsarin kira (mataki na masu sauraro) da kuma alamun shafi kamar sakon mai amfani. Ka duba /api/ don snippets.