Usuku olunye, inguqulo yesinye se-AI ibonise i-time line. Lesi xesha, i-ReAct (i-nope, akuyona i-JavaScript owaziwa futhi uyakuthanda). Thina ushiye mayelana ne-ReAct isampula esizayo esizayo ehlabathini we-AI agents. Ukuphendula + Ukuphendula Okokuqala yasungulwa ngo-2022 (eyenziwe ngempumelelo eminyakeni ye-AI), isampula ye-ReAct kuyinto ngokushesha emhlabeni wonke - futhi ngezizathu olungcono ... Funda ngokushesha lapho sincoma ukuthi kuyinto, indlela yokusebenza, futhi indlela yokusebenza ku-agent workflow yakho. I-AI wave? Nah. It's time to Re-Act! Yini i-ReAct Design Pattern? Ngaba uyazi, " “Ou mhlawumbe” ” Ugh... enye inguqulo React ngo-2025? Hhayi siye sinxazulululwe kuleli ... imizuzu? Kuyinto React ... kodwa ku-AI manje? Okuningi, Ngingathanda Izakhiwo ze-React! Hlola! We're talking about a different kind of ReAct here! Umhlaba we-AI - okuyinto etholakalayo kusuka ku-"I-Reasoning" + "I-Acting" - iyisisombululo sokucubungula lapho i-LLM ihlanganisa ukucubungula nokuphendula izinhlelo zokusebenza kakhulu noma ukukhiqiza imiphumela enhle futhi enhle. Ukuphendula 👇 Let's break it up nge analogy okusha! 👇 Umzekele ukuthi utshale umbhobho we-AI . Uma ushiye nje "ukwenza i-sandwich", uhlelo lwe-AI ephakeme ingatholela i-LLM izicelo kanye nokuthumela isitifiketi se-static. Ngiya ? Umdlalo olukhulu! Okokuqala, it : "Ukuvela - yintoni i-sandwich? Ngingathanda izakhiwo? Yini i-bread?" Ngemuva kwalokho : ivula i-freezer, ukuthatha ukuthi kuyimfuneko, izikhwama, izikhwama, futhi voilà-BLT ephelele! ReAct-powered agent reasons acts Ngokuvamile, ReAct akufanele kuphela. It Waze Ukusuka. Ukusuka. Ukusuka. izakhiwo, izakhiwo Ukusebenza Ukulungiswa okokuqala le pattern ku-2022 iphepha " “Ukuhlukanisa ku-2025 njengomthombo we-Agentic AI kanye ne-Agentic RAG-based agents. ReAct: Synergizing Ukuphendula nokuphendula ku-Language Models ReAct: Synergizing Ukuphendula nokuphendula ku-Language Models Okunye, kanjani lokhu kungenzeka, futhi indlela le design pattern ngokwenene ukusebenza? Thina ufunde! I-ReAct Origins: Indlela I-2022 I-Paper Yenza I-AI Workflow Revolution Ngemuva kwe-2022, i Imininingwane ezisekelwe kuleli khasi: ReAct: Synergizing Ukuphendula nokuphendula ku-Language Models “I-LLM’s’ abilities for reasoning (isib. Chain-of-thought-prompting) kanye ne-acting (isib. Ukwakhiwa kwe-action plan) zithunyelwe ikakhulukazi njengezinto ezahlukile. [Here, we] ukubuyekeza ukusetshenziswa kwe-LLM’s ukukhiqiza izimpendulo ezimbini kanye nemisebenzi esifundeni ngokuvamile ngokuvamile...” “I-LLM’s’ abilities for reasoning (isib. Chain-of-thought-prompting) kanye ne-acting (isib. Ukwakhiwa kwe-action plan) zithunyelwe ikakhulukazi njengezinto ezahlukile. [Here, we] ukubuyekeza ukusetshenziswa kwe-LLM’s ukukhiqiza izimpendulo ezimbini kanye nemisebenzi esifundeni ngokuvamile ngokuvamile...” Ngokuvamile: + = 💥. Okwangoku, i-LLM yaba ikakhulukazi ama-assistants e-brainy—i-generating text, ukuphendula imibuzo, ukudala ikhodi. Kodwa ke Ngaphansi kwe-2022 (Yep, lapho i-ChatGPT yasungulwa ngoNovemba 30), ama-developer basungula ukuqhuma i-LLM ku-workflows ye-software enhle. Izinto zilungele. Ukuguqulwa Fast forward to today: Ngena ngemvume ku – izinhlelo zokusebenza okuzenzakalelayo, zihlanganisa, ukucubungula, nokwenza izinto. Usuku I-AI Agents Kule New , isampula ye-ReAct - okwenziwe nje isampula se-academic - iyona omunye Ukwakhiwa kwama-I.I.I.Agents eyenziwe ngempumelelo. Ngaphezu kwalokho, i-IBM inikeza i-ReAct njenge-building block ye-agentic RAG workflows: AI “Agent” era Izakhiwo ezivamile Okay, ngakho-ke ReAct iyatholakala kumadlulayo ... kodwa kusetshenziselwa elandelayo. Sishayele DeLorean (88 MPH, baby! ⚡) - Sishayele ekhukhwini ukuze ubone kanjani le pattern kusebenza emzimbeni, futhi indlela yokusebenza. Ukusebenza kwe-React ku-Agentic AI Workflows Thola React njengoba MacGyver of AI Ngaphandle kokufaka nje impendulo efana ne-LLM yakho, izinhlelo ze-ReAct Ngena ngemva Ngemuva kwalokho . It is not magic ✨ – it is when chain-of-thought reasoning met real-world action. Thola ukunakekelwa Ukulungele Ngokuvamile, i-ReAct agent isekelwe a Ukubuyekezwa: Think 🤔 → Act 🛠️ → Observe 🔍 → Repeat 🔁 I-Reasoning (Think 🤔): Ukuqala nge-snap like "I-Plan a weekend trip to NYC." I-Agent ikhiqiza imibuzo: "Ngingathanda izindiza, i-hotel, kanye nenkinga le-attractions." Ukukhetha isinyathelo (Act ️): Ngokusekelwe isinyathelo se-agent, inikeza isixhobo (isibonelo, ngokusebenzisa ukuhlanganiswa kwe-MCP) - bheka, i-API yokufunda izindiza - futhi isebenza. Ukucaciswa (Ukucaciswa ): Izixhobo ivumela idatha (isib. Izinhlelo zokuhamba). Lokhu kusetshenziselwa kwama-agent, okuyinto zihlanganisa isinyathelo esilandelayo sokucaciswa. I-Agent isebenzisa imibuzo ezintsha ukhethe isixhobo esinye (isib. Ukukhangela i-hotel), inikeza idatha ezingaphezu, ukuhlaziywa kwegama yayo - konke ngaphakathi kwe-top-level loop. Loop (Repeat 🔁) Ungathanda ukuthi ukuchithwa kwe- "ukuba akuyona" ikhoyili. Kule ngalinye iteration, umphakeli: Yenza isinyathelo esisha sokucindezela. Khetha imishini elihle yokusebenza. Yenza umsebenzi. Ukuhlobisa imiphumela. Ukubuyekeza ukuba izinga lokuphumula. Ukulinganiswa okuqhubekayo kuze kube luhlobo lokuphendula noma indawo yokufinyelela. Indlela yokwenza ReAct Ngakho-ke, ufuna ukufaka i-ReAct ekusebenzeni ne-agents ye-real-world? Ngiya ku-common setup! Ukukhishwa kwe-show nge-A (Ukuza) I-agent ye-top-level, eyenziwe nge-LLM yakho ye-choice, inikeza isicelo sokuqala ku-Dedicated . Orchestrator Agent think CrewAI noma isakhiwo esifanayo I-Agent ye-Reasoning Waze Ngaphandle kokuphuma, I-Prompt ye-Original ku-liste esifanele ye-steps noma i-sub-tasks eyenziwe ngempumelelo. It is the brain, meticulously planning the strategy. Reasoning Agent Ukuhlukanisa Ngemuva kwalokho, lezi izicelo zithunyelwe ku-a Kuyinto lapho ingxubevange isitimela isitimela! Le agent kuyinto tool-wielder yakho, ehlanganisiwe ngqo nge-MCP server (ukwazi ukufinyelela idatha ezingaphandle noma izixhobo ezifana web scrapers noma databases) noma ukuxhumana . It's tasked with actually izindlela ezidingekayo. Acting Agent nezinye ama-agents ezijwayelekile ngokusebenzisa i-A2A protocols Ukusebenza Iziphumo zezi zokusebenza zangaphandle. Zihlanganisa ku- I-agent yenza ukubuyekeza imiphumela, ukunqoba ukuthi umsebenzi iyatholakala futhi ephelele, noma uma izinyathelo ezininzi zihlanganisa. Uma izinyathelo ezilandelayo zihlanganisa, isikhwama uqala ngokushesha, ukunikela ama-agent ngokushesha inqubo. Observing Agent Okuqhubekayo umzila wahlala kuze kube Ukubonisa umphumela "yakhelwe", ukuxhumana okuhlolwa okuqhubekayo kuze ku Ngiyaxolisa ku-Inquirer. Reasoning -> Acting -> Observing umphathi we-Agent Umphathi we-Orchestrator Njengoba ungakwazi ukubona, indlela elula ukufaka i-ReAct kuyinto nge-multi-agent setup! Nokho, Thola nje nge-single, elula, mini agent, futhi. Just check out the example in the video ngezansi: Ukulungele https://www.youtube.com/watch?v=pEMhPBQMNjg&embedable=true ReAct vs “regular” AI Workflows Aspect "Regular" AI Workflow ReAct-Powered AI Workflow Core Process Direct generation; single inference pass Iterative "Reasoning + Acting" loop; step-by-step thinking and execution External interaction May be limited to no external tool use Actively leverages tools Adaptability Less adaptable; relies on training data. Highly adaptable; refines strategy based on real-time feedback. Problem solving Best for straightforward, single-turn tasks. Excels at complex, multi-step problems requiring external info and dynamic solutions Feedback Loop Generally no explicit feedback for self-correction Explicit real-time feedback loop to refine reasoning and adjust actions Transparency Often a black box; hard to trace logic. High visibility; explicit Chain-of-Thought and sequential actions show reasoning and output at each step Use case fit Simple Q&A, content generation Complex tasks: trip planning, research, multi-tool workflows Implementation Simple; requires AI chat integrations Complex; requires loop logic, tool integration, and might involve a multi-agent architecture Core Process I-Direct Generation; I-Single Inference Pass I-Iterative "I-Rasoning + Acting" isilinganiso; ukuyila kanye nokwenza isilinganiso External interaction Ungafaki ku-no external tool usebenzisa Ukusebenza okuzenzakalelayo Instruments Adaptability Okungenani adjustable; isekelwe data ukuqeqeshwa. Ukuhlobisa kakhulu; Izinzuzo zokusekelwe ku-feedback e-real-time. Problem solving I-Best for Simple, i-Single-Turn Task. I-Excels ku-complex, i-multi-step problems requiring external info and dynamic solutions Feedback Loop Ngokuvamile akukho ukubuyekeza ngokuvumelana self-correction I-explicit-real-time feedback loop ukucubungula ukucubungula kanye nokuguqulwa kwezinto Transparency Ngokuvamile ibhokisi black; okungenani ukucindezeleka logic. Ukuhlobisa okuphezulu; Chain-of-Thought kanye nezimo ezisebenzayo zibonisa isisombululo kanye nokukhipha ngalinye iminyango Use case fit Simple Q & A, ukwakhiwa kwekhwalithi Izinqubo ezinzima: Ukuhlolwa, Ukuhlolwa, Imisebenzi ye-multi-tool Implementation Simple; kufuneka AI chat ukuhlanganiswa I-complex; inikeza i-luke logic, ukuhlanganiswa kwezixhobo, futhi ingatholakala isakhiwo se-multi-agent Izinzuzo nezinzuzo Thinks, Actes, Learns, and Course-corrects on the fly. I-Excels ku-complex, i-multi-step tasks eyenza i-info ye-external : I-Integrates nge-instruments ezisebenzayo namafutha ze-external. : Hlola wonke umqondo kanye nemisebenzi, ukwenza ukucubungula umlilo. 👍 Super accurate and adaptable 👍 Handles gnarly problems 👍 External tool power 👍 Transparent and debuggable Imininingwane engaphezu kwama-Moving Parts kuyinto engaphezu kwe-Design ne-Management. I-Iterative loops, izivakashi ze-external, kanye ne-orchestration overhead kungenzeka ukuthi izindleko zokusebenza zangaphambili futhi izivakashi zihlukile kakhulu (e-cost to pay for more power and accuracy). 👎 Increased complexity 👎 Higher latency and calls Yini kufuneka ukuba Master ReAct Thina siphinde – ngaphandle kwezindlela ezifanele, i-agent ye-ReAct ayikho kakhulu enhle kunoma iyiphi enye inqubo ye-AI yokusebenza. Izixhobo zibonisa ukubuyekeza ekusebenzeni. Ngaphandle kwabo, ama-agents akuyona kuphela ... ukuhlala kakhulu. Ku-Bright Data, sinamathela ukuxhumanisa ama-agents e-AI ku-tools ezinhle. Ngakho-ke, sinikeza isakhiwo ephelele ukucubungula lokhu. Akungekho indlela yokucubungula ama-agents akho, sinamathela: I-Data Packs: I-curated, i-real-time, i-AI-ready datasets enhle yokusebenza kwe-RAG. I-MCP servers: I-AI-ready servers ifakwe nge-tools ye-data parsing, i-browser control, i-format conversion, nokunye. ️ I-SERP APIs: Hlola i-APIs i-LLM yakho angakwazi ukufinyelela iziphumo ze-web ezingenalutho, ezingenalutho - eyakhelwe ama-RAG pipelines. Izibuyekezo ze-Agent: Izibuyekezo eziholile ze-AI ezivela ku-web, ukuhlangabezana nezinsizakalo ze-IP, ukuhlangabezana ne-CAPTCHAs, futhi zihlole. ️ Ukusebenza kwe-MCP Server ... Futhi le toolstack iyatholakala ngokushesha. ➡️ Hlola ukuthi isakhiwo se-AI & BI ye-Bright Data ingasiza ukuvikela ama-agent yakho ye-next-gen. ➡️ Take a look at what can unlock for your next-gen agents. Bright Data’s AI & BI infrastructure I-AI & BI Infrastructure ye-Bright Data [Extra] I-ReAct Cheat Sheet Ngaphambi kokufaka, ukuthatha imizuzu yokuhlanza emoyeni. Kubalulekile kakhulu (ne-confusion) mayelana ne-terms "ReAct" - ikakhulukazi njengoba amabhizinisi amaningi abasebenzisa ku-contexts ahlukene. Ngakho-ke, apha i-non-fluff glossary yokusiza ukugcina yonke okuhlobene: "ReAct design pattern": Isakhiwo se-AI esihlanganisa ukubuyekeza nokuphendula. Umthengisi wabheka okokuqala (njenge-ketch-of-thought ukubuyekeza), bese isebenza (njenge-web search), futhi ekugcineni inikeza impendulo enhle. "I-ReAct prompting": Isinyathelo se-prompt-engineering enikeza i-LLM ukubonisa inqubo yayo yokuxhumana isinyathelo esisodwa kanye nokuthatha imiphumela esisodwa. It yenzelwe ukwenza imibuzo enhle, enhle, futhi engaphansi kwe-hallucination. Funda kabanzi mayelana ne-ReAct prompting. “ReAct agentic pattern”: Just another name for saying “ReAct design pattern.” "I-ReAct Agent": Yonke i-AI agent elandelayo isilinganiso se-ReAct. I-ReAct inikeza imiyalezo mayelana ne-task, isebenza imiyalezo esekelwe kulesi isilinganiso (njenge-Calling a Tool), futhi ivumela impendulo. I-ReAct Agent Framework: I-architecture (noma i-library) ebonakalayo ukwakha ama-Agents e-ReAct. It inikeza ukuvelisa yonke i-logic ye-reason-act-answer emakhasini yakho ye-AI. Final Thoughts Ngaba ufunde ukuthi i-ReAct inikeza kanjani ku-AI – ikakhulukazi lapho ithi ama-agents e-AI. Uyazi ukuthi lokhu isampula sokucubungula etholakalayo, ukuthi itholakalisa umbhalo, futhi indlela yokusebenza okuzenzakalelayo ukuze ikhiqize izinhlelo zakho zokusebenza ze-agent. Njengoba sihlolwe, ukunikela lezi zokusebenza ze-next-generation kubaluleke uma ungenza isakhiwo se-AI kanye ne-toolchain efanelekayo ukuvikela ama-agents akho. Ku-Bright Data, mission yethu kuyimpendulo enhle: ukwenza i-AI enhle kakhulu, enhle kakhulu, futhi engatholakali kakhulu bonke, emhlabeni wonke. Ngesikhathi esilandelayo-ukugcina ukujabulela, ukujabulela, futhi ukwakha futha ye-AI.