When we make API calls to AI chatbots, we ask a question, and they give us an answer. But if we ask another question, the AI has no idea what we were talking about before. It forgets everything from the previous message because most free APIs or hosted AI models work on a one-time request system.<\/p>\n\n\n\n
That means if you are building a chatbot for your website or a small project, your bot will respond like it is meeting the user for the first time every single time. No memory, no context, just a new blank start with every question.<\/p>\n\n\n\n
In this guide, I will show you how to fix that. How to add chatbot memory using only free AI APIs and JavaScript, without any database or backend setup. It is lightweight, works anywhere, and makes your chatbot feel way smarter.<\/p>\n\n\n\n
When people talk to a chatbot, they expect it to remember the topic. That memory is called context<\/strong>.<\/p>\n\n\n\n For example:<\/p>\n\n\n If your chatbot forgets context, it will reply something like:<\/p>\n\n\n That is what happens when your chatbot has no memory. It forgets every previous message because AI APIs treat every request as new.<\/p>\n\n\n\n Most AI APIs, like Gemini or GPT, work like vending machines. You put in one prompt and you get one reply. Once it is done, the machine does not remember you ever came there.<\/p>\n\n\n\n So if you send this:<\/p>\n\n\n Then you send:<\/strong><\/p>\n\n\n Now your chatbot goes blank and says:<\/strong><\/p>\n\n\n That is because your chatbot does not remember that you were talking about Rust.<\/p>\n\n\n\n Here is how to make your chatbot remember like a pro.<\/p>\n\n\n\n That way, your chatbot always knows what has been happening in the conversation.<\/p>\n\n\n\n You can even use a free model like Gemini to make that summary automatically.<\/p>\n\n\n\n We will do this step by step and keep it simple.<\/p>\n\n\n\n First, create a small array that will hold all the chat messages.<\/p>\n\n\n Each message will have a role and content.<\/p>\n\n\n Whenever a message comes, we will store it.<\/p>\n\n\n Now comes the fun part. We will use Gemini to summarize older parts of the chat so we do not overload the tokens.<\/p>\n\n\n This small function will send your chat history to Gemini and return a short summary. If something goes wrong, it will return an empty string, so your chatbot never breaks.<\/p>\n\n\n\n Now we make a function that combines recent messages and the summary.<\/p>\n\n\n This function makes sure the AI always gets the right amount of context without sending the entire history every time.<\/p>\n\n\n\n Now we send both the summary and the latest messages to the model.<\/p>\n\n\n You can use Gemini, Claude, or any other model here. The process is the same.<\/p>\n\n\n\n Here is what happens behind the scenes:<\/p>\n\n\n\n User message is stored, older messages are summarized if needed, the next request is sent with the summary plus latest messages, the model replies, and everything repeats.<\/p>\n\n\n\n If your chatbot has long conversations, you can summarize older parts regularly.<\/p>\n\n\n This trick keeps the bot smart without wasting tokens.<\/p>\n\n\n\n This simple context system can be used in<\/p>\n\n\n\n Even if your API does not have built-in memory, you can easily simulate it.<\/p>\n\n\n\n Most AI chatbots forget what we said two seconds ago. But with a little effort, you can give them a working memory.<\/p>\n\n\n\n All you need is to store messages, summarize older parts, and send both the summary and the recent messages each time.<\/p>\n\n\n\n It works with any AI model, free or paid, and it makes your chatbot feel smarter and more natural.<\/p>\n\n\n\n I like to think of it as giving my chatbot a tiny brain that can remember just enough to keep a good conversation going.<\/p>\n","protected":false},"excerpt":{"rendered":" When we make API calls to AI chatbots, we ask a question, and they give us an answer. But if we ask another question, the AI has no idea what we were talking about before. It forgets everything from the previous message because most free APIs or hosted AI models work on a one-time request […]<\/p>\n","protected":false},"author":79,"featured_media":36070,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"_surecart_dashboard_logo_width":"180px","_surecart_dashboard_show_logo":true,"_surecart_dashboard_navigation_orders":true,"_surecart_dashboard_navigation_invoices":true,"_surecart_dashboard_navigation_subscriptions":true,"_surecart_dashboard_navigation_downloads":true,"_surecart_dashboard_navigation_billing":true,"_surecart_dashboard_navigation_account":true,"_uag_custom_page_level_css":"","footnotes":""},"categories":[9],"tags":[],"class_list":["post-36065","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-js"],"blocksy_meta":[],"uagb_featured_image_src":{"full":["https:\/\/codeforgeek.com\/wp-content\/uploads\/2025\/10\/Add-Chatbot-Memory-Without-Database.png",1024,608,false],"thumbnail":["https:\/\/codeforgeek.com\/wp-content\/uploads\/2025\/10\/Add-Chatbot-Memory-Without-Database-150x150.png",150,150,true],"medium":["https:\/\/codeforgeek.com\/wp-content\/uploads\/2025\/10\/Add-Chatbot-Memory-Without-Database-300x178.png",300,178,true],"medium_large":["https:\/\/codeforgeek.com\/wp-content\/uploads\/2025\/10\/Add-Chatbot-Memory-Without-Database-768x456.png",768,456,true],"large":["https:\/\/codeforgeek.com\/wp-content\/uploads\/2025\/10\/Add-Chatbot-Memory-Without-Database.png",1024,608,false],"1536x1536":["https:\/\/codeforgeek.com\/wp-content\/uploads\/2025\/10\/Add-Chatbot-Memory-Without-Database.png",1024,608,false],"2048x2048":["https:\/\/codeforgeek.com\/wp-content\/uploads\/2025\/10\/Add-Chatbot-Memory-Without-Database.png",1024,608,false]},"uagb_author_info":{"display_name":"Aditya Gupta","author_link":"https:\/\/codeforgeek.com\/author\/aditya\/"},"uagb_comment_info":0,"uagb_excerpt":"When we make API calls to AI chatbots, we ask a question, and they give us an answer. But if we ask another question, the AI has no idea what we were talking about before. It forgets everything from the previous message because most free APIs or hosted AI models work on a one-time request…","_links":{"self":[{"href":"https:\/\/codeforgeek.com\/wp-json\/wp\/v2\/posts\/36065","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/codeforgeek.com\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/codeforgeek.com\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/codeforgeek.com\/wp-json\/wp\/v2\/users\/79"}],"replies":[{"embeddable":true,"href":"https:\/\/codeforgeek.com\/wp-json\/wp\/v2\/comments?post=36065"}],"version-history":[{"count":0,"href":"https:\/\/codeforgeek.com\/wp-json\/wp\/v2\/posts\/36065\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/codeforgeek.com\/wp-json\/wp\/v2\/media\/36070"}],"wp:attachment":[{"href":"https:\/\/codeforgeek.com\/wp-json\/wp\/v2\/media?parent=36065"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/codeforgeek.com\/wp-json\/wp\/v2\/categories?post=36065"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/codeforgeek.com\/wp-json\/wp\/v2\/tags?post=36065"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}\nUser: What is the weather in Delhi\nBot: It is sunny and 28 degrees\nUser: What about tomorrow\n<\/pre><\/div>\n\n\n
\nBot: Sorry, which city\n<\/pre><\/div>\n\n\n
\n\n\n\nWhy Chatbots Forget Everything<\/h2>\n\n\n\n
\nMessage 1: What is Rust\nYou get\nBot: Rust is a programming language\n<\/pre><\/div>\n\n\n
\nMessage 2: Who created it\n<\/pre><\/div>\n\n\n
\nBot: I do not know what you mean by it\n<\/pre><\/div>\n\n\n
\n\n\n\nThe Simple Plan to Fix It<\/h2>\n\n\n\n
\n
\n\n\n\nLet Us Code This in JavaScript<\/h2>\n\n\n\n
Step 1: Store Messages<\/h3>\n\n\n\n
\nlet conversation = [];\n<\/pre><\/div>\n\n\n
\n{\n role: "user" | "assistant" | "system",\n content: "The message text"\n}\n<\/pre><\/div>\n\n\n\nfunction addMessage(role, content) {\n conversation.push({ role, content });\n}\n<\/pre><\/div>\n\n\n
\n\n\n\nStep 2: Summarise Old Messages Using Gemini Free Model<\/h3>\n\n\n\n
\nasync function summarizeContextWithGemini(conversation) {\n const text = conversation.map(m => `${m.role}: ${m.content}`).join("\\n");\n\n const apiKey = process.env.GEMINI_API_KEY;\n const model = "gemini-2.5-flash";\n\n const body = {\n contents: [\n {\n parts: [\n { text: "Summarize this conversation briefly, keeping only important details" },\n { text: text }\n ]\n }\n ],\n model: model,\n temperature: 0.7,\n max_output_tokens: 100\n };\n\n const response = await fetch(\n "https:\/\/generativelanguage.googleapis.com\/v1beta\/models\/" + model + ":generateContent",\n {\n method: "POST",\n headers: {\n "Content-Type": "application\/json",\n "x-goog-api-key": apiKey\n },\n body: JSON.stringify(body)\n }\n );\n\n if (!response.ok) {\n console.log("Gemini summarization failed", await response.text());\n return "";\n }\n\n const data = await response.json();\n const summary = data.candidates?.[0]?.content?.parts?.[0]?.text || "";\n return summary;\n}\n<\/pre><\/div>\n\n\n
\n\n\n\nStep 3: Build the Final Context for the Next Request<\/h3>\n\n\n\n
\nasync function buildContext() {\n const MAX_MESSAGES = 6;\n let recentMessages = conversation.slice(-MAX_MESSAGES);\n\n if (conversation.length > MAX_MESSAGES) {\n const summary = await summarizeContextWithGemini(conversation.slice(0, -MAX_MESSAGES));\n recentMessages = [\n { role: "system", content: "Summary of previous conversation: " + summary },\n ...recentMessages\n ];\n }\n\n return recentMessages;\n}\n<\/pre><\/div>\n\n\n
\n\n\n\nStep 4: Send the Message to the AI Model<\/h3>\n\n\n\n
\nasync function getAIResponse(userMessage) {\n addMessage("user", userMessage);\n\n const contextMessages = await buildContext();\n\n const response = await fetch("https:\/\/api.openai.com\/v1\/chat\/completions", {\n method: "POST",\n headers: {\n "Content-Type": "application\/json",\n "Authorization": `Bearer ${process.env.OPENAI_API_KEY}`,\n },\n body: JSON.stringify({\n model: "gpt-4o",\n messages: contextMessages,\n max_tokens: 200\n })\n });\n\n const data = await response.json();\n const reply = data.choices?.[0]?.message?.content || "No response";\n\n addMessage("assistant", reply);\n return reply;\n}\n<\/pre><\/div>\n\n\n
\n\n\n\nStep 5: Test the Chatbot Flow<\/h3>\n\n\n
\nasync function chatExample() {\n console.log(await getAIResponse("Hey, who are you"));\n console.log(await getAIResponse("What can you do"));\n console.log(await getAIResponse("Remind me what I asked earlier"));\n}\n<\/pre><\/div>\n\n\n
\n\n\n\nSave Tokens with Progressive Summarization<\/h2>\n\n\n\n
\nif (conversation.length > 20) {\n const summary = await summarizeContextWithGemini(conversation.slice(0, 15));\n conversation = [\n { role: "system", content: "Previous summary: " + summary },\n ...conversation.slice(15)\n ];\n}\n<\/pre><\/div>\n\n\n
\n\n\n\nWhere You Can Use This<\/h2>\n\n\n\n
\n
\n\n\n\nFinal Thoughts<\/h2>\n\n\n\n