From 1ade380bd73424f1a11639daa0c338b1b9092dea Mon Sep 17 00:00:00 2001
From: Matt Williams <m@technovangelist.com>
Date: Thu, 7 Dec 2023 11:48:25 -0800
Subject: [PATCH] Simple chat example for typescript

Signed-off-by: Matt Williams <m@technovangelist.com>
---
 examples/typescript-simplechat/client.ts    | 78 +++++++++++++++++++++
 examples/typescript-simplechat/package.json |  1 +
 examples/typescript-simplechat/readme.md    | 31 ++++++++
 3 files changed, 110 insertions(+)
 create mode 100644 examples/typescript-simplechat/client.ts
 create mode 100644 examples/typescript-simplechat/package.json
 create mode 100644 examples/typescript-simplechat/readme.md

diff --git a/examples/typescript-simplechat/client.ts b/examples/typescript-simplechat/client.ts
new file mode 100644
index 00000000..7e37fe30
--- /dev/null
+++ b/examples/typescript-simplechat/client.ts
@@ -0,0 +1,78 @@
+import * as readline from "readline";
+
+const model = "llama2";
+type Message = {
+  role: "assistant" | "user" | "system";
+  content: string;
+}
+const messages: Message[] = [{
+  role: "system",
+  content: "You are a helpful AI agent."
+}]
+
+const rl = readline.createInterface({
+  input: process.stdin,
+  output: process.stdout
+})
+
+async function chat(messages: Message[]): Promise<Message> {
+  const body = {
+    model: model,
+    messages: messages
+  }
+
+  const response = await fetch("http://localhost:11434/api/chat", {
+    method: "POST",
+    body: JSON.stringify(body)
+  })
+
+  const reader = response.body?.getReader()
+  if (!reader) {
+    throw new Error("Failed to read response body")
+  }
+  const content: string[] = []
+  while (true) {
+    const { done, value } = await reader.read()
+    if (done) {
+      break;
+    }
+    const rawjson = new TextDecoder().decode(value);
+    const json = JSON.parse(rawjson)
+
+    if (json.done === false) {
+      process.stdout.write(json.message.content);
+      content.push(json.message.content)
+      // messages.push({role: "system", content: text})
+    }
+
+  }
+  return { role: "assistant", content: content.join("") };
+}
+
+async function askQuestion(): Promise<void> {
+  return new Promise<void>((resolve) => {
+    rl.question("\n\nAsk a question: (press enter alone to quit)\n\n", async (user_input) => {
+      if (user_input.trim() === "") {
+        rl.close();
+        console.log("Thankyou. Goodbye.\n")
+        console.log("=======\nHere is the message history that was used in this conversation.\n=======\n")
+        messages.forEach(message => {
+          console.log(message)
+        })
+        resolve();
+      } else {
+        console.log();
+        messages.push({ role: "user", content: user_input });
+        messages.push(await chat(messages));
+        await askQuestion(); // Ask the next question
+      }
+    });
+  });
+}
+
+async function main() {
+  await askQuestion();
+
+}
+
+main();
\ No newline at end of file
diff --git a/examples/typescript-simplechat/package.json b/examples/typescript-simplechat/package.json
new file mode 100644
index 00000000..4ee1647d
--- /dev/null
+++ b/examples/typescript-simplechat/package.json
@@ -0,0 +1 @@
+{ "dependencies": { "@types/node": "^20.10.4", "prompt-sync": "^4.2.0", "readline": "^1.3.0" } }
\ No newline at end of file
diff --git a/examples/typescript-simplechat/readme.md b/examples/typescript-simplechat/readme.md
new file mode 100644
index 00000000..ea61bd8a
--- /dev/null
+++ b/examples/typescript-simplechat/readme.md
@@ -0,0 +1,31 @@
+# Simple Chat Example
+
+The **chat** endpoint is one of two ways to generate text from an LLM with Ollama. At a high level you provide the endpoint an array of objects with a role and content specified. Then with each output and prompt, you add more of those role/content objects, which builds up the history.
+
+## Review the Code
+
+You can see in the **chat** function that actually calling the endpoint is done simply with:
+
+```typescript
+const body = {
+  model: model,
+  messages: messages
+}
+
+const response = await fetch("http://localhost:11434/api/chat", {
+  method: "POST",
+  body: JSON.stringify(body)
+})
+```
+
+With the **generate** endpoint, you need to provide a `prompt`. But with **chat**, you provide `messages`. And the resulting stream of responses includes a `message` object with a `content` field.
+
+The final JSON object doesn't provide the full content, so you will need to build the content yourself. In this example, **chat** takes the full array of messages and outputs the resulting message from this call of the chat endpoint.
+
+In the **askQuestion** function, we collect `user_input` and add it as a message to our messages and that is passed to the chat function. When the LLM is done responding the output is added as another message to the messages array.
+
+At the end, you will see a printout of all the messages.
+
+## Next Steps
+
+In this example, all generations are kept. You might want to experiment with summarizing everything older than 10 conversations to enable longer history with less context being used.