docs: add example scripts and instructions.

Author: zicklag

.gitignore

@@ -2,6 +2,8 @@ target
 model
 ort
 
+example/conversations.json
+
 # pixi environments
 .pixi
 *.egg-info

README.md

@@ -44,6 +44,38 @@ The endpoint will return a list of items with their ID and the similarity score
 You can optionally add a `limit` query parameter which will set a limit on how many items are
 returned, for example: `POST /search?limit=20`. The default limit is `10`.
 
+## Example
+
+The repo includes a quick example of how to use it to search a database of chat messages using Deno.
+
+```bash
+cd example
+
+# First we need to start the search services
+docker compose up -d
+
+# Then we need to download the conversation database. That will download
+# the `conversations.json` file containing a bunch of sample messages.
+./download-conversations.sh
+
+# Now we need to index the messages. This may take a while to index the 188k
+# messages, but you can search the messages while they are being indexed and
+# if you don't care to index all of them you can cancel it at any time and
+# all of the messages that have already been indexed will still be searchable.
+deno run -A indexConversations.ts
+
+# After that's done ( or during indexing in another terminal ), we can search
+# for chat messages.
+deno run -A search.ts Do you like dogs?
+```
+
+Check out the scripts in the `example` folder to see how easy it is to search and index messages!
+
+### Qdrant Dashboard
+
+You can also access the QDrant web dashboard after running `docker compose up -d` by going to
+<http://localhost:6333/dashboard#/collections>.
+
 ## Deployment
 
 The service is easily deployed, along with Qdrant, using docker compose:

example/conversations.ts

@@ -0,0 +1,17 @@
+const conversations: {
+  [key: string]: {
+    article_url: string;
+    config: string;
+    content: {
+      message: string;
+      agent: string;
+      sentiment: string;
+      knowledge_source: string[];
+      turn_rating: string;
+    }[];
+  };
+} = JSON.parse(await Deno.readTextFile("./conversations.json"));
+
+export const messages: string[] = Object.values(conversations)
+  .map((x) => x.content.map((x) => x.message))
+  .flat();

example/download-conversations.sh

@@ -0,0 +1 @@
+wget https://github.com/alexa/Topical-Chat/raw/refs/heads/master/conversations/train.json -O conversations.json

example/indexConversations.ts

@@ -0,0 +1,9 @@
+import { messages } from "./conversations.ts";
+
+for (let i = 0; i < messages.length; i++) {
+  const message = messages[i];
+  await fetch(`http://localhost:3000/index/${i}`, {
+    method: "post",
+    body: message,
+  });
+}

example/search.ts

@@ -0,0 +1,19 @@
+import { messages } from "./conversations.ts";
+
+const query = Deno.args.join(" ");
+
+const resp = await fetch("http://localhost:3000/search?limit=20", {
+  method: "post",
+  body: query,
+});
+
+const results: { id: string; score: number }[] = await resp.json();
+
+console.log(
+  results
+    .map((x) => ({
+      message: messages[parseInt(x.id)],
+      score: x.score,
+    }))
+    .reverse()
+);