This is largely an experiment to get familiar with the openAI API, but it is functional in case anyone feels like playing with it. I realize that that there are simpler APIs available for moderation, but I was interested in exploring an LLM for moderation. The results are promising, although could do with further fine tuning. The API is nowhere near real-time (takes several seconds to respond), so the plugin is implemented with action filters. I do ike the feedback provided by the API and will probably expose that to users via a non-obtrusive tooltip once I've let it run for a while. The costs are reasonable, at least for a low traffic site like mine. It works out to roughly $0.001 per moderation, depending on whether topic context is enabled and the size of the posts. So, figure about $1 for 1000 posts. I'm sure similar results could be obtained with an open source LLM as well. As a side note, my day job is developer adjacent but I don't write a lot of code, so its probably not the most efficiently written plugin.

Moderation Plugin

Settings

• openAPI token - API key
• Moderator - User id for moderation activities
• Bypass moderation when user has reputation higher than this value - Leave empty or set to 0 if you want all users to be moderated, otherwise set to the threshold below which moderation will be triggered.
• Enable Notifications for Moderation Actions - Enable/Disable notifications for moderation actions
• Delete Moderated Posts - Enable this if you want moderated posts to be deleted rather than flagged
• Add Topic Context - Enable this if you want to include contextual content from the main post as well as the post being replied to (if applicable)
• Negative Behaviors - This is the list of behaviors that should be considered when moderating a post. The defaults are "Spam", "Profanity", "Racism", "Sexism", "Homophobia", "Aggressive Language”
• Code of Conduct/Terms of Service - (Optional) Add your Code of Conduct or Terms of Service to provide additional context for moderation of posts

Some examples of moderation results and feedback

Post 1

The Albanians are an ethnic group native to the Balkan Peninsula who share a common Albanian ancestry, culture, history and language. They primarily live in Albania, Kosovo, North Macedonia, Montenegro, Serbia as well as in Croatia, Greece, Italy and Turkey. They also constitute a large diaspora with several communities established across Europe, the Americas and Oceania.

Albanians have Paleo-Balkanic origins. Exclusively attributing these origins to the Illyrians, Thracians or other Paleo-Balkan people is still a matter of debate among historians and ethnologists.

Results - passed

Feedback
Thank you for your informative post on the Albanians. It's great that you've provided information on their ancestry, clulture and history. Your post is compliant with our code of conduct, and there is no evidence of negative behavior patterns. Keep up the good work!

{       "spam": false,
        "profanity": false,
        "racism": false,
        "sexism": false,
        "homophobia": false,
        "aggressive_language": false,
        "compliance": true,
        "passed": true,
        "feedback": "Thank you for your informative post on the Albanians. It's great that you've provided information on their ancestry, clulture and history. Your post is compliant with our code of conduct, and there is no evidence of negative behavior patterns. Keep up the good work!"
    }

Post 2

Added a negative (homophobic) term to the same post

Results : failed

Feedback
Hello, please note that the use of derogatory language directed towards individuals based on their sexual orientation is not allowed in our community. Please refrain from using such language in future posts. Additionally, please ensure that posts comply with our code of conduct that pledges to make participation in our community a harassment-free experience for everyone regardless of any personal characteristic. Thank you for your understanding and cooperation in maintaining a healthy and inclusive community.

{
        "spam": false,
        "profanity": true,
        "racism": false,
        "sexism": true,
        "homophobia": true,
        "aggressive_language": false,
        "compliance": false,
        "passed": false,
        "feedback": "Hello, please note that the use of derogatory language directed towards individuals based on their sexual orientation is not allowed in our community. Please refrain from using such language in future posts. Additionally, please ensure that posts comply with our code of conduct that pledges to make participation in our community a harassment-free experience for everyone regardless of any personal characteristic. Thank you for your understanding and cooperation in maintaining a healthy and inclusive community."
    }

Post 3

Replied to original post with minimum verbiage

Looks good!

Results: passed

Feedback
Thank you for your contribution! Please note that in order to maintain a high quality of discussion, we require substantive content that is meaningfully related to the subject under discussion. Please provide more information or context in your future posts. Also, please remember to be respectful and considerate of other community members' opinions and experiences. Thank you!

{
        "spam": false,
        "profanity": false,
        "racism": false,
        "sexism": false,
        "homophobia": false,
        "aggressive_language": false,
        "compliance_with_code_of_conduct": true,
        "passed": true,
        "friendly_feedback": "Thank you for your contribution! Please note that in order to maintain a high quality of discussion, we require substantive content that is meaningfully related to the subject under discussion. Please provide more information or context in your future posts. Also, please remember to be respectful and considerate of other community members' opinions and experiences. Thank you!"
    }

Post 4

Replied with hostility

This is full of c**p, you have no idea what you’re talking about!

Results: failed

Feedback
Your post contains aggressive language and profanity which is not acceptable on our community platform. We request you to refrain from such language and keep the conversation respectful and courteous. Additionally, it’s important to adhere to our code of conduct which states that “participation in our community must be a harassment-free experience for everyone”. We urge you to respect all members of our community and aim towards creating a diverse, inclusive and healthy environment for all.

{
        "Spam": false,
        "Profanity": true,
        "Racism": false,
        "Sexism": false,
        "Homophobia": false,
        "Aggressive Language": true,
        "Code of Conduct": false,
        "Passed": false,
        "feedback": "Your post contains aggressive language and profanity which is not acceptable on our community platform. We request you to refrain from such language and keep the conversation respectful and courteous. Additionally, it’s important to adhere to our code of conduct which states that “participation in our community must be a harassment-free experience for everyone”. We urge you to respect all members of our community and aim towards creating a diverse, inclusive and healthy environment for all."
    }

Post 5

Replied with comments on a different ethnic group (with Topic Context enabled)
Note: Context and Relevance are tricky and might need to be fine tuned.

Italians are an ethnic group native to the Italian geographical region and its neighboring insular territories. Italians share a common culture, history, ancestry and language. Their predecessors differ regionally, but generally include native populations such as the Etruscans, and the Italic peoples, including the Latins, from which the Romans emerged and helped create and evolve the modern Italian identity.

Results: failed

Feedback
Thank you for sharing information on the ethnic group Italians and their history. However, please ensure that your posts are directly related to the topic of discussion. In this case, while the topic is about Albanians, your post does not seem to contribute to the conversation. Please try to stay on topic in future postings.

{
  "spam": false,
  "profanity": false,
  "racism": false,
  "sexism": false,
  "homophobia": false,
  "aggressive_language": false,
  "code_of_conduct": false,
  "context_and_relevance": false,
  "passed": false,
  "feedback": "Thank you for sharing information on the ethnic group Italians and their history. However, please ensure that your posts are directly related to the topic of discussion. In this case, while the topic is about Albanians, your post does not seem to contribute to the conversation. Please try to stay on topic in future postings."
}

The plugin will now reply to any questions in a post (after it has passed moderation). This will obviously become much more powerful once openAI allows for training the gpt-3.5-turbo model on your site's data. In the meantime, it uses the context of the post to answer based on its current training set.

So, using the same topic as above and asking a question

is Albanian a Slavic language?

Generates a reply to the post

{                                                                                                                                      "spam": false,
  "profanity": false,
  "racism": false,
  "sexism": false,
  "homophobia": false,
  "aggressive_language": false,
  "code_of_conduct": true,
  "question": true,
  "answer": "No, Albanian is an Indo-European language and belongs to the group of Balkan languages, which also includes Slavic languages, but it is not a Slavic language. Albanian has its own unique grammar, vocabulary, and syntax that distinguish it from Slavic languages.",
  "passed": true,
  "feedback": "Thank you for your question. It's important to remember to keep discussions respectful and relevant to the topic at hand. Let's continue to learn and share knowledge in a positive and inclusive manner.",
  "timestamp": 1684860476156
}

AI Generated Response

No, Albanian is an Indo-European language and belongs to the group of Balkan languages, which also includes Slavic languages, but it is not a Slavic language. Albanian has its own unique grammar, vocabulary, and syntax that distinguish it from Slavic languages.

I'm trying to load a default set of global tags so they can be used for the autocomplete function when creating new topics. However, adding new tags in the 'Manage Tags' section doesn't seem to enable this functionality. Is this not the intended purpose of adding new tags through 'Manage Tags'?
I know I can add a set of tags to the whitelist for each category, but that would be a lot of duplicated tags and would also prevent users from adding new custom tags.

Is there any built-in way to achieve this?

I've been exploring the possibility of using a cloud based "serverless" DB as a drop-in replacement for mongodb. This is the only remaining component in the nodeBB architecture that is currently relatively difficult to scale without over-provisioning for peak capacity. There are a couple of candidates that appear to qualify on paper. The first being Cosmos DB

• Serverless deployments: Cosmos DB for MongoDB offers a serverless capacity mode. With Serverless, you're only charged per operation, and don't pay for the database when you don't use it.

• Free Tier: With Azure Cosmos DB free tier, you get the first 1000 RU/s and 25 GB of storage in your account for free forever, applied at the account level. Free tier accounts are automatically sandboxed so you never pay for usage.

I was able to dump and restore to Cosmos relatively easily

of course nothing ever works the first time out as nodeBB refused to start

2023-05-18T16:40:46.781Z [4564/21343] - error: MongoServerError: Error=2, Details='Response status code does not indicate success: BadRequest (400); Substatus: 0; ActivityId: 1a6747b1-c46c-4fb1-9847-f9a788f35376; Reason: (Response status code does not indicate success: BadRequest (400); Substatus: 0; ActivityId: 1a6747b1-c46c-4fb1-9847-f9a788f35376; Reason: (Response status code does not indicate success: BadRequest (400); Substatus: 0; ActivityId: 1a6747b1-c46c-4fb1-9847-f9a788f35376; Reason: (Message: {"Errors":["The index path corresponding to the specified order-by item is excluded."]}

I haven't had a chance to dig deeper, but will do so at some time.

The next candidate is Tigris, which is promising because it promises
MongoDB compatibilty as well as enhanced full text search.

• Cloud-native architecture: Tigris follows a modern composable architecture with loosely coupled components. Furthermore, compute is separated from storage, allowing for independent scaling and a more resilient system.
• Automatic database sharding: Tigris provides automatic database sharding, and shard keys are unnecessary as the data distribution is automatically handled. This includes both sharding and optimizing the number of shards. MongoDB Atlas requires you to specify and manage sharding to horizontally scale. With Tigris you get more time to focus on developing applications for your users.
  Cost-efficient and unlimited scalability: Unlike MongoDB Atlas, Tigris can scale to millions of records read and written per second and petabytes of data storage at a fraction of the cost

@oplik0 I'll have to play with Google's APIs to get a better feel for their relative performance, although if its anythling like my experience with Bard, I supect they're a couple of steps behind. This is becoming a very competitive space and I'm sure prices will drop rapidly. The price for gpt-3.5-turbo is already half that of gpt-4. I've implemented some of Google's moderation APIs in the past and they're fine for binary detection of toxicity, but the value that LLM's provide in this space is their ability to assess posts in context with the conversation flow and the subject matter of the topic. The qualitative feedback in human (albeit robotic) language is a nice bonus.

@phenomlab said in NodeBB Assets - Object Storage:

This has a lot of potential using workers and R2, and really needs proper documentation. As you know, it's taken me 4 hours to get this working!

Yeah, it only took me a few minutes to get it working as a POC this morning, but , as you experienced, getting that even partially operationalized took a lot longer.

The nice thing about implementing this is that it makes nodeBB essentially infinitely scalable. All static assets are handled by cloudflare workers while the nodeBB API can be deployed to droplets behind a load balancer. Spin up droplets as demand increases and then spin them back down when no longer needed.

@dave1904 ++ for this to be part of core. I had requested a similar feature last year, but it did not get much traction

Ideally, this should be implemented via the socket api so it can update in real-time and serve as a view into interesting conversations

That will work perfectly. I'll create a custom topic field for the total imported votes and add them to the calculated votes.

Appreciate the help!

@o-o Agreed, much as I love nodeBB I would love to see the UI decoupled from the API services. To be fair, the existing nodeBB architecture is already structured to allow for this. The only missing piece is a statically hostable front end written in Next.js (or Nuxt, Astro etc). I would especially like this as it would make nodeBB embeddable as a "forums as a service" provider without having to resort to subdomains or embedding in iframes. I imagine that the vast majority of nodeBB (and forum software in general) are using it as an adjunct to existing websites, and this would instantly make nodeBB a much more attractive option compared to the competition.

Thanks, that works perfectly. I'm handling the hook in my custom theme, but this probably needs to be in the https://community.nodebb.org/post/95161 plugin as well (without the importedVotes)

library.setTopicVotes = async function(hookData) {
    const topicData = await topics.getTopicFields(hookData.post.tid, ['cid', 'importedVotes']);
    const importedVotes = topicData.importedVotes || 0
    const voteData = await db.getSortedSetRangeByScoreWithScores(`tid:${hookData.post.tid}:posts:votes`, 0, -1, 1, '+inf');
    let allvotes = voteData.reduce((acc, cur) => acc + cur.score, 0);
    allvotes = allvotes + importedVotes
    await db.sortedSetAdd(`cid:${topicData.cid}:tids:votes`, allvotes, hookData.post.tid)
}

@eeeee I wasn't suggesting that the nodeBB team select Next.js, that choice should be based on whatever framework works best for nodeBB based on the choices available. My point is simply that the UI should be a separate component and not married to a nodeBB monolith.
BTW, I have used Next.js and it wouldn't be my personal choice for a new project. However, it has, by far, the largest installed base of production apps and if the goal is to appeal to the largest set of potential users of forum software, it would be an easy front runner for the "default" front end UI.
Keep in mind, that if the nodeBB team decides to go down this route, the front end
UI essentially becomes the equivalent of the current concept of themes. This allows anyone to use their framework of choice in their own implementation.

I was able to achieve the desired functionality by adding a custom "Share" button to the template for the private topics. This button triggers the move to predesignated categories based on the type of post.
Cheers

@The-SkyFoxx Decoupling the UI and choosing a framework are two separate architectural choices and not tied to each other. You can decouple the existing UI from the backend without rewriting the existing core or moving away from jQuery or breaking any existing functionality such as themes and client-side plugins.
At its most simplistic, it would be simply 2 express servers, one for hosting the client JS and static assets, while the second would be for the API services and authentication/session management. The UI express server could also be compiled to statically hosted javascript that would no longer need an express server and could be served from S3 buckets. This is easy enough to do for express based apps. The benefits of this approach are quite worthwhile. As the previous posts suggested, it allows for separate lifecycles for the UI and the core API services. It should also boost performance substantially as all static resources including the UI would be served from S3 buckets.
Optionally, the standalone API server now freed from hosting the UI, could be moved to an API server better optimized for pure API services such as fastify. You can lookup the various benchmarks, but Fastify or Restify are anywhere from 3X- 5X faster than express for API requests.
Lastly, I would also decouple the session management, but that's a "nice to have" that would let nodeBB be a truly serverless application that could run entirely on the cloud and be essentially endlessly scalable.

So basically fire the event:post_edited socket event?

					websockets.in(`topic_${tid}`).emit('event:post_edited', {
						post: {
							tid,
							pid,
							changed: true,
							content,
						},
						topic: {},
					});

Here is a slightly modified version of Baris' script that reads a list of tags from a JSON file (tags.json) and creates them

'use strict';
const nconf = require('nconf');
nconf.file({
  file: 'config.json',
});
nconf.defaults({
  base_dir: __dirname,
  views_dir: './build/public/templates',
  upload_path: 'public/uploads',
});
var db = require('./src/database');
const meta = require('./src/meta');
const categories = require('./src/categories');
const cache = require('./src/cache');
const tags = require('./tags.json');
db.init(function (err) {
  if (err) {
    console.log('NodeBB could not connect to your database. Error: ' + err.message);
    process.exit();
  }
  setTags(function (err) {
  console.log('setting');
    if (err) {
      console.error(err);
      process.exit();
    }
    console.log('done');
    process.exit();
  });
});
  async function createEmptyTag(tag) {
    if (!tag) {
      throw new Error('[[error:invalid-tag]]');
    }
    const isMember = await db.isSortedSetMember('tags:topic:count', tag);
    console.log(1, isMember);
    if (!isMember) {
      await db.sortedSetAdd('tags:topic:count', 0, tag);
      cache.del('tags:topic:count');
    }
    const allCids = await categories.getAllCidsFromSet('categories:cid');
    const isMembers = await db.isMemberOfSortedSets(
      allCids.map(cid => `cid:${cid}:tags`), tag
    );
    const bulkAdd = allCids.filter((cid, index) => !isMembers[index])
      .map(cid => ([`cid:${cid}:tags`, 0, tag]));
    await db.sortedSetAddBulk(bulkAdd);
    console.log('added ' + tag);
  };

async function setTags(callback) {
  tags.forEach(async (tag) => {
    console.log('adding tags', tag);
    await createEmptyTag(tag)
  });
      //callback();
  }

I agree that this should be built into the default theme. In my case, I added it to my custom theme by adding a script in the header.tpl template. The scripts below should be usable as is. It's a little more complex than absolutely neccessary, but in my case I needed to be able to match the mode of the parent site. The modes are kept in sync via a watcher on local storage. The script accepts 3 modes - system, dark and light. The system mode matches the OS mode preference. It also alllows to a page to be opened in a specific mode via a search query parameter, for example ?theme=light

    <script id="setTheme" async=false>
        function getParameterByName(name, url) {
            if (!url) url = window.location.href;
            name = name.replace(/[\[\]]/g, "\\$&");
            var regex = new RegExp("[?&]" + name + "(=([^&#]*)|&|#|$)"),
                results = regex.exec(url);
            if (!results) return null;
            if (!results[2]) return '';
            return decodeURIComponent(results[2].replace(/\+/g, " "));
        }

        function getInitialColorMode() {
            let persistedColorPreference = getParameterByName('theme', window.location.url);
            if (persistedColorPreference && typeof persistedColorPreference === 'string') {
                window.localStorage.setItem('system-theme', persistedColorPreference);
            } else {
                persistedColorPreference = window.localStorage.getItem('system-theme');
            }
            const hasPersistedPreference = typeof persistedColorPreference === 'string' && persistedColorPreference != 'system';
            if (hasPersistedPreference) {
                setColors(persistedColorPreference)
                return persistedColorPreference;
            }
            const mql = window.matchMedia('(prefers-color-scheme: dark)');
            const hasMediaQueryPreference = typeof mql.matches === 'boolean';
            if (hasMediaQueryPreference) {
                setColors(mql.matches ? 'dark' : 'light')
                return mql.matches ? 'dark' : 'light';
            }
            setColors('light')
            return 'light';
        }

        function setColors(value) {
            if (value == 'dark') {
                document.documentElement.style.setProperty('--bg-color', '#191919');
                document.documentElement.style.setProperty('--color', '#ffffffcc');
                document.documentElement.style.setProperty('--nav-bg-color', '#151515');
                document.documentElement.style.setProperty('--nav-border', '#b0b0b09c');
            } else {
                document.documentElement.style.setProperty('--bg-color', '#fafafa');
                document.documentElement.style.setProperty('--color', '#444444');
                document.documentElement.style.setProperty('--nav-bg-color', '#fafafa');
                document.documentElement.style.setProperty('--nav-border', '#eeeeee');
            }
        }
        async function setStart() {
            document.documentElement.classList.toggle("dark", getInitialColorMode() == 'dark');
            document.documentElement.classList.toggle("light", getInitialColorMode() != 'dark');
            window.addEventListener('storage', (e) => {
                if (e.key != 'system-theme') {
                    return;
                } else {
                    document.documentElement.classList.toggle("dark", getInitialColorMode() == 'dark');
                    document.documentElement.classList.toggle("light", getInitialColorMode() != 'dark');
                }
            })
            const darkMode = window.matchMedia('(prefers-color-scheme: dark)');
            darkMode.addListener((event) => {
                document.documentElement.classList.toggle("dark", getInitialColorMode() == 'dark')
                document.documentElement.classList.toggle("light", getInitialColorMode() != 'dark')
            });
        }
        setStart()
    </script>

@DownPW I use Backblaze for a couple of sites and since I haven't exceeded the free 10GB of storage, I haven't paid anything over the last year. In your case, you will have to pay for the extra 7GB of storage, but as I mentioned earlier that's less than 10 cents a month.

In terms of development, none should be required. The nice thing about using storage based on a industry standard like S3 is that it "just works" across all vendors. Add the S3 plugin, set your keys and you're good to go. All uploads will be saved to your bucket.

Another nice benefit is full data portability. I will be moving my storage to Cloudflare to consolidate my assets storage as mentioned in the other post. It's a very simple process using a tool like rclone and a simple DNS change to point to my new bucket(s)

I really like the preview feature where if you link to an existing topic, a preview is generated when you hover over that link. I would like to leverage this preview as an embed in external pages/websites in an iframe. Is there an existing route for the preview or would I need a plugin to expose it?
As a corollary to the above, how would I go about exposing the votes on a post to an external page? Is the socket API open for usage on a different site (same domain, but different subdomain)

I'm trying to build a WYSIWIG composer plugin using summernote with the redactor plugin as a starting point/template. However, at some point between hitting the submit button and the rendering of the final HTML, some filtering seems to be occurring. Is there any documentation available that would help me understand the flow from submission to saving to rendering so I can identify where the filtering happens?

In theory I have gigabit fiber, but I've never seen anything above 600 even when hardwired to ethernet. This is on WiFi