feat(bots/discord): add code to actually scan text files correctly

bots/discord/src/events/discord/messageCreate/messageScan.ts

@@ -59,26 +59,38 @@ withContext(on, 'messageCreate', async (context, msg) => {
         }
     }
 
-    if (msg.attachments.size > 0 && config.attachments?.scanAttachments) {
+    if (msg.attachments.size && config.attachments?.scanAttachments) {
         logger.debug(`Classifying message attachments for ${msg.id}`)
 
         for (const attachment of msg.attachments.values()) {
+            const mimeType = attachment.contentType?.split(';')?.[0]
+            if (!mimeType) return void logger.warn(`No MIME type for attachment: ${attachment.url}`)
+
             if (
                 config.attachments.allowedMimeTypes &&
-                !config.attachments.allowedMimeTypes.includes(attachment.contentType!)
+                !config.attachments.allowedMimeTypes.includes(mimeType)
             ) {
-                logger.debug(`Disallowed MIME type for attachment: ${attachment.url}, ${attachment.contentType}`)
+                logger.debug(`Disallowed MIME type for attachment: ${attachment.url}, ${mimeType}`)
                 continue
             }
 
-            if (attachment.contentType?.startsWith('text/') && attachment.size > (config.attachments.maxTextFileSize ?? 512 * 1000)) {
+            const isTextFile = mimeType.startsWith('text/')
+
+            if (isTextFile && attachment.size > (config.attachments.maxTextFileSize ?? 512 * 1000)) {
                 logger.debug(`Attachment ${attachment.url} is too large be to scanned, size is ${attachment.size}`)
                 continue
             }
 
             try {
+                let response: Awaited<ReturnType<typeof getResponseFromText>>['response'] | undefined
+
+                if (isTextFile) {
+                    const content = await (await fetch(attachment.url)).text()
+                    response = await getResponseFromText(content, filteredResponses, context, { skipApiRequest: true }).then(it => it.response)
+                } else {
                     const { text: content } = await api.client.parseImage(attachment.url)
-                const { response } = await getResponseFromText(content, filteredResponses, context, true)
+                    response = await getResponseFromText(content, filteredResponses, context, { onlyImageTriggers: true }).then(it => it.response)
+                }
 
                 if (response) {
                     logger.debug(`Response found for attachment: ${attachment.url}`)
@@ -89,8 +101,8 @@ withContext(on, 'messageCreate', async (context, msg) => {
 
                     break
                 }
-            } catch {
-                logger.error(`Failed to parse image: ${attachment.url}`)
+            } catch (e) {
+                logger.error(`Failed to parse attachment: ${attachment.url}`, e)
             }
         }
     }

bots/discord/src/utils/discord/messageScan.ts

@@ -9,7 +9,7 @@ export const getResponseFromText = async (
     responses: ConfigMessageScanResponse[],
     // Just to be safe that we will never use data from the context parameter
     { api, logger }: Omit<typeof import('src/context'), 'config'>,
-    ocrMode = false,
+    flags: { onlyImageTriggers?: boolean; skipApiRequest?: boolean } = {}
 ): Promise<
     Omit<ConfigMessageScanResponse, 'triggers'> & { label?: string; triggers?: ConfigMessageScanResponse['triggers'] }
 > => {
@@ -31,7 +31,7 @@ export const getResponseFromText = async (
             triggers: { text: textTriggers, image: imageTriggers },
         } = trigger
 
-        if (ocrMode) {
+        if (flags.onlyImageTriggers) {
             if (imageTriggers)
                 for (const regex of imageTriggers)
                     if (regex.test(content)) {
@@ -57,7 +57,7 @@ export const getResponseFromText = async (
     }
 
     // If none of the regexes match, we can search for labels immediately
-    if (!responseConfig.triggers && !ocrMode) {
+    if (!responseConfig.triggers && !flags.onlyImageTriggers && !flags.skipApiRequest) {
         logger.debug('No match from before regexes, doing NLP')
         const scan = await api.client.parseText(content)
         if (scan.labels.length) {
@@ -87,7 +87,7 @@ export const getResponseFromText = async (
     }
 
     // If we still don't have a response config, we can match all regexes after the initial label trigger
-    if (!responseConfig.triggers && ocrMode) {
+    if (!responseConfig.triggers && flags.onlyImageTriggers) {
         logger.debug('No match from NLP, doing after regexes')
         for (let i = 0; i < responses.length; i++) {
             const {