Generate, Review, and Publish Knowledge Base Article Suggestions from Videos, Session Replays, Slack, and APIs
This article explains how to use Brainfish to generate knowledge base article suggestions from multiple sources (uploaded videos, session replays, Slack threads, and external integrations via the Suggestions API) and review them in a single, unified Review Screen. It outlines the step-by-step process for uploading a video or queuing a session for discovery, describes what happens during and after processing, shows how to review the generated drafts and their structure (including video snippets), and explains how to accept (publish) or reject suggestions.
Overview
Brainfish can automatically generate knowledge base article suggestions from multiple sources—then route all drafts into a single, unified Review Screen where you can edit, accept (publish), or reject them.
Sources that can generate suggestions:
- Uploaded videos
- Session recordings (via Discover)
- Slack threads (via @Brainfish tagging)
- Suggestions API (external integrations)
How the Suggestion Pipeline Works (New Article vs. Precise Edit Suggestions)
When Brainfish discovers useful knowledge from an input source (for example, a document, transcript, video, or message thread), it runs a high-level workflow:
- Discovery extracts useful knowledge from a source
- The system scans the input and identifies information that is useful for the knowledge base.
- The system classifies the extracted knowledge
- It determines whether the extracted content should result in:
- A new knowledge article, or
- An update to an existing knowledge article (delivered as precise edit suggestions)
- It determines whether the extracted content should result in:
Output types
- New article: Created when the extracted knowledge does not map well to existing content.
- Precise edit suggestions (individual suggestions): Produced when the extracted knowledge aligns with an existing knowledge article and is best applied as incremental improvements.
What is an “individual suggestion”?
An individual suggestion refers to the “suggest precise edits” experience for existing knowledge articles. Instead of creating a brand-new article, the system proposes targeted changes (additions, edits, or clarifications) to improve or update content that already exists.
Note: This article describes the product’s high-level behavior; details such as matching logic, thresholds, or all supported source types may vary by account and are not covered here.
The Unified Review Screen
All AI-generated article suggestions are consolidated into one review queue. This replaces older, per-source review workflows.
Access the Review Screen
- Navigate to Knowledge from the main application menu.
- Look for the notification banner indicating Articles are waiting for your review.
- Click Review to open the unified review queue.
- Suggestions appear in reverse chronological order.
- Each suggestion displays its source (for example: Video Upload, Session Recording, Slack, API).
Review and act on suggestions
For each suggestion in the queue:
- Read the proposed content.
- Optionally edit the content before publishing/applying it.
- Take the appropriate action (for example, accept/publish a new article draft, or apply edits to update an existing article).
- The queue automatically advances to the next suggestion.
Note: The Review Screen is currently being rolled out to all customers. If you don’t see it yet, contact hello@brainfi.sh.
Update an existing article vs create a new article (what to expect)
When generating knowledge from video uploads (and other supported sources), Brainfish follows a suggestion-first workflow.
- There is no explicit option at upload/submission time to choose “update an existing article” versus “create a new article.”
- Brainfish generates suggestions first, then you decide what to do during review.
- Depending on how well the discovered knowledge matches your existing KB, you may see:
- A new article draft, intended to be published as a new KB article, or
- A precise edit suggestion (individual suggestion), intended to update an existing KB article with targeted changes.
Generate Knowledge Base Articles from Uploaded Videos
How to upload videos for article generation
-
Navigate to Knowledge.
-
Select Articles (Managed CMS).
-
Select any collection.
-
Click Upload a file or drag and drop a video to the file upload box.
-
Once the video is uploaded, processing the video will automatically take place.
:::info Navigating away from the page will not affect video processing.
:::
- Once the video is finished processing, one or more article suggestions will appear in the unified Review Screen (the number depends on the video contents).
Video requirements and limits
- Supported format:
MP4 - Maximum video duration (for generating articles): 15 minutes
- This limit can be increased by request to the Brainfish team.
- Maximum file size (for videos embedded in articles): 500 MB
Tips for best results
- Compress videos before uploading to speed up processing.
- Use a resolution of 720p or higher for better output quality.
How the video analysis works
When you upload a video, the AI performs a deep analysis using both computer vision and audio processing:
- Watching: It uses computer vision to identify what is happening on the screen (e.g., button clicks, menu navigation, UI changes).
- Listening: It transcribes and analyzes the voiceover or spoken audio to understand the intent and context behind the actions.
This combination helps ensure that the resulting documentation accurately reflects the product functionality demonstrated in the video.
What gets generated
After analysis, Brainfish generates draft articles that may include:
- Overview of the topic
- Step-by-step instructions (numbered actions)
- Summary recap
- Video snippets embedded and cued to the most relevant segments for each section
Reviewing Generated Articles and Video Snippets
https://www.youtube.com/watch?v=RShaIIu7xls&t=17%E2%80%8B
Summary
Once the video analysis is complete, the platform generates draft articles. This section covers how to review these drafts, understand their structure, and utilize the integrated video snippet feature.
Reviewing Generated Content
After processing, the system presents a list of draft articles based on the distinct topics detected in your video.
- Select a draft: Click on a title from the list of generated articles (e.g., "The Improvement Engine").
- Check the structure: The AI automatically organizes content into logical sections, such as:
- Overview: A high-level explanation of the topic.
- Step-by-Step Instructions: Numbered actions for the user to follow.
- Summary: A concluding recap.
Video Snippets
A key feature of these articles is the embedded video player.
- Contextual Playback: The article includes a video player that is automatically cued to the specific segment (snippet) relevant to that article.
- Verification: This allows users to see exactly where the information came from.
- Usage: Users can choose to read the text or watch the short clip, which is particularly useful when these answers are surfaced via in-app assistants or Slack integrations.
Generate Knowledge Base Articles from Session Replays
CleanShot 2025-12-08 at 08.52.09.mp4 1920x1080
Summary
This section demonstrates how to use the Discover feature within a session replay to analyze user actions and automatically generate new knowledge base items.
When to Use This
Use this workflow when you identify a session that contains valuable user interactions or questions that should be documented. This allows the system to learn from the session and suggest relevant knowledge base articles.
Step-by-Step Instructions
- You can find Session Replays in the Knowledge > External data tab if you have observe and learn turned on in a widget from the Distribution screen
- While viewing a recording in the Session Replay page, you can open one to determine if the content is valuable for training.
- Click the green Discover button located in the top right corner of the player.
- Navigate to the Knowledge dashboard from the main menu.
- Find the processed session in the unified Review Screen (hovering over the item will indicate that its source is a Session Recording).
Generate Knowledge Base Article Suggestions from Slack
If your workspace has Brainfish’s Slack integration enabled, Brainfish can generate article suggestions from relevant Slack threads.
- In Slack, tag @Brainfish on a message or within a thread you want to turn into documentation.
- The resulting suggestion will appear in the unified Review Screen with Slack as its source.
Generate Knowledge Base Article Suggestions via the Suggestions API
Teams can generate article suggestions from external systems (for example, support tools, forms, or internal apps) via the Suggestions API.
- When an integration submits a suggestion via the API, it will appear in the unified Review Screen with API as its source.
FAQs
Q: Can I choose, at upload time, whether Brainfish updates an existing article vs creates a new one? A: Not currently. Brainfish generates suggestions first and routes them to the Review Screen; depending on the match, you may receive a new article draft or a precise edit suggestion for an existing article.
Q: What happens to the videos after upload? A: The videos will be saved in storage together with your other media uploads. They will be kept as reference for the generated articles.
Q: How long will video processing last? A: Video processing time will depend on how long the video is.
Q: What’s the maximum video duration I can upload? A: Currently, you cannot upload a video longer than 15 minutes, however this is able to be increased upon request to the Brainfish team.
Q: What video formats are supported?
A: MP4.
Q: Is there a maximum file size for video uploads? A: When embedding videos in articles, the maximum file size is 500 MB.
Q: Are videos compressed? A: Videos are not compressed before they are processed; however, we recommend compressing files to speed up upload.
Q: How many articles are created per video? A: Every video will generate at least one article. Additional articles may be created automatically to capture multiple topics.
Q: What are the required dimensions? A: There are no set dimensions for your video uploads, but we recommend 720p and above for best results.
Q: Can I edit a suggestion before accepting it? A: Yes. You can edit any suggestion in the review queue before publishing it to the knowledge base.