InputBuffer routes incoming feedback into named buffers. You define what belongs in each buffer, and the AI reads that definition to sort every new feedback item without you touching it.

Help me design InputBuffer buffers for this product.

Buffers are named topic groups. Every piece of incoming user feedback is read by an AI and
routed into the buffer whose description and routing instructions fit it best.

Read this codebase and figure out what users of this product are most likely to complain
about or ask for. Then propose 3 to 5 buffers. For each one give me:

- Name: a short label, such as "Bug Reports" or "Billing".
- Description: one sentence explaining what belongs in it. The AI reads this to route.
- Routing instructions: explicit plain-English rules, including what does NOT belong.
  Example: "Route here only when the user reports a crash, error, or unexpected behavior.
  Not for feature requests."

Constraints:
- Keep the buffers non-overlapping. Two buffers that could both plausibly accept the same
  feedback produce low confidence scores and land the feedback in a manual review queue.
- Prefer fewer, clearer buffers over many narrow ones.
- Base the names on what this product actually does. Don't give me generic categories.

Output them as a table I can work from. I'll create them at Settings, then Buffers,
in the InputBuffer dashboard.

Buffers guide: https://inputbuffer.io/docs/buffers

Report anything about InputBuffer that gets in your way, at any point while we are working with
it: initial setup, and the styling, configuration, and debugging that follow. Include the times you
had to correct your own understanding of InputBuffer, such as a request field the API ignored, a
widget attribute that had no effect, or an error message read from the wrong key. A correction
unrelated to InputBuffer does not qualify. Show me the text, and send it once I agree:

curl -X POST https://inputbuffer.io/api/v0/feedback \
  -H "Authorization: Bearer ibw_96e7f571-64ad-494a-a125-afed734d3fdc_op1CeDTiptGa8fWRU25tKcgx6uubJ2wOlWvM0JbpY8Q" \
  -H "Content-Type: application/json" \
  -H "X-IB-Client: your-tool/version (your-model)" \
  -d '{ "description": "What you were trying to do, what happened instead, what would have helped" }'

That token is InputBuffer's own and accepts nothing but feedback, so there is nothing to set up.
Leave out "submitted_by". Send one report per problem, not one per attempt.

1. 1

   Navigate to Buffers

   Open your organization dashboard and click Buffers in the left sidebar.

2. 2

   Create your first buffer

   Click Add Buffer in the top right. Fill in the fields:

   • Name: a short label for this category. Examples: Bug Reports, Feature Requests, Billing, Performance.
   • Description (optional, but worth writing): one sentence explaining what this buffer is for. The AI uses the description to decide what belongs here.

   Click Create Buffer. The buffer appears in your list.

   Buffers

   Bug Reports

   Things that are broken or not working as expected.

   0 feedback items

   Routing instructions are added after the buffer exists. Click the pencil icon on the buffer's row to open them, and see Routing instructions below for how to write them.

3. 3

   Submit feedback and let the AI route it

   Once you have at least one buffer, every new feedback item that arrives is analyzed by the AI classifier. It reads the title and description, compares them against all of your buffers' names, descriptions, and routing instructions, and makes a routing decision.

   What happens to each feedback item:

   • At 80% confidence or higher, InputBuffer assigns the feedback to the matching buffer. Nothing for you to do.
   • Between 50% and 80%, the feedback needs your review. It appears on the Review page, where you can confirm the AI's choice, or open it and assign it manually.
   • Below 50%, the feedback stays unclassified. You can assign it manually from the feedback detail page.

   When auto-create buffers is enabled, the AI can also suggest a new buffer if no existing one fits well. It will only do this when its confidence in the new buffer exceeds the buffer creation threshold (default 40%). All three thresholds are adjustable in Settings → AI Settings.

   Note·

   When you're just starting out, aim for 3–5 buffers with clear, non-overlapping purposes. The AI routes best when each buffer has a distinct meaning. Overlapping buffers (e.g., "Bugs" and "App Bugs") lead to low confidence and more manual review.

4. 4

   View feedback inside a buffer

   Click a buffer's name in the list to open its detail page. You'll see all feedback routed here, sorted newest first.

   Buffers

   Bug Reports

   Things that are broken or not working as expected.

   Routing: Route here only when the user reports a crash, error, or unexpected behavior.

   12 feedback items

   Nothing happens when I click download

   I click the download button on my invoice and nothing happens. No file, no error.

   Mar 1

   Export times out on large projects

   Exporting anything over a few thousand rows spins for a minute and then fails.

   Feb 27

   From here you can read individual feedback items, add internal notes, and track how feedback in this category evolves over time.

Routing instructions

Routing instructions are the most direct way to control what the AI puts in a buffer. Write them as plain-English rules describing what belongs, and optionally what does not.

Examples:

• Route here only when the user reports a crash, error, or unexpected behavior.
• Billing and payment issues only. Not for login problems or account setup.
• Feature requests and suggestions for new functionality. Not for bugs.
• Performance complaints: slowness, timeouts, loading delays. Not for errors.

Tips:

• Use "only when" and "not for". The exclusions matter as much as the inclusions.
• Mention keywords users actually write: "crash", "slow", "billing", "can't login".
• Keep instructions to one to three sentences, so the rule stays easy to read and easy to revise.

Reviewing low-confidence classifications

When the AI's confidence falls between 50% and 80%, the feedback lands on the Review page instead of being assigned automatically. Click Review in the sidebar to see it.

For each item you'll see the AI's suggested buffer and its confidence score. You can:

• Mark reviewed: accept the AI's routing decision.
• Leave it for now and come back later.

Your recent corrections are included as worked examples in the next classification prompt, so the AI sees how you've been routing this kind of feedback and follows suit.

AI-suggested buffers

As your feedback accumulates, the AI may suggest new buffers for patterns it notices that don't fit your existing categories. Suggestions appear at the top of the Buffers page under an AI suggested heading, as dashed cards.

Each draft shows:

• The suggested buffer name
• The AI's reason for suggesting it
• How many feedback items are already associated with it

For each draft you can:

• Approve: the draft becomes an active buffer and starts receiving new feedback.
• Delete: the suggestion is set aside, and the AI won't propose that name again. Feedback the AI had associated with it becomes unclassified.

Deleting a suggestion is not permanent. Rejected suggestions collect in a rejected suggestions section further down the Buffers page, and each one has a Restore button that brings it back as a draft.

AI-suggested buffers also appear on the Review page if you prefer to manage all pending items in one place.

Editing a buffer

To change a buffer's name, description, or routing instructions, click the pencil icon on the right side of any buffer row in the list. This opens an edit modal with the same fields as creation, plus an icon picker.

Changes take effect right away for new feedback. Feedback already routed to the buffer is not re-classified.

Next steps

• Feedback Widget: collect feedback directly from your web app or docs site, with no backend required
• API getting started: authenticate, then browse the reference for every endpoint, parameter, and response schema