Docs: image and PDF use cases + misc touchups

[anyacherniss]

Closes: #210
Closes: #322
Closes: #432

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Preview	Comments	Updated (UTC)
anotherai	Ready	Preview	Comment	Oct 7, 2025 5:33pm
anotherai-docs	Ready	Preview	Comment	Oct 7, 2025 5:33pm

[pierrevalade]

@anyacherniss does this use-case works?
is Claude/Cursor able to fetch the image via their url to understand what the image is about?

[anyacherniss]

It appears so yes, I asked Claude to analyze an image from https://anotherai.dev/experiments/01999b76-0e5c-7086-4849-3ef9e28f0ff6, which reused inputs from https://anotherai.dev/experiments/01999b4a-5300-73c2-9ff7-a7e75094c124.

The description initially had some errors: initially it said 4 donuts, not three, and I made up asking about a "green thing" to see if it would actually note that it doesn't exist. After prompting it to recheck, it appears it can view and analyze the image.

This is Sonnet 4.5

[anyacherniss]

@claude can you review and suggest any improvements that either:

• increase the readability of the changes, particularly for new/prospective users who have minimal context about the product
• fix technical inaccuracies

[claude]

Claude finished @anyacherniss's task —— View job

---

PR Review: Documentation Improvements for Image and PDF Use Cases

I've reviewed all the documentation changes in this PR. The overall structure and content are strong, but I've found several areas where readability can be improved for new/prospective users and a few technical inaccuracies that should be addressed.

✅ Strengths

• Excellent step-by-step guides for both image and PDF processing agents
• Good use of callouts and accordions to organize information
• Helpful visual examples with screenshots
• Clear workflow for experimentation and optimization

🔍 Issues Found

1. Missing Space in Prompt Examples (image-agents.mdx:41, pdf-processing-agents.mdx:41)

Create a new AnotherAI agent that provides a list of detected foods and their approximate calorie count from a given imagewhere the average response time is less than 4 seconds

Missing space between "image" and "where". Should be:

Create a new AnotherAI agent that provides a list of detected foods and their approximate calorie count from a given image where the average response time is less than 4 seconds

2. Inconsistent Capitalization (Both files)

Line 58: "User identication" → "User identification"

3. Confusing Wording for Dataset Structure (image-agents.mdx:101-103, pdf-processing-agents.mdx:101-103)

The current wording is unclear for new users:

"A dataset is a collection of images in a file format that your AI assistant can understand. Datasets can contain expected outputs, but do not need to."

Suggested rewrite:

"A dataset is a JSON file containing URLs or file paths to your images/PDFs along with optional expected outputs. This allows you to test multiple inputs at once."

4. Technical Inaccuracy in Callout (image-agents.mdx:144-156)

The callout about file path references is partially incorrect based on the review comments. Anya noted that dragged and dropped PDFs in Claude Code do include file paths. Consider clarifying or removing this callout to avoid confusion.

5. Grammatical Error (Both files)

Line 97: "There a couple ways to do this" → "There are a couple ways to do this"

6. Unclear JSON Structure (image-agents.mdx:124, 130, 135)

The JSON examples have trailing commas that make them invalid:

{
  "image_url": "https://example.com/images/photo1.jpg",
}

Should be:

{
  "image_url": "https://example.com/images/photo1.jpg"
}

7. Duplicate Content Issue

The PDF processing document appears to be largely copied from the image agents document with minimal changes. While the structure is appropriate, consider:

• Adding PDF-specific considerations (e.g., multi-page handling, text extraction vs visual analysis)
• Including different examples that showcase PDF-specific use cases
• Mentioning any limitations or differences in PDF processing vs image processing

8. Missing Context for New Users (user-feedback.mdx:33-45)

The example jumps into asking the AI assistant to create a function without explaining:

• What the annotations API is
• How completion IDs are obtained
• What the API key is and where to get it

Consider adding a brief introduction or linking to relevant documentation.

9. Inconsistent Image References

Some image references use generic placeholder names (e.g., /images/add-annotation-completion-output.png) which may not exist. Ensure all image paths are correct and images are included in the PR.

💡 Suggestions for Improvement

1. Add a Brief Introduction: Both the image and PDF guides could benefit from a 2-3 sentence introduction explaining what these agents do and common use cases before diving into the tutorial.

2. Clarify Local vs URL Inputs: Add a clear section explaining when to use URLs vs local files, and any limitations of each approach.

3. Include Troubleshooting Section: Consider adding common issues users might encounter (e.g., image size limits, supported formats, timeout issues).

4. Link Between Related Docs: Add cross-references between the image and PDF processing guides since they share similar workflows.

5. Real-World Examples: Include more concrete examples of what users might build (e.g., "receipt scanner", "document classifier", "product image analyzer").

✅ Todo List

• Examine all changed documentation files
• Review image-agents.mdx for readability and accuracy
• Review pdf-processing-agents.mdx for readability and accuracy
• Review other modified documentation files
• Compile and provide comprehensive feedback

The documentation additions are valuable and well-structured overall. Addressing these issues will make the guides more accessible to new users and technically accurate.

---

Branch • Job run

[pierrevalade]

how does "upload a PDF to a AI assistant" works? is there a screenshot or video that we can show?

[anyacherniss]

I can record a video, yes. Adding as a todo for later in the process - want to get the text part polished first

[anyacherniss]

Added the video in 4a997d1. Due to edits I made to the content in this section, I chose to put the video a bit lower on the page. I feel strongly that placing it under this text is not the best place for it, given the additional content I've added since you reviewed.