Nothing happens after Analyze with AI

[marco-brandizi]

❓ Question

Hi all. I'm trying this plug-in, I followed the instructions, installed my Gemini API key, then I selected a few emails and right-clicked on "Analyze with AI". Nothing happened after that. I did it a couple of times.

🔧 Environment

Thunderbird

• Version:149.0.2 (aarch64)
• OS: macOS 26.4.1 (25E253)

Ollama

• Version: NA
• Model: NA

AutoSort+

• Version: 1.2.3.2

📝 What I've Already Tried

• Checked the documentation
• [NA] Tested Ollama directly: ollama run model "test"
• Ran: curl http://localhost:11434/api/tags
• Checked browser console (Ctrl+Shift+J)
• Searched existing issues

🎯 Additional Context

Provide any relevant details, screenshots, or code examples.

---

⚡ Quick Troubleshooting

For Ollama setup questions:

1. Is Ollama running? ollama serve
2. Is model installed? ollama list
3. Can you chat with it? ollama run tinyllama "test"

For AutoSort+ questions:

1. Settings → Provider: Ollama selected?
2. Test Connection passing?
3. Check console logs during analysis (Ctrl+Shift+J)

For Thunderbird API questions:

• Version compatibility needed?
• Manifest permissions set?
• Tested in developer mode?

[Nigel1992]

Hi,

Thanks for your report!

Before sharing the logs, could you please try the latest pre-release version first to see if it resolves the issue?
You can download it here: v1.2.3.4 Pre-release

If the problem persists, please open the Browser Console in Thunderbird and post the log contents here:

How to open the Browser Console:

1. Go to the Menu (the three horizontal lines) in the top right corner.
2. Navigate to Tools > Developer Tools.
3. Select Browser Console (or press Ctrl+Shift+J / Cmd+Shift+J on Mac).
4. Perform the action that is causing the issue.
5. Copy the logs and paste them here.

Important

Privacy Note: Before posting, please review the logs and remove or mask any sensitive information. Make sure to hide:

• Authentication tokens or API keys.
• Full email addresses or passwords.
• Personal server addresses or private IP addresses.

This will help us see exactly what's happening under the hood while keeping your account secure!

[marco-brandizi]

Thanks @Nigel1992. I can see what the problem is from the console log:

Error analyzing email: Error: All 1 Gemini API keys have reached their daily limit (20/day each). Please wait for reset or add more API keys in settings.

I'll try ollama. But these errors should be reported to the end user, only the experts can figure out using the dev tools.

Moreover, I am already paying for a github/copilot account and supporting it would be great, I can't afford all these AI services.

[marco-brandizi]

@Nigel1992 I've configured ollama, it seems to be using it, but now it gives new errors, see the attached log. Thanks.

[Nigel1992]

Thanks — I wasn’t aware of the new public preview for the Copilot SDK; I’ll implement support for it in the extension soon.

Thanks for the logs — very helpful. Quick summary and next steps:

• Gemini quota: you’re right — those quota errors were previously only logged to the console and shown as transient notifications, so end users often miss them. I’ve updated the extension to persist the latest error and show a dismissible banner in Settings → AI Settings so users will see the problem without opening dev tools.

• Ollama logs: the injected tab received a non-JSON response and the code called response.json(), which caused JSON.parse: unexpected character. Likely causes:

  • The configured Ollama Server URL points to the UI or a proxy that returns HTML, not the API endpoint.
  • The Ollama instance returned an HTML error page (or auth page) instead of JSON.
  • The server requires an auth token and returned an auth/error page.

Quick checks you can run locally (replace the port/url if needed):

curl -v http://localhost:61837/api/tags
curl -v -X POST http://localhost:61837/api/chat \
  -H "Content-Type: application/json" \
  -d '{"model":"qwen3.5:9b","messages":[{"role":"user","content":"ping"}],"stream":false}'

If you can paste the curl output (or the HTML/error body), that will help pinpoint whether it’s a server-side issue or a small client-side safeguard we can add right away.

What I’ll do next:

• Add Copilot SDK support as requested.
• Update the Ollama tab-injection flow to fall back to response.text() when response.json() fails and show the raw response in the UI/notifications so we don’t throw and so the error is visible to end users.
• Follow up with a small PR/patch and instructions here when ready.

Thanks — I’ll update this issue once Copilot SDK support and the Ollama fix are published.

— Nigel

[Nigel1992]

Pre-release test build with the Ollama tab-injection fallback and improved error visibility.
Download: https://github.com/Nigel1992/AutoSort-Plus/releases/download/v1.2.4-pre1/autosortplus-v1.2.4-pre1.xpi

This is a pre-release test build — please try the Ollama flow again and paste any server response if errors persist.