Disable safety checking for Gemini

[xl0]

Gemini is quite over-zealous with the default safety settings, triggering on completely benign tasks.
Ideally, we should make this configurable, but until someone implements it, let's just set the safety thresholds to the max.

---

🚀	This description was created by Ellipsis for commit e993889

Summary:

This PR updates the POST function in app/api/chat/google/route.ts to disable safety checks by setting all harm block thresholds to BLOCK_NONE for the Google Generative AI model.

Key points:

• Updated POST function in app/api/chat/google/route.ts
• Imported HarmBlockThreshold, HarmCategory, SafetySetting from @google/generative-ai
• Set all safety thresholds to BLOCK_NONE for various categories
• Configured GoogleGenerativeAI model with new safety settings

---

Generated with ❤️ by ellipsis.dev

[ellipsis-dev]

❌ Changes requested. Reviewed everything up to e993889 in 29 seconds

More details

• Looked at 50 lines of code in 1 files
• Skipped 0 files when reviewing.
• Skipped posting 0 drafted comments based on config settings.

Workflow ID: wflow_JbqOXGi0VZt4uCVW

---

Want Ellipsis to fix these issues? Tag @ellipsis-dev in a comment. You can customize Ellipsis with 👍 / 👎 feedback, review rules, user-specific overrides, quiet mode, and more.

[xl0]

@mckaywrigley ping

[samrahimi]

Yes, this and the addition of the gemini-pro-1.5-exp-0801 update should both be merged... I should have checked the PRs before I implemented both myself last night. Together they make gemini into an uncensored frontier model, which is a big deal (the 0801 update actually loses its guardrails near completely when you turn off the safety checked and provide an appropriate system message... plus its performance is currently topping the leaderboards, a huge improvement over the status quo)