AI Settings
Configure Gemini-backed AI and validate connectivity for the v2 UI.
AI configuration status
The backend AI stack (Gemini + ATOMIC kernel) is configured entirely via environment variables on the Dev VM. Use this page to see current status and test candidate keys and models without persisting secrets.
Reads GOOGLE_GEMINI_API_KEY inside the backend container. Managed via the VM's .env, not this UI.
Resolved from GOOGLE_GEMINI_MODEL when set; otherwise defaults to gemini-2.5-flash.
When configured, the same Gemini stack powers both the floating AI chat widget and AI Reports page.
Gemini API key
Use this section to quickly test a Gemini API key and model combination. Keys are never stored; they are used only for the one-off validation call.
- Visit Google AI Studio
- Click "Get API key" and choose a project
- Copy the generated key
- Configure it as
GOOGLE_GEMINI_API_KEYon Dev VM 220
This field is used only to call the /api/v1/ai/test endpoint. It is never written to disk.
If left blank, the backend will fall back to its default model (typically gemini-2.5-flash).
Model comparison
For most portfolio analysis workflows, Gemini 2.5 Flash is the recommended choice: fast, cost-efficient, and capable of handling long context. Pro variants are ideal for the heaviest reasoning workloads or narrative-heavy reports.
- 2.5 Flash: primary model for chat + AI Reports. Great balance of speed and quality.
- 2.5 Pro: most capable model, best for deep analysis and complex “what-if” questions.
- 3 Pro Preview: early access to next-gen capabilities. Expect occasional rough edges.
- Flash/Pro Latest: tracks Google's latest releases without code changes.
Currently testing with gemini-2.5-flash (Gemini 2.5 Flash (Recommended)).