GMI model 'deepseek-ai/DeepSeek-V4-Flash' unavailable — summarization fallback needed

[graycyrus]

Summary

The summarization pipeline fails when the configured GMI model deepseek-ai/DeepSeek-V4-Flash is not available for the user's organization, with no fallback.

Sentry Issue

• OPENHUMAN-TAURI-CC — 1 event

Error: OpenHuman API error (404 Not Found): {"success":false,"error":"GMI model 'deepseek-ai/DeepSeek-V4-Flash' is not available for your organization. Set GMI_SUMMARIZATION_MODEL_ID to a model ID returned by GET /gmi/v1/models."}
Environment: staging
Affected platform: macOS (MacBookAir7,2, x86_64, Bengaluru, India)

Analysis

• The error message from the API is clear: the model isn't provisioned for the org
• The API suggests setting GMI_SUMMARIZATION_MODEL_ID — but this env var is not accessible to end users
• This blocks the entire summarization flow for affected users

Suggested Fix

1. Auto-discover available models — on startup or first summarization call, query GET /gmi/v1/models and select the best available summarization model
2. Fallback chain — if the configured model isn't available, try a default list of known summarization models before failing
3. Surface the config option — if GMI_SUMMARIZATION_MODEL_ID is user-configurable, expose it in Settings; if not, handle it server-side

Priority: High