[Bug]: Model picker list stops rendering before reaching Ollama Cloud (~30th of 54 providers), no crash shown to user #582

Affected area

Android app — model picker (/model selector, same list that showed the DeepSeek duplicate-key crash in #309)

What happened?

Opening the model picker on a session with many authenticated providers, the visible list renders providers in order (Anthropic, Nous Portal, OpenRouter, Mixture of Agents, ChatGPT/Codex Subscription, xAI Grok OAuth, GitHub Copilot, ...) and then simply stops — it never reaches Ollama Cloud, even though Ollama Cloud has valid credentials and models. No crash dialog is shown to the user; the list just ends short.

Root cause confirmed NOT to be backend/API

Called the exact same endpoint the app consumes directly:

GET /api/model/options  (api_server, :8642)

The payload correctly returns 54 provider rows including:

ollama-cloud  Ollama Cloud  24 models

at position 30 of 54. So the API/backend picker payload (hermes_cli.inventory.build_model_options_payload → list_authenticated_providers) is correct and complete — this is purely a client-side rendering issue in the Android app's model picker list.

Suspected relation to #309

#309 (closed) was an IllegalArgumentException: Key "DeepSeek:deepseek-v4-flash" was already used crash in the same LazyColumn-based model picker, caused by a duplicate composable key. Given the picker renders a flat list across many provider sections (some providers legitimately share model IDs across catalogs — e.g. glm-5.3-flash" appears both under ollama-cloudand underz-ai/openrouter` aggregator rows), it's plausible a similar duplicate-key collision is silently truncating the LazyColumn's rendering (or throwing and being swallowed) somewhere around section ~7-8, before Ollama Cloud's section (~30th row) is reached — but this time with no visible crash dialog, unlike #309.

Environment

• Backend: Hermes Agent v0.21.2 (self-hosted Mac mini, git install)
• Confirmed via direct /api/model/options call against the exact backend the app was paired to
• Provider count: 54 total, 11 with total_models > 0

Expected

Every authenticated provider with models should render in the picker, including Ollama Cloud and everything after it in payload order (opencode-zen, opencode-free, etc.).

Suggested places to look

• The model picker's LazyColumn/list composable — check for a key derivation that could collide across provider sections with overlapping model names (same pattern as [Bug]: Crash — IllegalArgumentException: Key "DeepSeek:deepseek-v4-flash" was already used. If you are us #309's fix, if it was a key-uniqueness fix rather than a data fix)
• Confirm whether the crash is happening silently (caught/swallowed exception) vs. the list is being truncated by a size/pagination limit client-side