FreeInference chat and embeddings
- Add FreeInference to the provider picker and key import flow.
- Support tested Qwen3.6 35B chat, including streaming, and BGE-M3 embeddings with 1,024-dimensional vectors.
- Validate upstream model identities and embedding shapes; preserve normal routing errors and failover.
- Avoid treating FreeInference's public model roster as API-key validation. Real requests establish key health.
The signed catalog adds two entries without removing any existing models. Premium receives the new catalog immediately; Free receives these entries through the unchanged 30-day release gate. Update the app or stable Docker image to use the new adapter.
FreeInference offers a shared daily research allowance. It logs prompts and responses for research, so do not send sensitive data.
Verification: live normal/streaming chat and embedding calls through the app adapter, full regression tests, type checks, migrations, desktop bundles, and staged day-0/day-29/day-30 catalog checks.