Files
nofx/api
tinkle-community 2f8c6111b1 feat: restrict AI models to claw402 gateway, default GPT-5.6 Sol
- Offer only claw402 (x402 USDC) models: remove direct-API providers
  (deepseek/qwen/openai/claude/gemini/grok/kimi/minimax) and BlockRun
  from supported models, defaults, and the config UI
- Sync model catalog with the claw402 catalog (GET /api/v1/catalog):
  GPT-5.6 Sol/Terra/Luna, Claude Fable 5 / Opus 4.8, DeepSeek-V4
  Flash/Pro, GLM-5 — official naming, brand icons, verified prices
- Default model gpt-5.6 (Sol) across backend and frontend
- fix: omit temperature for GPT-5/o-series reasoning models; they
  reject non-default values and fail the whole request
- fix: hot-reload running traders when a model config is saved under
  its provider key (previously required a restart to take effect)
- fix: allow saving model config edits without re-entering the stored
  key; backend preserves the existing key when api_key is empty
- fix: record AI charges with the effective model name so per-call
  cost tracking matches the invoked model and price
2026-07-28 18:16:43 +09:00
..
2026-04-25 16:18:45 +08:00