wehub-resource-sync
23 lines
1.2 KiB
TOML
23 lines
1.2 KiB
TOML
name = "Venice AI"
|
|
env = ["VENICE_API_KEY"]
|
|
npm = "venice-ai-sdk-provider"
|
|
# Raw HTTP reasoning controls (sources accessed 2026-06-25):
|
|
# - POST /api/v1/chat/completions accepts either
|
|
# `reasoning = { effort = "..." }` or top-level `reasoning_effort`; the flat
|
|
# field wins if both are sent. `reasoning = { enabled = false }` is Venice's
|
|
# preferred toggle, while supported legacy thinking models also accept
|
|
# `venice_parameters = { disable_thinking = true }`.
|
|
# - Alpha POST /api/v1/responses is stateless and accepts only nested
|
|
# `reasoning = { effort = "..." }`; E2EE models must use Chat instead.
|
|
# https://docs.venice.ai/guides/features/reasoning-models
|
|
# https://docs.venice.ai/swagger.yaml
|
|
# Model-specific effort enums come from live GET /api/v1/models fields
|
|
# `supportsReasoningEffort` and `reasoningEffortOptions`. Venice exposes no
|
|
# dedicated reasoning-token budget: `max_completion_tokens` caps reasoning plus
|
|
# visible output, and `model_spec.maxCompletionTokens` supplies that total cap.
|
|
# https://api.venice.ai/api/v1/models?type=text
|
|
# https://docs.venice.ai/api-reference/endpoint/models/list
|
|
# https://docs.venice.ai/api-reference/endpoint/chat/completions
|
|
doc = "https://docs.venice.ai"
|
|
# api = "https://api.venice.ai/api/v1"
|