Skip to content

Add NEAR AI provider and DeepSeek V4.1 Flash Abliterated - #192

Open
0xadvait wants to merge 1 commit into
OpenGradient:mainfrom
0xadvait:advait/nearai-provider
Open

0xadvait wants to merge 1 commit into
OpenGradient:mainfrom
0xadvait:advait/nearai-provider

Conversation

@0xadvait

@0xadvait 0xadvait commented Nov 4, 2026

Copy link
Copy Markdown
Contributor

Adds NEAR AI Cloud (OpenAI-compatible, cloud-api.near.ai/v1) as a provider, with one model: deepseek-v4.1-flash-abliterated -> deepseek-ai/DeepSeek-V4.1-Flash-1017-abliterated.

  • Pricing: $0.30 in / $1.20 out per MTok, from cloud-api.near.ai/v1/models.
  • Parameters: NEAR's listing names no sampling params, but temperature: 0, tools, streamed usage and image input all work on the live API. reasoning_effort: "medium" is a 400 on this model (it takes none/low/high/xhigh/max), so it is sent as high, like the Kimi mapping.
  • Key: NEARAI_API_KEY in the host .env, injected as nearai_api_key by run-enclave.sh.

Tests: CI's pytest command (522 passed) and make lint clean. Also streamed through get_chat_model_cached against the live API with medium, none and an image turn.

DeepSeek V4.1 Flash with refusals removed, on NEAR AI Cloud's OpenAI-compatible
API: $0.30 in / $1.20 out per MTok, 1M context, tools, images. NEAR's V4.1
rejects reasoning_effort "medium", so it is sent as "high".
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant