Skip to content

Add PRXVT AI provider#2635

Open
PRXVT wants to merge 2 commits into
anomalyco:devfrom
PRXVT:add-prxvt-ai-provider
Open

Add PRXVT AI provider#2635
PRXVT wants to merge 2 commits into
anomalyco:devfrom
PRXVT:add-prxvt-ai-provider

Conversation

@PRXVT

@PRXVT PRXVT commented Jun 17, 2026

Copy link
Copy Markdown

Add PRXVT AI provider

Adds PRXVT AI, an OpenAI-compatible inference gateway, as a new provider.

  • SDK / endpoint: @ai-sdk/openai-compatible against https://api.prxvt.ai/v1
    (OpenAI Chat Completions shape, e.g. POST /v1/chat/completions)
  • Auth: PRXVT_API_KEYAuthorization: Bearer prxvt-sk-…
  • Logo: monochrome currentColor SVG, transparent background, centered viewBox

Models

20 served models, with names, context windows, modalities, and pricing.
This includes PRXVT MoA, a mixture-of-agents model that routes across frontier models and synthesizes a
final answer.

Validation

All provider and model configs validate against the core schema and the base_model merge logic (bun validate
equivalent) provider valid, all 20 models valid.

PRXVT added 2 commits June 17, 2026 02:31
PRXVT AI (https://prxvt.ai) is an OpenAI-compatible inference gateway.
Endpoints follow the OpenAI Chat Completions shape
(https://api.prxvt.ai/v1/chat/completions) with bearer-token auth.

- provider.toml: @ai-sdk/openai-compatible, api https://api.prxvt.ai/v1,
  env PRXVT_API_KEY
- logo.svg: monochrome currentColor mark, transparent background, tightly
  centered viewBox (matches existing provider logo conventions)
- 20 served models, with pricing/context taken from the live
  GET /v1/models catalog, including PRXVT MoA (mixture-of-agents over
  frontier models, synthesized by Claude Opus 4.8)
- Wrapped models reuse provider-agnostic facts via base_model and only
  declare PRXVT-specific cost; Hermes 4 405B, DeepSeek V3.1, and PRXVT MoA
  are defined in full since no shared models/ metadata exists for them
- Reasoning models that stream interleaved thinking (deepseek-v3.1,
  deepseek-v4-flash, gemini-3.1-pro, kimi-k2.6, hy3-preview) declare
  reasoning_options effort levels and interleaved reasoning_details,
  verified against the live API
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