Add ZeroGPU connector

connectors/zerogpu/README.md

@@ -0,0 +1,47 @@
+# ZeroGPU
+
+Submitted by ZeroGPU (owner: vendor).
+
+## API documentation
+
+- `POST /v1/chat/completions`: https://docs.zerogpu.ai/api-reference/chat-completions
+- `POST /v1/responses`: https://docs.zerogpu.ai/api-reference/responses
+- `GET /v1/me` (the probe): https://docs.zerogpu.ai/integrations/sprites#how-key-validation-works
+  (our Sprites guide, which we publish when this connector ships; "The probe" below
+  documents the endpoint in full until then)
+- Model IDs used in the examples: https://docs.zerogpu.ai/docs/model-catalog
+
+## The probe
+
+`GET /v1/me` returns the key, project and organization the key belongs to:
+`{"object": "api_key", "id": ..., "project_id": ..., "organization_id": ...}`.
+We added it for this connector. `GET /v1/models` is public and answers 200
+without a key, so it can't tell a good key from a bad one. `/v1/me` is free and
+read-only: it runs no inference, never checks the balance or uses a free trial
+call, records no usage, and is sent with `Cache-Control: no-store`. It reads the
+key's row on every call, so a revoked key fails immediately.
+
+A wrong or revoked key gets `403` with the plain-text body `Invalid credentials`,
+hence the `errors` rule. A missing key gets `401`. The recording in
+`fixtures/zerogpu/probe_200.json` has the real response shape with the three IDs
+replaced by placeholders.
+
+## Project scoping
+
+Each ZeroGPU API key belongs to exactly one project, and the API derives the
+project from the key. The `x-project-id` header is ignored by the API, so it is
+not in `forward_request_headers`.
+
+## Response headers
+
+`x-request-id` comes back on every response (support uses it to trace a
+request). `retry-after` comes back on `429` rate-limit responses.
+
+## Icons
+
+Both icons come from ZeroGPU's own logo, the dark favicon served by our
+dashboard at https://platform.zerogpu.ai/assets/favicon_dark.svg. ZeroGPU owns the mark and submits it here for
+this listing. `icon.svg` is the logo as we draw it (black disc, green ring and
+halves) without the square favicon background, cropped to the circle.
+`mark.svg` is the ring and the two halves alone, with every fill set to
+`currentColor`.

connectors/zerogpu/connector.json

@@ -0,0 +1,96 @@
+{
+  "$schema": "https://raw.githubusercontent.com/superfly/sprites-connectors/main/v1/schema.json",
+  "format_version": 1,
+  "name": "zerogpu",
+  "display_name": "ZeroGPU",
+  "base_url": "https://api.zerogpu.ai",
+  "description": "Run AI inference with ZeroGPU's open-weight models and specialized small models.",
+  "long_description": "Bring a ZeroGPU API key to run OpenAI-compatible inference on open-weight models and ZeroGPU's specialized small language models (ZLMs) for generation, classification and extraction. Saving, replacing and testing a key reads GET /v1/me, which returns the key, project and organization it belongs to; the check is free, runs no inference and does not look at your balance, so a valid key on an account without credit still saves. All forwarded inference is charged to your ZeroGPU account. The gateway keeps the key off your Sprite; the key's own project is used for every request. Supports chat completions and the Responses API, including SSE streaming, plus GET /v1/me. Model listing, embeddings, moderations, audio, batches, files, API-key management, billing and account routes are not forwarded.",
+  "website_url": "https://platform.zerogpu.ai/dashboard",
+  "auth": {"header": "x-api-key"},
+  "allowed_paths": [
+    {"methods": ["POST"], "pattern": "/v1/chat/completions"},
+    {"methods": ["POST"], "pattern": "/v1/responses"},
+    {"methods": ["GET"], "pattern": "/v1/me"}
+  ],
+  "forward_response_headers": ["x-request-id", "retry-after"],
+  "probe": {
+    "method": "GET",
+    "path": "/v1/me",
+    "label": "key identity",
+    "expect": [
+      ["object", "eq", "api_key"],
+      ["id", "text"],
+      ["project_id", "text"],
+      ["organization_id", "text"],
+      ["error", "absent"]
+    ],
+    "errors": [{"status": 403, "reason": "invalid_api_key"}]
+  },
+  "examples": [
+    {
+      "title": "Chat with an open-weight model (charged to your ZeroGPU account)",
+      "method": "POST",
+      "path": "/v1/chat/completions",
+      "json": {
+        "model": "meta-llama/llama-3.1-8b-instruct",
+        "messages": [{"role": "user", "content": "Write a one-line tagline for a coffee shop."}],
+        "max_tokens": 60,
+        "stream": false
+      },
+      "notes": [
+        "Set stream to true for SSE. Other models are listed at docs.zerogpu.ai/docs/model-catalog."
+      ]
+    },
+    {
+      "title": "Classify text into IAB categories with a ZLM (charged to your ZeroGPU account)",
+      "method": "POST",
+      "path": "/v1/chat/completions",
+      "json": {
+        "model": "zlm-v1-iab-classify-edge",
+        "messages": [{"role": "user", "content": "Best trail running shoes for beginners, with a guide to cushioning and grip."}]
+      },
+      "notes": [
+        "The reply's message content is a JSON string of IAB content and audience categories with scores."
+      ]
+    },
+    {
+      "title": "Responses API (charged to your ZeroGPU account)",
+      "method": "POST",
+      "path": "/v1/responses",
+      "json": {
+        "model": "openai/gpt-oss-120b",
+        "input": "Summarize what an edge inference network is in two sentences.",
+        "max_output_tokens": 300
+      }
+    }
+  ],
+  "messages": {
+    "400": "ZeroGPU rejected the request; check the model ID and request body",
+    "402": "ZeroGPU balance exhausted; add credits or enable autopay at platform.zerogpu.ai",
+    "403": "ZeroGPU rejected the API key; check it or create a new one at platform.zerogpu.ai",
+    "404": "ZeroGPU model not found; check the model ID at docs.zerogpu.ai/docs/model-catalog",
+    "429": "ZeroGPU rate limit reached; retry after the retry-after interval"
+  },
+  "discovery": {
+    "categories": ["ai-inference"],
+    "keywords": ["llm", "language models", "open-weight models", "small language models", "classification", "extraction", "iab", "openai-compatible"]
+  },
+  "icon_svg": {"file": "icon.svg"},
+  "icon_bg_class": "bg-green-500/20",
+  "presentation": {
+    "tagline": "Open-weight models and specialized small models without storing your ZeroGPU key on a Sprite.",
+    "setup": "Log in at platform.zerogpu.ai/dashboard, open API Keys and create a key (it starts with `zgpu-api-`).",
+    "notices": [
+      {
+        "title": "Billing",
+        "body": "Inference is charged to your ZeroGPU account. Saving, replacing and testing a key read `GET /v1/me`, which is free and does not check your balance; an account without credit saves its key, and inference then returns 402 until you add credit."
+      },
+      {
+        "title": "Scope",
+        "body": "Only `POST /v1/chat/completions` and `POST /v1/responses` (including SSE streaming) and `GET /v1/me` (the key's own key, project and organization IDs) are forwarded. Requests run in the key's own project. Model listing, embeddings, moderations, audio, batches, files, API keys, billing and account routes are not forwarded."
+      }
+    ],
+    "mark_svg": {"file": "mark.svg"}
+  }
+}