Welcome to ava-ai

You've been given a personal API key to a fast hosted LLM. Any tool that speaks the OpenAI API works. You need exactly two values:

Base URLhttps://ava-ai.duckdns.org/v1
API keythe sk-… key you received privately

Models

NameTypeNotes
gpt-oss-120bchatdefault — fastest, 131k context
gpt-oss-120b-alt3chatsame model, alternate route — try if the default is ever down or slow
llama-3.1-8bchatsmall + very cheap — drafts, summaries, bulk work
gpt-oss-20bchatsmaller gpt-oss — cheapest chat option
llama-3.3-70bchatcapable 70B generalist
qwen3.6-27bchat + visionnewer Qwen, reads images — ⚠ pricier output
zai-glm-4.7chat⚠ premium coder/agents — drains budget ~5× faster
gemma-4-31bchat + visionreads images/screenshots — mid-price
Flash-2-RCVchat + visionreasoning/coding/vision — rate-limited, best for lighter use
bge-small-en-v1.5embeddings384 dimensions

Use gpt-oss-120b by default — it's the fastest. More models are added as we grow; call /v1/models anytime for the current list. All spend on every model counts against the same personal budget.

Desktop chat apps

Any chat app that supports a custom OpenAI-compatible provider works:

  1. In settings, add a new provider of type OpenAI-compatible
  2. Base URL: https://ava-ai.duckdns.org/v1 · API key: your key
  3. Refresh the model list (or add gpt-oss-120b by hand), then pick it in a new chat

Coding agents & CLI tools

Most of these read an OpenAI-compatible endpoint from environment variables:

OPENAI_BASE_URL=https://ava-ai.duckdns.org/v1
OPENAI_API_KEY=sk-YOUR-KEY

If yours uses a config file instead, point its base URL and API key at the same two values. Note: some tools only show models you list explicitly — they do not auto-discover them. If your model list comes up empty, declare the ones you want:

gpt-oss-120b
gpt-oss-120b-alt3
gpt-oss-20b
llama-3.1-8b
llama-3.3-70b
qwen3.6-27b
zai-glm-4.7
gemma-4-31b
Flash-2-RCV

curl

curl https://ava-ai.duckdns.org/v1/chat/completions \
  -H "Authorization: Bearer sk-YOUR-KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"gpt-oss-120b","messages":[{"role":"user","content":"hello"}]}'

Python (openai SDK)

from openai import OpenAI
client = OpenAI(base_url="https://ava-ai.duckdns.org/v1", api_key="sk-YOUR-KEY")
r = client.chat.completions.create(model="gpt-oss-120b",
    messages=[{"role": "user", "content": "hello"}])
print(r.choices[0].message.content)

Budget & limits

Your key has a monthly dollar budget and a rate limit (30 requests/min, 2 parallel). Check where you stand anytime at ava-ai.duckdns.org/usage — paste your key, nothing is stored. If you hit your budget you'll get a clear quota exceeded error; message the admin for a bump or wait for the monthly reset.

Privacy: your prompts and responses are never stored — this proxy only meters token counts. The IP address each request comes from is logged for abuse protection. Keep your key private; anyone holding it spends your budget.