Skip to content

Home

banner

anyask is a single-responsibility Python package for one job: call an LLM provider, get its raw response back. No routing, no fallback, no retries — you always name the provider and model explicitly, and you always get the same normalized AskResponse shape back regardless of which of the 18 supported vendors you called.

import anyask

response = anyask.ask(
    "Say hello in one word.",
    provider="anthropic",
    model="claude-haiku-4-5-20251001",
    api_key="sk-...",  # or set ANTHROPIC_API_KEY
)

print(response.content)          # "Hello"
print(response.usage)            # TokenUsage(prompt_tokens=..., completion_tokens=..., total_tokens=...)
print(response.finish_reason)    # "end_turn" (raw, provider-specific - never normalized)

It is designed for:

  • anyone who wants a single, uniform call signature across every major LLM vendor without adopting a full agent/orchestration framework
  • applications that already do their own retry, fallback, and routing logic and just need a thin, predictable client underneath
  • teams that want to add or drop a provider by installing/removing one pip extra, with no import-time cost for providers they don't use

Features

  • One function, 18 vendors — ask()/ask_async() take the same provider="..."/model="..." signature for OpenAI, Anthropic, Gemini, Vertex AI, Azure, Mistral, Bedrock, Cohere, OCI, and nine OpenAI-compatible vendors (DeepSeek, Groq, xAI, Together AI, Fireworks AI, Cerebras, Perplexity, OpenRouter, Moonshot)
  • Frozen, normalized AskResponse — content, usage, finish_reason (kept raw and provider-specific, never coerced), provider, model, and the untouched raw SDK response
  • Lazy per-provider imports — installing none of the 18 provider extras still gives a fully working import anyask; each provider's SDK is only imported the moment it's actually used
  • No routing/fallback magic — you always name the provider and model; anyask never silently swaps one for another
  • get_provider() for reuse — construct a provider once and call generate_sync()/generate() many times instead of rebuilding an SDK client on every call
  • anyask check <provider>/anyask ready — a small CLI that answers "is this provider's SDK installed and its credentials set," without making a network call

What anyask does (and doesn't) do

prompt is always a plain string; AskResponse.content is always a plain string. That single-string-in, single-string-out contract is what "raw response, one function, no magic" means in practice, and it draws a hard line around the scope:

Does: one text prompt → one text completion, the same call signature across all 18 providers, with provider-native extras (temperature, max_tokens, reasoning, and anything else a given provider's SDK call accepts) reachable via **kwargs, and everything else the SDK response carries reachable via AskResponse.raw.

Doesn't: multi-turn conversations (no messages=[...] history — every ask() is one independent turn), system prompts (not part of the contract — a few providers happen to accept one through a provider-specific **kwargs name, like Anthropic's/Bedrock's system= or Gemini's system_instruction=, but that's incidental, not guaranteed, and silently doesn't work on the rest), multi-modal input (text only — no images, audio, or files), structured/multi-modal output (content is always plain text; tool calls or other structured parts are only reachable via raw), or streaming (not implemented by any provider yet). See SPEC.md's out-of-scope list for the full picture.

Explore the Docs