Skip to content

Dependency Footprint

anyask supports 18 LLM providers, but no single install needs all of their SDKs at once. pip install anyask installs only PyYAML (used by the built-in default-model catalog) — every provider SDK is an opt-in extra via pip install anyask[<provider>] (see Installation). This page documents what a representative sample of extras actually pull in, verified with real, clean-venv installs — useful when you're sizing a container image or just curious where the weight goes.

Bare install: 2 packages

pip install anyask

Installs only anyask itself plus its single dependency, PyYAML. This is fully functional on its own — import anyask, anyask.ask(...), anyask.get_provider(...) all work for constructing/inspecting objects. Only calling a provider requires its extra:

>>> import anyask
>>> anyask.ask("hi", provider="anthropic", model="claude-haiku-4-5-20251001")
ImportError: Provider 'anthropic' requires additional dependencies that
are not installed. Install with: pip install anyask[anthropic]

Per-provider package counts (measured sample)

Extra Total packages (incl. anyask) What makes up the difference
[bedrock] 9 Reuses boto3's own credential chain — no httpx/pydantic stack, just boto3/botocore + their small support libs (jmespath, s3transfer, python-dateutil, six, urllib3)
[oci] 15 Per-request cryptographic signing needs cryptography + pyOpenSSL + PyJWT (no bearer API key at all), plus circuitbreaker/crc32c/pytz support libs
[anthropic] 18 httpx (sync+async client) + pydantic (typed request/response models), plus jiter/distro/docstring_parser
[openai] (+ 9 aliases below) 19 Same httpx + pydantic stack as anthropic, plus tqdm/colorama
[mistral] 31 Heaviest — a full OpenTelemetry SDK (opentelemetry-api/-sdk/exporters) plus protobuf, googleapis-common-protos, and invoke (a task-runner library), on top of its own httpx+pydantic stack

[openai] also covers deepseek, groq, xai, together, fireworks, cerebras, perplexity, openrouter, and moonshot at no extra package cost — all nine subclass the same OpenAI-compatible chat- completions client (anyask/providers/openai_compat.py) and need nothing beyond the openai package itself.

[gemini]/[vertexai] (both use google-genai), [azure], and [cohere] were not individually re-verified in a clean venv for this page — they weren't called out as a lightest/heaviest/unusual case in the sample above, so treat their counts as "similar shape to anthropic/ openai's httpx+pydantic stack, plus google-auth→cryptography for the Google-backed ones" rather than exact numbers.

Package count and disk size are two different axes. oci pulls in fewer total packages (15) than mistral (31) despite being the single heaviest file on PyPI (~36MB) — a heavy SDK author can produce either a few large files or many small ones.

Why some SDKs are heavier than others

The differences trace back to each provider's own upstream SDK design, not anything this package controls:

  • HTTP client choice: httpx (async-capable, used by openai, anthropic, and most httpx-based SDKs) pulls in httpcore, h11, anyio, and sniffio on top of itself. bedrock reuses boto3's own request machinery instead.
  • Request/response validation: pydantic (used by most of the httpx-based SDKs) brings its own compiled-Rust pydantic_core plus annotated-types/typing-inspection.
  • Credential machinery: oci needs cryptography+pyOpenSSL+ PyJWT because it signs every request itself rather than sending a bearer token; bedrock needs nothing extra since boto3 resolves AWS credentials internally.
  • Unrelated SDK features: mistral's OpenTelemetry stack (for built-in tracing) has nothing to do with making a chat-completions call, but ships as a hard requirement of that SDK regardless.

How this was verified

Each number above comes from an actual pip install -e ".[<extra>]" into a fresh, empty virtual environment (not from reading pyproject.toml alone), followed by pip list:

python -m venv /tmp/anyask-check-<extra>
/tmp/anyask-check-<extra>/bin/pip install -e ".[<extra>]"
/tmp/anyask-check-<extra>/bin/pip list

Verified extras: bare (no extra), anthropic, openai, bedrock, oci, mistral — the same representative sample (lightest/heaviest/ unusual) called out above. Counts were produced on Python 3.14; exact transitive versions will differ slightly on Python 3.9/3.12 (this package's actual CI matrix), but the package count and relative ordering should hold. See anyask/registry.py for the lazy-loading mechanism that makes per-extra installs possible in the first place.