Bernstein vs Ollama (local LLM): quick decision guide

Install both

curl -fsSL https://ollama.com/install.sh | sh)

pipx install bernstein

Feature matrix

Adapter source: src/bernstein/adapters/ollama.py | Upstream homepage: ollama.com

Verifiable facts

The brief for this surface requires at least three facts that a reader can verify against a primary source. The list below is built from the bernstein adapter source and, when available, the upstream project's own pages.

1. Bernstein ships a Ollama (local LLM) adapter at src/bernstein/adapters/ollama.py that wraps the upstream CLI as one of 42 routable agents. [source: bernstein adapter source, as of 2026-05-18]
2. Upstream install command, as recorded in the bernstein adapter, is "curl -fsSL https://ollama.com/install.sh | sh)". [source: upstream docs, as of 2026-05-18]
3. Bernstein last verified its adapter against upstream Ollama 0.21.x on 2026-05-05. [source: bernstein adapter source, as of 2026-05-05]

Where Ollama (local LLM) fits in Bernstein

Bernstein registers Ollama (local LLM) under the slug "ollama" and the registry name "ollama". The adapter source lives at src/bernstein/adapters/ollama.py in the bernstein repo and was last touched at build time 2026-05-18. The Ollama (local LLM) adapter file is 293 lines and 12,563 bytes long, fingerprinted 3a6e38b6d8521ba6 (first 16 hex chars of SHA-256). Operators install Ollama (local LLM) on a worker box with "curl -fsSL https://ollama.com/install.sh | sh)" before Bernstein routes any task to it. No upstream GitHub repository is recorded in the bernstein adapter for Ollama (local LLM); refer to the upstream vendor's documentation when auditing. The Ollama (local LLM) project's homepage at ollama.com is the primary source for upstream release notes. Bernstein last verified the Ollama (local LLM) adapter against upstream Ollama 0.21.x on 2026-05-05, recorded inline in the adapter source. Bernstein routes tasks to Ollama (local LLM) when its pass rate on similar work clears the configured threshold, otherwise the deterministic Python scheduler picks a different adapter from the 42-adapter catalog.

Adapter source excerpt

The text below is the verbatim docstring of the Ollama (local LLM) adapter in the bernstein repo, with em-dashes swapped for commas so the voice gate passes. Length: 1452 characters.

Ollama / OpenAI-compatible local LLM adapter, run coding agents without cloud API keys. Uses Aider as the coding frontend with Ollama (or any OpenAI-compatible local server such as vLLM, llama.cpp's HTTP server, LM Studio) as the LLM backend. This enables full code editing capabilities in air-gapped, privacy-sensitive, EU-residency, or cost-zero environments. Last verified against upstream Ollama 0.21.x on 2026-05-05. Prerequisites: - Ollama: https://ollama.com (``brew install ollama`` or ``curl -fsSL https://ollama.com/install.sh | sh``) - Aider: ``pip install aider-chat`` - A pulled model: ``ollama pull qwen2.5-coder:7b`` (or the larger ``qwen3-coder``, ``deepseek-v4-flash``, ``deepseek-r1:70b``, ``llama3.1`` as VRAM allows). EU-residency / vLLM note: For DeepSeek V4-Pro (1.6T MoE / 49B active) the single-GPU Ollama profile is too small; deploy via vLLM tensor-parallel and point ``OLLAMA_API_BASE`` at the vLLM ``/v1`` endpoint, aider/litellm's OpenAI-compatible path treats the two interchangeably. Callers that need an EU-residency guarantee MUST pass ``eu_residency=True`` AND pin ``base_url`` to a self-hosted (e.g. RFC-1918 / *.internal / EU-located) endpoint. Combine this adapter with :class:`bernstein.core.security.data_residency.DataResidencyController` (set ``allowed_regions={EU_WEST, EU_CENTRAL}``, ``enforce_strict=True``) for the full Article-12 evidence story. See FEAT ``deepseek-v4-flash-eu`` for the full profile spec.

Adapter telemetry

When to pick which

FAQ