Compare
Bernstein vs CLM (sovereign LLM): quick decision guide
CLM sovereign LLM adapter, drives a customer-side CLM gateway.
Page built on 2026-05-18 from data/adapters-meta.json. Every claim below links to its primary source.
Install both
CLM (sovereign LLM)
No install command recorded in the bernstein adapter source as of 2026-05-18.
Bernstein
pipx install bernsteinApache-2.0. Deterministic Python scheduler.
Feature matrix
| Capability | CLM (sovereign LLM) | Bernstein |
|---|---|---|
| Install method | Not recorded | pipx install bernstein |
| License | Not recorded | Apache-2.0 |
| Authentication | Not recorded | Per-agent credential scoping (no shared key) |
| Multi-agent orchestration | One agent in a terminal | CLM (sovereign LLM) plus 41 other adapters in parallel worktrees |
| MCP support | Not measured | Yes |
| Parallel-safe in worktrees | Not measured | Yes (designed around git worktrees) |
| HMAC-chained audit log | No | Yes (RFC 2104 SHA-256 chain in .sdd/) |
| Deterministic scheduler | Not applicable (single-agent CLI) | Yes (Deterministic Python scheduler) |
Adapter source: src/bernstein/adapters/clm.py
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.
- Bernstein ships a CLM (sovereign LLM) adapter at src/bernstein/adapters/clm.py that wraps the upstream CLI as one of 42 routable agents. [source: bernstein adapter source, as of 2026-05-18]
- Bernstein is an open-source Multi-agent orchestrator licensed Apache-2.0, with a deterministic Python scheduler that routes work across CLI agents in parallel git worktrees. [source: bernstein repo, as of 2026-05-18]
- Bernstein writes an HMAC-SHA256 chained audit log under .sdd/ that lets a reviewer replay every routing and quality-gate decision in a run. [source: bernstein repo, as of 2026-05-18]
Where CLM (sovereign LLM) fits in Bernstein
Bernstein registers CLM (sovereign LLM) under the slug "clm" and the registry name "clm". The adapter source lives at src/bernstein/adapters/clm.py in the bernstein repo and was last touched at build time 2026-05-18. The CLM (sovereign LLM) adapter file is 461 lines and 17,843 bytes long, fingerprinted e388967a6eac0a53 (first 16 hex chars of SHA-256). No upstream GitHub repository is recorded in the bernstein adapter for CLM (sovereign LLM); refer to the upstream vendor's documentation when auditing. The bernstein adapter file for CLM (sovereign LLM) does not yet carry a "Last verified against upstream" line; this means the adapter still tracks an unpinned upstream binary. Bernstein routes tasks to CLM (sovereign 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 CLM (sovereign LLM) adapter in the bernstein repo, with em-dashes swapped for commas so the voice gate passes. Length: 1244 characters.
CLM sovereign LLM adapter, drives a customer-side CLM gateway. Some sovereign-AI vendors deploy a customer-side Cyber Language Model (CLM) served behind NVIDIA NIM, which exposes an OpenAI-compatible HTTP API (TensorRT-LLM + Triton). This adapter spawns ``aider`` configured to talk to that gateway via ``OPENAI_API_BASE`` / ``OPENAI_API_KEY``, unlocking Bernstein's HMAC audit chain, lineage trail, and fingerprint memoisation for engineering workflows against CLM. Phase 1, adapter MVP. Phase 2 partial, tool-calling allowlist (T578) wired into the OpenAI-compatible ``tools=[]`` request shape, lethal-trifecta refusal, and streaming-assembly helper whose lineage payload always carries the full response, never just the first chunk. Phase 2.5 (this module), opt-in mTLS to the customer gateway, reusing :class:`bernstein.core.protocols.cluster.cluster_tls.TLSConfig` plumbing. When ``CLM_CERT_FILE`` / ``CLM_KEY_FILE`` / ``CLM_CA_FILE`` are set, the adapter routes the spawn through a small launcher (:mod:`bernstein.adapters.clm_tls_launcher`) that monkey-patches :class:`httpx.Client` defaults before importing aider, so the OpenAI SDK transparently presents the customer-issued client cert. See ``docs/adapters/clm.md`` for configuration.Adapter telemetry
| Registry name | clm |
|---|---|
| Adapter class | CLM (sovereign LLM) |
| Source file | src/bernstein/adapters/clm.py |
| Source file size | 461 lines, 17,843 bytes |
| Source SHA-256 | e388967a6eac0a531ef3acec74512367e304ea85bf201c459df4e67d15508b35 |
| Category bucket | openai-family |
| Upstream repo | Not derivable from adapter source |
| Upstream homepage | Not recorded |
| Last verified upstream | No "Last verified" line in adapter source |
| Operator-curated overlay | No (programmatic page) |
When to pick which
Choose CLM (sovereign LLM)
Reach for CLM (sovereign LLM) when the work is a single thread that fits one agent: in a single-process terminal session, designed for single-instance use per repo. Auth model is configured per upstream docs. You skip the orchestrator round-trip and get the smallest possible surface between you and the model.
Choose Bernstein
Wrap CLM (sovereign LLM) under Bernstein when the goal splits into parallel tasks, when you want an HMAC-chained audit log on every routing decision, or when a deterministic Python scheduler (no LLM picking who runs what) is a hard requirement. Bernstein also runs Claude, Gemini, and Aider against the same plan so an OpenAI miss is caught by a cross-model verifier.
FAQ
Does Bernstein replace CLM (sovereign LLM)?
No. Bernstein wraps CLM (sovereign LLM) as one of 42 CLI adapters and routes tasks to it based on per-task pass-rate history. CLM (sovereign LLM) keeps running unchanged; Bernstein decides when it gets work.
Can I run CLM (sovereign LLM) alongside other agents in the same repo?
Yes. Each agent runs in its own git worktree under .worktrees/, so file edits never collide. Bernstein merges results back to the trunk only after the configured quality gates (lint, types, tests) pass.
Is this comparison page handwritten?
No. The template is fixed; every fact and every link is pulled from the bernstein adapter source in the master branch and (when available) the upstream project's own pages. The data extractor lives at scripts/gen-compare-data.mjs. No LLM writes the prose.