Bernstein vs TongyiDeepResearch: quick decision guide

Install both

pipx install bernstein

Feature matrix

Adapter source: src/bernstein/adapters/tongyi_deepresearch.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.

1. Bernstein ships a TongyiDeepResearch adapter at src/bernstein/adapters/tongyi_deepresearch.py that wraps the upstream CLI as one of 42 routable agents. [source: bernstein adapter source, as of 2026-09-26]
2. 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 2026-09-26]
3. Bernstein records lineage and a replay journal on every run; with the audit log enabled it also writes an HMAC-SHA256 chained record under .sdd/ that lets a reviewer replay every routing and quality-gate decision. [source: bernstein repo, as of 2026-09-26]

Where TongyiDeepResearch fits in Bernstein

Bernstein registers TongyiDeepResearch under the slug "tongyi-deepresearch" and the registry name "tongyi_deepresearch". The adapter source lives at src/bernstein/adapters/tongyi_deepresearch.py in the bernstein repo and was last touched at build time 2026-09-26. The TongyiDeepResearch adapter file is 89 lines and 3,693 bytes long, fingerprinted d0c727c703bc7ff1 (first 16 hex chars of SHA-256). No upstream GitHub repository is recorded in the bernstein adapter for TongyiDeepResearch; refer to the upstream vendor's documentation when auditing. The bernstein adapter file for TongyiDeepResearch does not yet carry a "Last verified against upstream" line; this means the adapter still tracks an unpinned upstream binary. Bernstein routes tasks to TongyiDeepResearch when its pass rate on similar work clears the configured threshold, otherwise the deterministic Python scheduler picks a different adapter from the 40+ adapter catalog.

Adapter source excerpt

The module docstring of the TongyiDeepResearch adapter, as it stands in the bernstein repo. Punctuation is normalised for the web; the wording is the author's. Length: 1438 characters.

Tongyi DeepResearch adapter for Bernstein. Runs Tongyi DeepResearch's ReAct agent (search, visit, scholar, python tools) on the task prompt from an operator-provided checkout, and records the answer and every page its visit tool opened as a verifiable run artifact (:mod:`.deep_research_artifact`). Upstream serves the planning model from a local inference server; here both the planner and the page summariser go through the agent's own gateway endpoint and key instead (see :mod:`.tongyi_deepresearch_runner`). Operator settings: * ``BERNSTEIN_TONGYI_DEEPRESEARCH_OPENAI_BASE_URL`` / ``_OPENAI_API_KEY`` / ``_MODEL`` - required; the gateway endpoint, this agent's key, the model; * ``BERNSTEIN_TONGYI_DEEPRESEARCH_HOME`` - required; the checkout whose ``inference/`` directory holds the agent; * ``BERNSTEIN_TONGYI_DEEPRESEARCH_PYTHON`` - the interpreter with its requirements installed (default ``python3``); * ``BERNSTEIN_TONGYI_DEEPRESEARCH_SUMMARY_MODEL`` - the page-summary model (default: the planning model); * ``BERNSTEIN_TONGYI_DEEPRESEARCH_SERPER_BASE_URL`` / ``_JINA_BASE_URL`` - where the search/scholar and visit tools send what they would send to ``google.serper.dev`` and ``r.jina.ai`` (default: those hosts); the endpoint must speak the same protocol; * the tool settings the agent reads itself (``SERPER_KEY_ID``, ``JINA_API_KEYS``, ``SANDBOX_FUSION_ENDPOINT``, ``MAX_LLM_CALL_PER_RUN``, ...) are let through unchanged.

Adapter telemetry

When to pick which

FAQ