Skip to content
Python

Add Cerebras, Hugging Face, Cohere and SiliconFlow as compat providers

Rohit C Prasadb849b53d27bd
diff --git a/README.md b/README.md
index c6db7374..fb145804 100644
--- a/README.md
+++ b/README.md
@@ -87,7 +87,7 @@ Unattended runs never self-approve: their asks park in an inbox until a human an
Model access is yours: pick a provider, paste your key, switch anytime. Supported out of the box:
-**OpenAI · Anthropic · Google Gemini · BytePlus Ark · Volcengine Ark Agent Plan · Atlas Cloud · Inkling (Thinking Machines) · GLM (Z.ai) · DeepSeek · Kimi (Moonshot) · Qwen · MiniMax · Mistral · Grok (xAI) · Solar (Upstage)** - plus aggregators **OrcaRouter** and **OpenRouter**, open-weight models via **Together** and **Fireworks**, and fully local models via **Ollama**.
+**OpenAI · Anthropic · Google Gemini · BytePlus Ark · Volcengine Ark Agent Plan · Atlas Cloud · Inkling (Thinking Machines) · GLM (Z.ai) · DeepSeek · Kimi (Moonshot) · Qwen · MiniMax · Mistral · Grok (xAI) · Solar (Upstage) · Cohere** - plus aggregators **OrcaRouter**, **OpenRouter** and **Hugging Face Inference Providers**, open-weight models via **Together**, **Fireworks**, **Cerebras** and **SiliconFlow**, and fully local models via **Ollama**, llama.cpp, vLLM, LM Studio, Jan, Docker Model Runner or mlx-lm.
A curated model list marks what we've verified for tool-calling work. Adding any model string works at your own risk.
diff --git a/coworker/providers/matrix.py b/coworker/providers/matrix.py
index 5fd31245..6ed6eeed 100644
--- a/coworker/providers/matrix.py
+++ b/coworker/providers/matrix.py
@@ -17,7 +17,8 @@ showing a made-up denominator. Values entered 2026-07-28 from vendor docs; verif
the id refresh.
Resellers: Together + Fireworks + OpenRouter + Nebius + OrcaRouter + NVIDIA NIM
-+ Groq (landed 2026-10-25, after the deferral of 2026-07-04).
++ Groq (landed 2026-10-25, after the deferral of 2026-07-04) + Cerebras, Hugging Face
+Inference Providers and SiliconFlow (2026-10-26).
"""
from __future__ import annotations
@@ -221,6 +222,19 @@ MATRIX: dict[str, ModelEntry] = {
# 2026-08-12 and solar-pro3 retires 2026-10-30 (docs/models/history).
"upstage:solar-pro4": ModelEntry("Solar Pro 4 · Upstage", _AGENTIC, 512_000),
"upstage:solar-mini4": ModelEntry("Solar Mini 4 · Upstage", _AGENTIC, 512_000),
+ # -- Cerebras / Hugging Face / Cohere / SiliconFlow (added 2026-10-26) --------
+ # Cerebras: inference-docs.cerebras.ai/models/openai-oss — "Tool Calling" listed,
+ # "65k tokens (free tier), 131k tokens (paid)"; the paid window is stored.
+ "cerebras:gpt-oss-120b": ModelEntry("GPT-OSS 120B · via Cerebras", _AGENTIC, 131_072),
+ # Hugging Face Inference Providers: router.huggingface.co/v1/models (public listing)
+ # reports supports_tools=true and context_length 1048576 for GLM-5.2 at its providers.
+ "huggingface:zai-org/GLM-5.2": ModelEntry("GLM-5.2 · via Hugging Face", _AGENTIC, 1_048_576),
+ # Cohere: docs.cohere.com/docs/models lists command-a-plus-05-2026 at 128k; the
+ # tool-use guide and the Compatibility API docs use this id in their examples.
+ "cohere:command-a-plus-05-2026": ModelEntry("Command A+ · Cohere", _AGENTIC, 128_000),
+ # SiliconFlow: siliconflow.com/blog/glm-5-2-api-guide — id zai-org/GLM-5.2, "1M-token
+ # context window" (pricing table: 1049K); tool calling among the listed workflows.
+ "siliconflow:zai-org/GLM-5.2": ModelEntry("GLM-5.2 · via SiliconFlow", _AGENTIC, 1_048_576),
# -- SambaNova ---------------------------------------------------------------
# SambaCloud's OpenAI-compatible API; ids and windows are the `context_length` values
# api.sambanova.ai/v1/models reports (keyless, read 2026-10-25). The two 128k rows are
diff --git a/coworker/providers/registry.py b/coworker/providers/registry.py
index 225bbb16..d8a70c43 100644
--- a/coworker/providers/registry.py
+++ b/coworker/providers/registry.py
@@ -842,6 +842,41 @@ DESCRIPTORS: list[ProviderDescriptor] = [
recommended_model="llama-3.3-70b-versatile",
env_key="GROQ_API_KEY",
),
+ # Four more OpenAI-compatible vendors (2026-10-26); endpoints, key pages and the
+ # recommended ids read from each vendor's docs that day (sources in matrix.py).
+ _compat(
+ "cerebras",
+ "Cerebras",
+ base_url="https://api.cerebras.ai/v1",
+ recommended_model="gpt-oss-120b",
+ env_key="CEREBRAS_API_KEY",
+ ),
+ # Hugging Face routes each repo id to one of its partner providers; model ids are
+ # the Hub repo ids, optionally with a `:provider` suffix to pin one.
+ _compat(
+ "huggingface",
+ "Hugging Face (Inference Providers)",
+ base_url="https://router.huggingface.co/v1",
+ recommended_model="zai-org/GLM-5.2",
+ env_key="HF_TOKEN",
+ endpoint_help="Prefilled with the Inference Providers router. Model ids are Hugging Face repo ids, e.g. zai-org/GLM-5.2.",
+ ),
+ _compat(
+ "cohere",
+ "Cohere",
+ base_url="https://api.cohere.ai/compatibility/v1",
+ recommended_model="command-a-plus-05-2026",
+ env_key="COHERE_API_KEY",
+ endpoint_help="Prefilled with Cohere's Compatibility API, which speaks the OpenAI chat format.",
+ ),
+ _compat(
+ "siliconflow",
+ "SiliconFlow",
+ base_url="https://api.siliconflow.com/v1",
+ recommended_model="zai-org/GLM-5.2",
+ env_key="SILICONFLOW_API_KEY",
+ endpoint_help="Prefilled with SiliconFlow's international endpoint. China mainland: https://api.siliconflow.cn/v1 (keys from cloud.siliconflow.cn).",
+ ),
ProviderDescriptor(
name="ollama",
title="Ollama",
diff --git a/coworker/server/manager.py b/coworker/server/manager.py
index 85f508bb..8f578b1b 100644
--- a/coworker/server/manager.py
+++ b/coworker/server/manager.py
@@ -4354,6 +4354,11 @@ class SessionManager:
],
"atlascloud": ["zai-org/glm-5.2"],
"upstage": ["solar-pro4", "solar-mini4"],
+ # Verified 2026-10-26 (sources in providers/matrix.py).
+ "cerebras": ["gpt-oss-120b"],
+ "huggingface": ["zai-org/GLM-5.2"],
+ "cohere": ["command-a-plus-05-2026"],
+ "siliconflow": ["zai-org/GLM-5.2"],
}
def _suggested_models(self, name: str) -> list[str]:
diff --git a/surfaces/gui/src/providers/ProviderSetup.tsx b/surfaces/gui/src/providers/ProviderSetup.tsx
index 460e41d9..bd7ce79a 100644
--- a/surfaces/gui/src/providers/ProviderSetup.tsx
+++ b/surfaces/gui/src/providers/ProviderSetup.tsx
@@ -59,6 +59,10 @@ export const KEY_HELP: Record<string, { url: string; label: string; labelKey?: s
atlascloud: { url: "https://www.atlascloud.ai/console/api-keys", label: "atlascloud.ai" },
groq: { url: "https://console.groq.com/keys", label: "console.groq.com" },
upstage: { url: "https://console.upstage.ai/api-keys", label: "console.upstage.ai" },
+ cerebras: { url: "https://cloud.cerebras.ai", label: "cloud.cerebras.ai" },
+ huggingface: { url: "https://huggingface.co/settings/tokens", label: "huggingface.co" },
+ cohere: { url: "https://dashboard.cohere.com/api-keys", label: "dashboard.cohere.com" },
+ siliconflow: { url: "https://cloud.siliconflow.com", label: "cloud.siliconflow.com" },
};
export type Verify = { state: "idle" | "testing" | "ok" | "error"; msg?: string };
diff --git a/surfaces/gui/src/providers/logos.ts b/surfaces/gui/src/providers/logos.ts
index a2087b60..731a10c4 100644
--- a/surfaces/gui/src/providers/logos.ts
+++ b/surfaces/gui/src/providers/logos.ts
@@ -113,6 +113,11 @@ export const PROVIDER_ORDER = [
"novita",
"groq",
"upstage",
+ // No vendored marks yet for these four: the gallery shows the neutral monogram.
+ "cerebras",
+ "huggingface",
+ "cohere",
+ "siliconflow",
];
export function providerRank(name: string): number {
diff --git a/tests/test_providers.py b/tests/test_providers.py
index 6785471b..ce92ec82 100644
--- a/tests/test_providers.py
+++ b/tests/test_providers.py
@@ -292,6 +292,10 @@ COMPAT_VENDORS = {
"nvidia": "https://integrate.api.nvidia.com/v1",
"atlascloud": "https://api.atlascloud.ai/v1",
"upstage": "https://api.upstage.ai/v1",
+ "cerebras": "https://api.cerebras.ai/v1",
+ "huggingface": "https://router.huggingface.co/v1",
+ "cohere": "https://api.cohere.ai/compatibility/v1",
+ "siliconflow": "https://api.siliconflow.com/v1",
}
@@ -474,6 +478,10 @@ def test_compat_models_route_and_get_tool_capabilities():
"orcarouter:anthropic/claude-sonnet-5",
"atlascloud:zai-org/glm-5.2",
"upstage:solar-pro4",
+ "cerebras:gpt-oss-120b",
+ "huggingface:zai-org/GLM-5.2",
+ "cohere:command-a-plus-05-2026",
+ "siliconflow:zai-org/GLM-5.2",
):
prefix = model.split(":", 1)[0]
assert router._provider_name(model) == prefix