@quantumclaw/quantumclaw 2026.3.22

package/docs/concepts/model-providers.md

@@ -0,0 +1,607 @@
+---
+summary: "Model provider overview with example configs + CLI flows"
+read_when:
+  - You need a provider-by-provider model setup reference
+  - You want example configs or CLI onboarding commands for model providers
+title: "Model Providers"
+---
+
+# Model providers
+
+This page covers **LLM/model providers** (not chat channels like WhatsApp/Telegram).
+For model selection rules, see [/concepts/models](/concepts/models).
+
+## Quick rules
+
+- Model refs use `provider/model` (example: `opencode/claude-opus-4-6`).
+- If you set `agents.defaults.models`, it becomes the allowlist.
+- CLI helpers: `quantumclaw onboard`, `quantumclaw models list`, `quantumclaw models set <provider/model>`.
+- Provider plugins can inject model catalogs via `registerProvider({ catalog })`;
+  QuantumClaw merges that output into `models.providers` before writing
+  `models.json`.
+- Provider manifests can declare `providerAuthEnvVars` so generic env-based
+  auth probes do not need to load plugin runtime. The remaining core env-var
+  map is now just for non-plugin/core providers and a few generic-precedence
+  cases such as Anthropic API-key-first onboarding.
+- Provider plugins can also own provider runtime behavior via
+  `resolveDynamicModel`, `prepareDynamicModel`, `normalizeResolvedModel`,
+  `capabilities`, `prepareExtraParams`, `wrapStreamFn`, `formatApiKey`,
+  `refreshOAuth`, `buildAuthDoctorHint`,
+  `isCacheTtlEligible`, `buildMissingAuthMessage`,
+  `suppressBuiltInModel`, `augmentModelCatalog`, `isBinaryThinking`,
+  `supportsXHighThinking`, `resolveDefaultThinkingLevel`,
+  `isModernModelRef`, `prepareRuntimeAuth`, `resolveUsageAuth`, and
+  `fetchUsageSnapshot`.
+- Note: provider runtime `capabilities` is shared runner metadata (provider
+  family, transcript/tooling quirks, transport/cache hints). It is not the
+  same as the [public capability model](/plugins/architecture#public-capability-model)
+  which describes what a plugin registers (text inference, speech, etc.).
+
+## Plugin-owned provider behavior
+
+Provider plugins can now own most provider-specific logic while QuantumClaw keeps
+the generic inference loop.
+
+Typical split:
+
+- `auth[].run` / `auth[].runNonInteractive`: provider owns onboarding/login
+  flows for `quantumclaw onboard`, `quantumclaw models auth`, and headless setup
+- `wizard.setup` / `wizard.modelPicker`: provider owns auth-choice labels,
+  legacy aliases, onboarding allowlist hints, and setup entries in onboarding/model pickers
+- `catalog`: provider appears in `models.providers`
+- `resolveDynamicModel`: provider accepts model ids not present in the local
+  static catalog yet
+- `prepareDynamicModel`: provider needs a metadata refresh before retrying
+  dynamic resolution
+- `normalizeResolvedModel`: provider needs transport or base URL rewrites
+- `capabilities`: provider publishes transcript/tooling/provider-family quirks
+- `prepareExtraParams`: provider defaults or normalizes per-model request params
+- `wrapStreamFn`: provider applies request headers/body/model compat wrappers
+- `formatApiKey`: provider formats stored auth profiles into the runtime
+  `apiKey` string expected by the transport
+- `refreshOAuth`: provider owns OAuth refresh when the shared `pi-ai`
+  refreshers are not enough
+- `buildAuthDoctorHint`: provider appends repair guidance when OAuth refresh
+  fails
+- `isCacheTtlEligible`: provider decides which upstream model ids support prompt-cache TTL
+- `buildMissingAuthMessage`: provider replaces the generic auth-store error
+  with a provider-specific recovery hint
+- `suppressBuiltInModel`: provider hides stale upstream rows and can return a
+  vendor-owned error for direct resolution failures
+- `augmentModelCatalog`: provider appends synthetic/final catalog rows after
+  discovery and config merging
+- `isBinaryThinking`: provider owns binary on/off thinking UX
+- `supportsXHighThinking`: provider opts selected models into `xhigh`
+- `resolveDefaultThinkingLevel`: provider owns default `/think` policy for a
+  model family
+- `isModernModelRef`: provider owns live/smoke preferred-model matching
+- `prepareRuntimeAuth`: provider turns a configured credential into a short
+  lived runtime token
+- `resolveUsageAuth`: provider resolves usage/quota credentials for `/usage`
+  and related status/reporting surfaces
+- `fetchUsageSnapshot`: provider owns the usage endpoint fetch/parsing while
+  core still owns the summary shell and formatting
+
+Current bundled examples:
+
+- `anthropic`: Claude 4.6 forward-compat fallback, auth repair hints, usage
+  endpoint fetching, and cache-TTL/provider-family metadata
+- `openrouter`: pass-through model ids, request wrappers, provider capability
+  hints, and cache-TTL policy
+- `github-copilot`: onboarding/device login, forward-compat model fallback,
+  Claude-thinking transcript hints, runtime token exchange, and usage endpoint
+  fetching
+- `openai`: GPT-5.4 forward-compat fallback, direct OpenAI transport
+  normalization, Codex-aware missing-auth hints, Spark suppression, synthetic
+  OpenAI/Codex catalog rows, thinking/live-model policy, and
+  provider-family metadata
+- `google` and `google-gemini-cli`: Gemini 3.1 forward-compat fallback and
+  modern-model matching; Gemini CLI OAuth also owns auth-profile token
+  formatting, usage-token parsing, and quota endpoint fetching for usage
+  surfaces
+- `moonshot`: shared transport, plugin-owned thinking payload normalization
+- `kilocode`: shared transport, plugin-owned request headers, reasoning payload
+  normalization, Gemini transcript hints, and cache-TTL policy
+- `zai`: GLM-5 forward-compat fallback, `tool_stream` defaults, cache-TTL
+  policy, binary-thinking/live-model policy, and usage auth + quota fetching
+- `mistral`, `opencode`, and `opencode-go`: plugin-owned capability metadata
+- `byteplus`, `cloudflare-ai-gateway`, `huggingface`, `kimi-coding`,
+  `modelstudio`, `nvidia`, `qianfan`, `synthetic`, `together`, `venice`,
+  `vercel-ai-gateway`, and `volcengine`: plugin-owned catalogs only
+- `qwen-portal`: plugin-owned catalog, OAuth login, and OAuth refresh
+- `minimax` and `xiaomi`: plugin-owned catalogs plus usage auth/snapshot logic
+
+The bundled `openai` plugin now owns both provider ids: `openai` and
+`openai-codex`.
+
+That covers providers that still fit QuantumClaw's normal transports. A provider
+that needs a totally custom request executor is a separate, deeper extension
+surface.
+
+## API key rotation
+
+- Supports generic provider rotation for selected providers.
+- Configure multiple keys via:
+  - `QUANTUMCLAW_LIVE_<PROVIDER>_KEY` (single live override, highest priority)
+  - `<PROVIDER>_API_KEYS` (comma or semicolon list)
+  - `<PROVIDER>_API_KEY` (primary key)
+  - `<PROVIDER>_API_KEY_*` (numbered list, e.g. `<PROVIDER>_API_KEY_1`)
+- For Google providers, `GOOGLE_API_KEY` is also included as fallback.
+- Key selection order preserves priority and deduplicates values.
+- Requests are retried with the next key only on rate-limit responses (for example `429`, `rate_limit`, `quota`, `resource exhausted`).
+- Non-rate-limit failures fail immediately; no key rotation is attempted.
+- When all candidate keys fail, the final error is returned from the last attempt.
+
+## Built-in providers (pi-ai catalog)
+
+QuantumClaw ships with the pi‑ai catalog. These providers require **no**
+`models.providers` config; just set auth + pick a model.
+
+### OpenAI
+
+- Provider: `openai`
+- Auth: `OPENAI_API_KEY`
+- Optional rotation: `OPENAI_API_KEYS`, `OPENAI_API_KEY_1`, `OPENAI_API_KEY_2`, plus `QUANTUMCLAW_LIVE_OPENAI_KEY` (single override)
+- Example models: `openai/gpt-5.4`, `openai/gpt-5.4-pro`
+- CLI: `quantumclaw onboard --auth-choice openai-api-key`
+- Default transport is `auto` (WebSocket-first, SSE fallback)
+- Override per model via `agents.defaults.models["openai/<model>"].params.transport` (`"sse"`, `"websocket"`, or `"auto"`)
+- OpenAI Responses WebSocket warm-up defaults to enabled via `params.openaiWsWarmup` (`true`/`false`)
+- OpenAI priority processing can be enabled via `agents.defaults.models["openai/<model>"].params.serviceTier`
+- OpenAI fast mode can be enabled per model via `agents.defaults.models["<provider>/<model>"].params.fastMode`
+- `openai/gpt-5.3-codex-spark` is intentionally suppressed in QuantumClaw because the live OpenAI API rejects it; Spark is treated as Codex-only
+
+```json5
+{
+  agents: { defaults: { model: { primary: "openai/gpt-5.4" } } },
+}
+```
+
+### Anthropic
+
+- Provider: `anthropic`
+- Auth: `ANTHROPIC_API_KEY` or `claude setup-token`
+- Optional rotation: `ANTHROPIC_API_KEYS`, `ANTHROPIC_API_KEY_1`, `ANTHROPIC_API_KEY_2`, plus `QUANTUMCLAW_LIVE_ANTHROPIC_KEY` (single override)
+- Example model: `anthropic/claude-opus-4-6`
+- CLI: `quantumclaw onboard --auth-choice token` (paste setup-token) or `quantumclaw models auth paste-token --provider anthropic`
+- Direct API-key models support the shared `/fast` toggle and `params.fastMode`; QuantumClaw maps that to Anthropic `service_tier` (`auto` vs `standard_only`)
+- Policy note: setup-token support is technical compatibility; Anthropic has blocked some subscription usage outside Claude Code in the past. Verify current Anthropic terms and decide based on your risk tolerance.
+- Recommendation: Anthropic API key auth is the safer, recommended path over subscription setup-token auth.
+
+```json5
+{
+  agents: { defaults: { model: { primary: "anthropic/claude-opus-4-6" } } },
+}
+```
+
+### OpenAI Code (Codex)
+
+- Provider: `openai-codex`
+- Auth: OAuth (ChatGPT)
+- Example model: `openai-codex/gpt-5.4`
+- CLI: `quantumclaw onboard --auth-choice openai-codex` or `quantumclaw models auth login --provider openai-codex`
+- Default transport is `auto` (WebSocket-first, SSE fallback)
+- Override per model via `agents.defaults.models["openai-codex/<model>"].params.transport` (`"sse"`, `"websocket"`, or `"auto"`)
+- Shares the same `/fast` toggle and `params.fastMode` config as direct `openai/*`
+- `openai-codex/gpt-5.3-codex-spark` remains available when the Codex OAuth catalog exposes it; entitlement-dependent
+- Policy note: OpenAI Codex OAuth is explicitly supported for external tools/workflows like QuantumClaw.
+
+```json5
+{
+  agents: { defaults: { model: { primary: "openai-codex/gpt-5.4" } } },
+}
+```
+
+### OpenCode
+
+- Auth: `OPENCODE_API_KEY` (or `OPENCODE_ZEN_API_KEY`)
+- Zen runtime provider: `opencode`
+- Go runtime provider: `opencode-go`
+- Example models: `opencode/claude-opus-4-6`, `opencode-go/kimi-k2.5`
+- CLI: `quantumclaw onboard --auth-choice opencode-zen` or `quantumclaw onboard --auth-choice opencode-go`
+
+```json5
+{
+  agents: { defaults: { model: { primary: "opencode/claude-opus-4-6" } } },
+}
+```
+
+### Google Gemini (API key)
+
+- Provider: `google`
+- Auth: `GEMINI_API_KEY`
+- Optional rotation: `GEMINI_API_KEYS`, `GEMINI_API_KEY_1`, `GEMINI_API_KEY_2`, `GOOGLE_API_KEY` fallback, and `QUANTUMCLAW_LIVE_GEMINI_KEY` (single override)
+- Example models: `google/gemini-3.1-pro-preview`, `google/gemini-3-flash-preview`
+- Compatibility: legacy QuantumClaw config using `google/gemini-3.1-flash-preview` is normalized to `google/gemini-3-flash-preview`
+- CLI: `quantumclaw onboard --auth-choice gemini-api-key`
+
+### Google Vertex and Gemini CLI
+
+- Providers: `google-vertex`, `google-gemini-cli`
+- Auth: Vertex uses gcloud ADC; Gemini CLI uses its OAuth flow
+- Caution: Gemini CLI OAuth in QuantumClaw is an unofficial integration. Some users have reported Google account restrictions after using third-party clients. Review Google terms and use a non-critical account if you choose to proceed.
+- Gemini CLI OAuth is shipped as part of the bundled `google` plugin.
+  - Enable: `quantumclaw plugins enable google`
+  - Login: `quantumclaw models auth login --provider google-gemini-cli --set-default`
+  - Note: you do **not** paste a client id or secret into `quantumclaw.json`. The CLI login flow stores
+    tokens in auth profiles on the gateway host.
+
+### Z.AI (GLM)
+
+- Provider: `zai`
+- Auth: `ZAI_API_KEY`
+- Example model: `zai/glm-5`
+- CLI: `quantumclaw onboard --auth-choice zai-api-key`
+  - Aliases: `z.ai/*` and `z-ai/*` normalize to `zai/*`
+
+### Vercel AI Gateway
+
+- Provider: `vercel-ai-gateway`
+- Auth: `AI_GATEWAY_API_KEY`
+- Example model: `vercel-ai-gateway/anthropic/claude-opus-4.6`
+- CLI: `quantumclaw onboard --auth-choice ai-gateway-api-key`
+
+### Kilo Gateway
+
+- Provider: `kilocode`
+- Auth: `KILOCODE_API_KEY`
+- Example model: `kilocode/anthropic/claude-opus-4.6`
+- CLI: `quantumclaw onboard --kilocode-api-key <key>`
+- Base URL: `https://api.kilo.ai/api/gateway/`
+- Expanded built-in catalog includes GLM-5 Free, MiniMax M2.5 Free, GPT-5.2, Gemini 3 Pro Preview, Gemini 3 Flash Preview, Grok Code Fast 1, and Kimi K2.5.
+
+See [/providers/kilocode](/providers/kilocode) for setup details.
+
+### Other bundled provider plugins
+
+- OpenRouter: `openrouter` (`OPENROUTER_API_KEY`)
+- Example model: `openrouter/anthropic/claude-sonnet-4-6`
+- Kilo Gateway: `kilocode` (`KILOCODE_API_KEY`)
+- Example model: `kilocode/anthropic/claude-opus-4.6`
+- MiniMax: `minimax` (`MINIMAX_API_KEY`)
+- Moonshot: `moonshot` (`MOONSHOT_API_KEY`)
+- Kimi Coding: `kimi-coding` (`KIMI_API_KEY` or `KIMICODE_API_KEY`)
+- Qianfan: `qianfan` (`QIANFAN_API_KEY`)
+- Model Studio: `modelstudio` (`MODELSTUDIO_API_KEY`)
+- NVIDIA: `nvidia` (`NVIDIA_API_KEY`)
+- Together: `together` (`TOGETHER_API_KEY`)
+- Venice: `venice` (`VENICE_API_KEY`)
+- Xiaomi: `xiaomi` (`XIAOMI_API_KEY`)
+- Vercel AI Gateway: `vercel-ai-gateway` (`AI_GATEWAY_API_KEY`)
+- Hugging Face Inference: `huggingface` (`HUGGINGFACE_HUB_TOKEN` or `HF_TOKEN`)
+- Cloudflare AI Gateway: `cloudflare-ai-gateway` (`CLOUDFLARE_AI_GATEWAY_API_KEY`)
+- Volcengine: `volcengine` (`VOLCANO_ENGINE_API_KEY`)
+- BytePlus: `byteplus` (`BYTEPLUS_API_KEY`)
+- xAI: `xai` (`XAI_API_KEY`)
+- Mistral: `mistral` (`MISTRAL_API_KEY`)
+- Example model: `mistral/mistral-large-latest`
+- CLI: `quantumclaw onboard --auth-choice mistral-api-key`
+- Groq: `groq` (`GROQ_API_KEY`)
+- Cerebras: `cerebras` (`CEREBRAS_API_KEY`)
+  - GLM models on Cerebras use ids `zai-glm-4.7` and `zai-glm-4.6`.
+  - OpenAI-compatible base URL: `https://api.cerebras.ai/v1`.
+- GitHub Copilot: `github-copilot` (`COPILOT_GITHUB_TOKEN` / `GH_TOKEN` / `GITHUB_TOKEN`)
+- Hugging Face Inference example model: `huggingface/deepseek-ai/DeepSeek-R1`; CLI: `quantumclaw onboard --auth-choice huggingface-api-key`. See [Hugging Face (Inference)](/providers/huggingface).
+
+## Providers via `models.providers` (custom/base URL)
+
+Use `models.providers` (or `models.json`) to add **custom** providers or
+OpenAI/Anthropic‑compatible proxies.
+
+Many of the bundled provider plugins below already publish a default catalog.
+Use explicit `models.providers.<id>` entries only when you want to override the
+default base URL, headers, or model list.
+
+### Moonshot AI (Kimi)
+
+Moonshot uses OpenAI-compatible endpoints, so configure it as a custom provider:
+
+- Provider: `moonshot`
+- Auth: `MOONSHOT_API_KEY`
+- Example model: `moonshot/kimi-k2.5`
+
+Kimi K2 model IDs:
+
+[//]: # "moonshot-kimi-k2-model-refs:start"
+
+- `moonshot/kimi-k2.5`
+- `moonshot/kimi-k2-0905-preview`
+- `moonshot/kimi-k2-turbo-preview`
+- `moonshot/kimi-k2-thinking`
+- `moonshot/kimi-k2-thinking-turbo`
+
+[//]: # "moonshot-kimi-k2-model-refs:end"
+
+```json5
+{
+  agents: {
+    defaults: { model: { primary: "moonshot/kimi-k2.5" } },
+  },
+  models: {
+    mode: "merge",
+    providers: {
+      moonshot: {
+        baseUrl: "https://api.moonshot.ai/v1",
+        apiKey: "${MOONSHOT_API_KEY}",
+        api: "openai-completions",
+        models: [{ id: "kimi-k2.5", name: "Kimi K2.5" }],
+      },
+    },
+  },
+}
+```
+
+### Kimi Coding
+
+Kimi Coding uses Moonshot AI's Anthropic-compatible endpoint:
+
+- Provider: `kimi-coding`
+- Auth: `KIMI_API_KEY`
+- Example model: `kimi-coding/k2p5`
+
+```json5
+{
+  env: { KIMI_API_KEY: "sk-..." },
+  agents: {
+    defaults: { model: { primary: "kimi-coding/k2p5" } },
+  },
+}
+```
+
+### Qwen OAuth (free tier)
+
+Qwen provides OAuth access to Qwen Coder + Vision via a device-code flow.
+The bundled provider plugin is enabled by default, so just log in:
+
+```bash
+quantumclaw models auth login --provider qwen-portal --set-default
+```
+
+Model refs:
+
+- `qwen-portal/coder-model`
+- `qwen-portal/vision-model`
+
+See [/providers/qwen](/providers/qwen) for setup details and notes.
+
+### Volcano Engine (Doubao)
+
+Volcano Engine (火山引擎) provides access to Doubao and other models in China.
+
+- Provider: `volcengine` (coding: `volcengine-plan`)
+- Auth: `VOLCANO_ENGINE_API_KEY`
+- Example model: `volcengine/doubao-seed-1-8-251228`
+- CLI: `quantumclaw onboard --auth-choice volcengine-api-key`
+
+```json5
+{
+  agents: {
+    defaults: { model: { primary: "volcengine/doubao-seed-1-8-251228" } },
+  },
+}
+```
+
+Available models:
+
+- `volcengine/doubao-seed-1-8-251228` (Doubao Seed 1.8)
+- `volcengine/doubao-seed-code-preview-251028`
+- `volcengine/kimi-k2-5-260127` (Kimi K2.5)
+- `volcengine/glm-4-7-251222` (GLM 4.7)
+- `volcengine/deepseek-v3-2-251201` (DeepSeek V3.2 128K)
+
+Coding models (`volcengine-plan`):
+
+- `volcengine-plan/ark-code-latest`
+- `volcengine-plan/doubao-seed-code`
+- `volcengine-plan/kimi-k2.5`
+- `volcengine-plan/kimi-k2-thinking`
+- `volcengine-plan/glm-4.7`
+
+### BytePlus (International)
+
+BytePlus ARK provides access to the same models as Volcano Engine for international users.
+
+- Provider: `byteplus` (coding: `byteplus-plan`)
+- Auth: `BYTEPLUS_API_KEY`
+- Example model: `byteplus/seed-1-8-251228`
+- CLI: `quantumclaw onboard --auth-choice byteplus-api-key`
+
+```json5
+{
+  agents: {
+    defaults: { model: { primary: "byteplus/seed-1-8-251228" } },
+  },
+}
+```
+
+Available models:
+
+- `byteplus/seed-1-8-251228` (Seed 1.8)
+- `byteplus/kimi-k2-5-260127` (Kimi K2.5)
+- `byteplus/glm-4-7-251222` (GLM 4.7)
+
+Coding models (`byteplus-plan`):
+
+- `byteplus-plan/ark-code-latest`
+- `byteplus-plan/doubao-seed-code`
+- `byteplus-plan/kimi-k2.5`
+- `byteplus-plan/kimi-k2-thinking`
+- `byteplus-plan/glm-4.7`
+
+### Synthetic
+
+Synthetic provides Anthropic-compatible models behind the `synthetic` provider:
+
+- Provider: `synthetic`
+- Auth: `SYNTHETIC_API_KEY`
+- Example model: `synthetic/hf:MiniMaxAI/MiniMax-M2.5`
+- CLI: `quantumclaw onboard --auth-choice synthetic-api-key`
+
+```json5
+{
+  agents: {
+    defaults: { model: { primary: "synthetic/hf:MiniMaxAI/MiniMax-M2.5" } },
+  },
+  models: {
+    mode: "merge",
+    providers: {
+      synthetic: {
+        baseUrl: "https://api.synthetic.new/anthropic",
+        apiKey: "${SYNTHETIC_API_KEY}",
+        api: "anthropic-messages",
+        models: [{ id: "hf:MiniMaxAI/MiniMax-M2.5", name: "MiniMax M2.5" }],
+      },
+    },
+  },
+}
+```
+
+### MiniMax
+
+MiniMax is configured via `models.providers` because it uses custom endpoints:
+
+- MiniMax (Anthropic‑compatible): `--auth-choice minimax-api`
+- Auth: `MINIMAX_API_KEY`
+
+See [/providers/minimax](/providers/minimax) for setup details, model options, and config snippets.
+
+### Ollama
+
+Ollama ships as a bundled provider plugin and uses Ollama's native API:
+
+- Provider: `ollama`
+- Auth: None required (local server)
+- Example model: `ollama/llama3.3`
+- Installation: [https://ollama.com/download](https://ollama.com/download)
+
+```bash
+# Install Ollama, then pull a model:
+ollama pull llama3.3
+```
+
+```json5
+{
+  agents: {
+    defaults: { model: { primary: "ollama/llama3.3" } },
+  },
+}
+```
+
+Ollama is detected locally at `http://127.0.0.1:11434` when you opt in with
+`OLLAMA_API_KEY`, and the bundled provider plugin adds Ollama directly to
+`quantumclaw onboard` and the model picker. See [/providers/ollama](/providers/ollama)
+for onboarding, cloud/local mode, and custom configuration.
+
+### vLLM
+
+vLLM ships as a bundled provider plugin for local/self-hosted OpenAI-compatible
+servers:
+
+- Provider: `vllm`
+- Auth: Optional (depends on your server)
+- Default base URL: `http://127.0.0.1:8000/v1`
+
+To opt in to auto-discovery locally (any value works if your server doesn’t enforce auth):
+
+```bash
+export VLLM_API_KEY="vllm-local"
+```
+
+Then set a model (replace with one of the IDs returned by `/v1/models`):
+
+```json5
+{
+  agents: {
+    defaults: { model: { primary: "vllm/your-model-id" } },
+  },
+}
+```
+
+See [/providers/vllm](/providers/vllm) for details.
+
+### SGLang
+
+SGLang ships as a bundled provider plugin for fast self-hosted
+OpenAI-compatible servers:
+
+- Provider: `sglang`
+- Auth: Optional (depends on your server)
+- Default base URL: `http://127.0.0.1:30000/v1`
+
+To opt in to auto-discovery locally (any value works if your server does not
+enforce auth):
+
+```bash
+export SGLANG_API_KEY="sglang-local"
+```
+
+Then set a model (replace with one of the IDs returned by `/v1/models`):
+
+```json5
+{
+  agents: {
+    defaults: { model: { primary: "sglang/your-model-id" } },
+  },
+}
+```
+
+See [/providers/sglang](/providers/sglang) for details.
+
+### Local proxies (LM Studio, vLLM, LiteLLM, etc.)
+
+Example (OpenAI‑compatible):
+
+```json5
+{
+  agents: {
+    defaults: {
+      model: { primary: "lmstudio/minimax-m2.5-gs32" },
+      models: { "lmstudio/minimax-m2.5-gs32": { alias: "Minimax" } },
+    },
+  },
+  models: {
+    providers: {
+      lmstudio: {
+        baseUrl: "http://localhost:1234/v1",
+        apiKey: "LMSTUDIO_KEY",
+        api: "openai-completions",
+        models: [
+          {
+            id: "minimax-m2.5-gs32",
+            name: "MiniMax M2.5",
+            reasoning: false,
+            input: ["text"],
+            cost: { input: 0, output: 0, cacheRead: 0, cacheWrite: 0 },
+            contextWindow: 200000,
+            maxTokens: 8192,
+          },
+        ],
+      },
+    },
+  },
+}
+```
+
+Notes:
+
+- For custom providers, `reasoning`, `input`, `cost`, `contextWindow`, and `maxTokens` are optional.
+  When omitted, QuantumClaw defaults to:
+  - `reasoning: false`
+  - `input: ["text"]`
+  - `cost: { input: 0, output: 0, cacheRead: 0, cacheWrite: 0 }`
+  - `contextWindow: 200000`
+  - `maxTokens: 8192`
+- Recommended: set explicit values that match your proxy/model limits.
+- For `api: "openai-completions"` on non-native endpoints (any non-empty `baseUrl` whose host is not `api.openai.com`), QuantumClaw forces `compat.supportsDeveloperRole: false` to avoid provider 400 errors for unsupported `developer` roles.
+- If `baseUrl` is empty/omitted, QuantumClaw keeps the default OpenAI behavior (which resolves to `api.openai.com`).
+- For safety, an explicit `compat.supportsDeveloperRole: true` is still overridden on non-native `openai-completions` endpoints.
+
+## CLI examples
+
+```bash
+quantumclaw onboard --auth-choice opencode-zen
+quantumclaw models set opencode/claude-opus-4-6
+quantumclaw models list
+```
+
+See also: [/gateway/configuration](/gateway/configuration) for full configuration examples.