feat(runtime): add reasoning toggle for ollama

docs/config-reference.md

@@ -50,6 +50,18 @@ Notes:
 - Setting `max_tool_iterations = 0` falls back to safe default `10`.
 - If a channel message exceeds this value, the runtime returns: `Agent exceeded maximum tool iterations (<value>)`.
 
+## `[runtime]`
+
+| Key | Default | Purpose |
+|---|---|---|
+| `reasoning_enabled` | unset (`None`) | Global reasoning/thinking override for providers that support explicit controls |
+
+Notes:
+
+- `reasoning_enabled = false` explicitly disables provider-side reasoning for supported providers (currently `ollama`, via request field `think: false`).
+- `reasoning_enabled = true` explicitly requests reasoning for supported providers (`think: true` on `ollama`).
+- Unset keeps provider defaults.
+
 ## `[gateway]`
 
 | Key | Default | Purpose |

docs/providers-reference.md

@@ -67,6 +67,21 @@ credential is not reused for fallback providers.
 - Cross-region inference profiles supported (e.g., `us.anthropic.claude-*`).
 - Model IDs use Bedrock format: `anthropic.claude-sonnet-4-6`, `anthropic.claude-opus-4-6-v1`, etc.
 
+### Ollama Reasoning Toggle
+
+You can control Ollama reasoning/thinking behavior from `config.toml`:
+
+```toml
+[runtime]
+reasoning_enabled = false
+```
+
+Behavior:
+
+- `false`: sends `think: false` to Ollama `/api/chat` requests.
+- `true`: sends `think: true`.
+- Unset: omits `think` and keeps Ollama/model defaults.
+
 ### Kimi Code Notes
 
 - Provider ID: `kimi-code`