From 60f39f83828dbc18f020123e77fa4e305314bc7c Mon Sep 17 00:00:00 2001
From: can1357 <me@can.ac>
Date: Sat, 3 Jan 2026 03:40:24 +0100
Subject: [PATCH] feat(coding-agent): added CLI args and env vars for model
 role overrides

- Added --smol and --slow CLI args for ephemeral model role overrides
- Added PI_SMOL_MODEL and PI_SLOW_MODEL env var support
- Overrides are applied at runtime without persisting to settings.json
---
 packages/coding-agent/CHANGELOG.md    |  2 ++
 packages/coding-agent/src/cli/args.ts |  8 ++++++++
 packages/coding-agent/src/main.ts     | 11 +++++++++++
 3 files changed, 21 insertions(+)

diff --git a/packages/coding-agent/CHANGELOG.md b/packages/coding-agent/CHANGELOG.md
index 0e740886c..aed24a98b 100644
--- a/packages/coding-agent/CHANGELOG.md
+++ b/packages/coding-agent/CHANGELOG.md
@@ -18,6 +18,8 @@
 - `pi/<role>` model aliases in Task tool agent definitions (e.g., `model: pi/smol, haiku, flash, mini`)
 - Small model auto-discovery using priority chain: haiku > flash > mini
 - Slow model auto-discovery using priority chain: gpt-5.2-codex > codex > gpt > opus > pro
+- CLI args for model roles: `--smol <model>` and `--slow <model>` (ephemeral, not persisted)
+- Env var overrides: `PI_SMOL_MODEL` and `PI_SLOW_MODEL`
 - Title generation now uses configured small model from settings
 
 - LSP diagnostics on edit: Edit tool can now return LSP diagnostics after editing code files. Disabled by default to avoid noise during multi-edit sequences. Enable via `lsp.diagnosticsOnEdit` setting.
diff --git a/packages/coding-agent/src/cli/args.ts b/packages/coding-agent/src/cli/args.ts
index d291d7148..96fa4ec8f 100644
--- a/packages/coding-agent/src/cli/args.ts
+++ b/packages/coding-agent/src/cli/args.ts
@@ -12,6 +12,8 @@ export type Mode = "text" | "json" | "rpc";
 export interface Args {
 	provider?: string;
 	model?: string;
+	smol?: string;
+	slow?: string;
 	apiKey?: string;
 	systemPrompt?: string;
 	appendSystemPrompt?: string;
@@ -69,6 +71,10 @@ export function parseArgs(args: string[]): Args {
 			result.provider = args[++i];
 		} else if (arg === "--model" && i + 1 < args.length) {
 			result.model = args[++i];
+		} else if (arg === "--smol" && i + 1 < args.length) {
+			result.smol = args[++i];
+		} else if (arg === "--slow" && i + 1 < args.length) {
+			result.slow = args[++i];
 		} else if (arg === "--api-key" && i + 1 < args.length) {
 			result.apiKey = args[++i];
 		} else if (arg === "--system-prompt" && i + 1 < args.length) {
@@ -148,6 +154,8 @@ ${chalk.bold("Usage:")}
 ${chalk.bold("Options:")}
   --provider <name>              Provider name (default: google)
   --model <id>                   Model ID (default: gemini-2.5-flash)
+  --smol <id>                    Small/fast model for lightweight tasks (or PI_SMOL_MODEL env)
+  --slow <id>                    Slow/reasoning model for thorough analysis (or PI_SLOW_MODEL env)
   --api-key <key>                API key (defaults to env vars)
   --system-prompt <text>         System prompt (default: coding assistant prompt)
   --append-system-prompt <text>  Append text or file contents to the system prompt
diff --git a/packages/coding-agent/src/main.ts b/packages/coding-agent/src/main.ts
index 2848f3a75..beb0e08ac 100644
--- a/packages/coding-agent/src/main.ts
+++ b/packages/coding-agent/src/main.ts
@@ -356,6 +356,17 @@ export async function main(args: string[]) {
 
 	const settingsManager = SettingsManager.create(cwd);
 	time("SettingsManager.create");
+
+	// Apply model role overrides from CLI args or env vars (ephemeral, not persisted)
+	const smolModel = parsed.smol ?? process.env.PI_SMOL_MODEL;
+	const slowModel = parsed.slow ?? process.env.PI_SLOW_MODEL;
+	if (smolModel || slowModel) {
+		const roleOverrides: Record<string, string> = {};
+		if (smolModel) roleOverrides.small = smolModel;
+		if (slowModel) roleOverrides.slow = slowModel;
+		settingsManager.applyOverrides({ modelRoles: roleOverrides });
+	}
+
 	initTheme(settingsManager.getTheme(), isInteractive);
 	time("initTheme");