feat(rag): add --embed-model flag with nomic-embed-text default

src/tai/cli.py

@@ -12,7 +12,7 @@ from rich.panel import Panel
 from rich.rule import Rule
 from rich.text import Text
 
-from tai.ai_client import DEFAULT_AI_HOST, DEFAULT_MODEL, AIClient, AIConfig
+from tai.ai_client import DEFAULT_AI_HOST, DEFAULT_EMBED_MODEL, DEFAULT_MODEL, AIClient, AIConfig
 from tai.ai_guardrails import validate_ai_response
 from tai.collectors import CollectionReport, collect_from_plan
 from tai.input_parser import InputValidationError, build_request
@@ -110,6 +110,13 @@ def run(
             help="Disable RAG; send full diagnostics to AI instead of retrieved chunks.",
         ),
     ] = False,
+    embed_model: Annotated[
+        str,
+        typer.Option(
+            "--embed-model",
+            help="Embedding model for RAG. Must be pulled in Ollama on the AI host.",
+        ),
+    ] = DEFAULT_EMBED_MODEL,
 ) -> None:
     """Start an interactive troubleshooting session scaffold."""
     try:
@@ -145,7 +152,7 @@ def run(
     if not (probe or collect or analyze or interactive):
         return  # nothing SSH-related requested
 
-    ai_config = AIConfig(host=ai_host, model=model, api_key=ai_key)
+    ai_config = AIConfig(host=ai_host, model=model, api_key=ai_key, embed_model=embed_model)
     logger = SessionLogger.create(log_file) if log_file else None
     if analyze or interactive:
         console.print(f"[cyan]AI:[/cyan] {AIClient(ai_config).summary()}")