fix(projects): route auto-summary through main model

Project auto-summaries were using the 3B background model, but the
task — synthesizing a coherent paragraph over ~10 notes — is well past
what 3B can do reliably. Evidence on dev: "doging conversation
hygiene", "MCPview). MCP).", trailing stray quotes, and hallucinated
topics ("AI regulation").

Route through the user's default chat model instead. Project summary
regeneration is rare (only when a project changes) so the KV cache
eviction cost on the main model is negligible.

Title generation, tag suggestions, and RSS classification continue to
use the background model — those tasks are within what 3B handles.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

src/fabledassistant/services/projects.py

@@ -119,12 +119,17 @@ async def generate_project_summary(user_id: int, project_id: int) -> None:
             f"Recent notes:\n{note_snippets}"
         )
 
+        # Project summarization is synthesis over ~10 notes — too hard for the
+        # 3B background model, which produces incoherent, repetitive output
+        # (misspellings, broken parens, cliché filler). Route through the main
+        # chat model instead. These are rare events (only when a project
+        # changes), so the main model's KV cache eviction cost is negligible.
         from fabledassistant.services.llm import generate_completion
         from fabledassistant.config import Config
         from fabledassistant.services.settings import get_setting
         messages = [{"role": "user", "content": prompt}]
-        bg_model = await get_setting(user_id, "background_model", Config.OLLAMA_BACKGROUND_MODEL)
+        main_model = await get_setting(user_id, "default_model", Config.OLLAMA_MODEL)
-        summary = await generate_completion(messages, model=bg_model, max_tokens=400, num_ctx=2048)
+        summary = await generate_completion(messages, model=main_model, max_tokens=400, num_ctx=2048)
         if not summary:
             return