#13401: fix: slug generator should use agent's primary model instead of hardcoded anthropic default

Fixes #13396 ## Problem When using OpenClaw with **Bedrock-only** configuration (no Anthropic API key), the slug generator fails with: ``` FailoverError: No API key found for provider "anthropic" ``` This occurs because `runEmbeddedPiAgent` (used by slug generator) falls back to the hardcoded default `anthropic/claude-opus-4-6` instead of respecting the agent's configured primary model. ## Solution The slug generator now resolves models in this order: 1. Explicit `slugGenerator.model` config (if provided) 2. Agent's primary model (from agent config) 3. Falls back to `runEmbeddedPiAgent`'s default resolution ## Changes - Added `slugGenerator.model` field to `OpenClawConfig` type - Modified `llm-slug-generator.ts` to resolve model using `resolveAgentModelPrimary` - Pass resolved model to `runEmbeddedPiAgent` ## Testing Tested with Bedrock-only configuration: - Agent primary model: `amazon-bedrock/us.anthropic.claude-haiku-4-5-20251001-v1:0` - No Anthropic API key configured - Slug generator now uses Bedrock model successfully ## Impact This fixes startup failures for users with: - Bedrock-only setup - Azure OpenAI-only setup - Any non-Anthropic provider configuration <!-- greptile_comment --> <h2>Greptile Overview</h2> <h3>Greptile Summary</h3> This PR updates the LLM-based session slug generator to pick a model from configuration instead of relying on `runEmbeddedPiAgent`’s hardcoded defaults (which can force the `anthropic` provider and break “Bedrock-only”/non‑Anthropic setups). It adds an optional `slugGenerator.model` field to `OpenClawConfig`, then in `src/hooks/llm-slug-generator.ts` resolves the effective model as `cfg.slugGenerator.model` first, otherwise the default agent’s primary model via `resolveAgentModelPrimary(cfg, agentId)`, and passes that through to `runEmbeddedPiAgent`. Overall, the change is small and consistent with existing model selection helpers (`resolveAgentModelPrimary`). I did not find any new functional issues in the changed code paths. <h3>Confidence Score: 5/5</h3> - This PR is safe to merge with minimal risk. - Changes are limited in scope, use an existing model-resolution helper, and correctly pass an optional `model` override down to `runEmbeddedPiAgent` without altering other execution logic. - No files require special attention. <!-- greptile_other_comments_section --> _{(2/5) Greptile learns from your feedback when you react with thumbs up/down!} <!-- /greptile_comment -->