Feature: AI-powered model recommendations for cost optimization

[GalDayan]

Feature Request (from @sivan)

Add intelligent model recommendation based on session analysis to optimize token usage and LLM costs.

Requested behavior:
Clawatch should:

1. Analyze recent sessions and their patterns
2. Identify opportunities where a cheaper/smaller model could handle the workload
3. Suggest specific model alternatives that would reduce costs without impacting quality
4. Show potential savings estimate

Example use cases:

• Simple queries that don't need GPT-4 → suggest GPT-3.5-turbo
• Repetitive tasks → suggest cheaper models
• Sessions with low complexity → recommend cost-effective alternatives

Value:

• Help users optimize their LLM spending automatically
• Provide actionable insights on model selection
• Reduce operational costs without manual analysis

Context:
Requested by Sivan during internal testing feedback session.

Related Slack thread:
https://genwayhq.slack.com/archives/C0AFHCR4933/p1773248511101549

---

Auto-filed by Dor (PM) on behalf of team feedback

[GalDayan]

User validation from Sivan:

"Yes I would change. I am tired of pausing tasks because I'm out of tokens."

Key insight: This isn't just cost optimization - it's a productivity blocker. Users are hitting token limits and having to pause work.

Reframe:

• Problem: Users burn expensive tokens on simple tasks, hit limits, can't finish real work
• Solution: Smart model routing to stretch token budgets further
• Impact: More work gets done with same token budget

Suggested priority: P1 (unblocks productivity, not just cost savings)

cc <@U04UUS3T5K5> - this might be higher priority than initially thought