Live data from Hacker News

Viewing profile — sparacha

sparacha

HN member
Joined
Mon, Jul 01, 2024, 11:25 PM UTC
HN karma
18
Public activity
56 items

About sparacha

No profile information was provided.

Recent public activity

  1. story
    Show HN: Signals – finding the most informative agent traces without LLM judges

    Hey HN Salman, Shuguang and Adil here from Katanemo Labs (a DigitalOcean company). Wanted to introduce our latest research on agentic systems called Signals. If you've been buildin…

  2. story
    Show HN: Preference-aware routing for OpenClaw via Plano

    Hey HN! OpenClaw is notorious about its token usage, and for many the price of Opus 4.6 can be cost prohibitive for personal projects. The usual workaround is “just switch to a che…

  3. story
  4. story
  5. story
  6. story
  7. comment
    Comment #45455933

    Hi HN — we’re the team behind Arch-Router [1], A 1.5B preference-aligned LLM router that guides model selection by matching queries to user-defined domains (e.g., travel) or action…

  8. story
  9. comment
    Comment #44772050

    Hey! I built this. AMA. The model router is built into the proxy layer here: https://github.com/katanemo/archgw

  10. comment
    Comment #44652881

    But you can also use tokens to implement routing decisions in a proxy. You can make RBAC natively available to all agents outside code. The incremental feature work in code vs an o…

  11. comment
    Comment #44651074

    There is liteLLM, OpenRouter, Arch (although that’s an edge/service proxy for agents) and now this. We all need a new problem to solve

  12. comment
    Comment #44546887

    That’s an example of what the edge component could do. Did you give the preference-based automatic routing a try?

  13. story
  14. comment
    Comment #44439016

    RouteLLM is essentially a benchmark-driven approach. Their framework chooses between a weak and a strong model and helps developers optimize for a metric called APGR (Average Perfo…

  15. comment
    Comment #44438708

    https://news.ycombinator.com/item?id=44436031

  16. comment
    Comment #44438547

    Arch is developer friendly, but designed for enterprise-grade customers in mind. The core contributors of Envoy redesigned the proxy substrate to handle prompts - offering somethin…

  17. comment
    Comment #44438150

    Overall performance degrades from 93.17 -> 92.99 with a quantized version

  18. comment
    Comment #44438075

    Can you share more about your evaluation setup? I would love to see the specific usage pattern as we have tested our model against smaller LLMs and foundational models and our resu…

  19. comment
    Comment #44437364

    yes - we have already published a quantized version here: https://huggingface.co/katanemo/Arch-Router-1.5B.gguf . The performance difference with a quant version is negligible. I'l…

  20. comment
    Comment #44436070

    Hi HN! I am one of the co-authors of the paper. If there are any questions about our approach, I would love to answer them.

  21. story
  22. story
  23. story
  24. story
  25. comment
    Comment #43731567

    good one!