sparacha

6 ms·

1.
Show HN: Signals – finding the most informative agent traces without LLM judges(arxiv.org)
3 points by sparacha 6mo ago|0 comments
2.
Show HN: Preference-aware routing for OpenClaw via Plano(github.com)
1 points by sparacha 7mo ago|0 comments
3.
The two agentic loops – how to build and scale agentic apps(planoai.dev)
1 points by sparacha 8mo ago|0 comments
4.
by sparacha 1y ago
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 types (e.g., image editing). Offering a practical mechanism
5.
by sparacha 1y ago
Hey! I built this. AMA. The model router is built into the proxy layer here: https://github.com/katanemo/archgw
6.
by sparacha 1y ago
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 out of process server is the trade off. One gets you going su
7.
by sparacha 1y ago
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
8.
by sparacha 1y ago
That’s an example of what the edge component could do. Did you give the preference-based automatic routing a try?
9.
Arch-Function LLMs promise lightning-fast agentic AI for enterprise workflows(venturebeat.com)
2 points by sparacha 1y ago|0 comments
10.
by sparacha 1y ago
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 Performance Gap Recovered) — a measure of how much of the stronge
11.
by sparacha 1y ago
https://news.ycombinator.com/item?id=44436031
12.
by sparacha 1y ago
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 something that is battle tested in terms of resiliency, speed, and d
13.
by sparacha 1y ago
Overall performance degrades from 93.17 -> 92.99 with a quantized version
14.
by sparacha 1y ago
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 results show things differently. Of course, routing policies sho
15.
by sparacha 1y ago
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'll run another analysis and update
16.
by sparacha 1y ago
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.
17.
by sparacha 1y ago
good one!
18.
by sparacha 1y ago
build MCP servers that get plugged into different apps like Claude desktop. But what if you want to go from FastAPI functions and build your own agentic app - added bonus have common tool calls be blazing fast. Just updated https:/&#x
19.
The rise of intelligent infrastructure for LLM applications(archgw.com)
3 points by sparacha 2y ago|1 comments
20.
by sparacha 2y ago
Opik is an evaluation tool first. Arch is a proxy server built on top of Envoy so it borrows from a very robust observability source. They both are complimentary in many ways
21.
by sparacha 2y ago
Speed. And separately, instruction fine-tuning an LLM for a specialized task like function calling or guardrails == better performance. Even Anthropic and other model providers suggest you separate tasks for LLMs to improve overall user exp
22.
by sparacha 2y ago
Woud love feedback. See if it is useful, or what adaptations would make it useful.
23.
by sparacha 2y ago
Leaderboards are getting harder and harder as a decision tool. What does it mean to be better 0.7% or 1.6%. How does that help me? Is higher always better? What are the trade offs? Evals continue be the hardest most important parts of LLMs
24.
Show HN: ArchGW – An open-source intelligent proxy server for prompts(github.com)
39 points by sparacha 2y ago|7 comments
25.
ArchGW: Open-source, AI-native (edge and LLM) proxy for prompt traffic(github.com)
2 points by sparacha 2y ago|1 comments
26.
by sparacha 2y ago
Why We Built ArchGW? Traditional application architectures separate routing, security, and observability from business logic - so that developers can move faster without the tax of reinventing the wheel. LLM applications should be no differ
27.
by sparacha 2y ago
OP - congrats on this launch! I think there are a lot of useful bits here, especially the communication abstractions for MCP and having clear recipes to implement patterns mentioned by Anthropic team is a big plus.
28.
by sparacha 2y ago
If you are building a state machine then you should use tools that enable you to do that (temporal, langgraph, etc) by orchestrating multple LLM calls - in that case archgw offers intelligent routing to your COT agent and enables you to tra
29.
by sparacha 2y ago
Arch-Function our fast, open source LLM does most of the heavy lifting on extracting parameter values from a user prompt, gathering more information from the user, determining the right set of functions to call downstream. Its designed for
30.
Show HN: Arch GW – Distributed gateway for agents, engineered with small LLMs(docs.archgw.com)
7 points by sparacha 2y ago|0 comments

More ›