AdditionalWeb107

• I built a Claude Code Router TUI

  Posted by AdditionalWeb107@reddit | LocalLLaMA | View on Reddit | 2 comments

• Signals – finding the most informative agent traces without LLM judges (arxiv.org)

  Posted by AdditionalWeb107@reddit | LocalLLaMA | View on Reddit | 0 comments

• Plano reaches 5K GH stars as I continue to help devs build agents locally

  Posted by AdditionalWeb107@reddit | LocalLLaMA | View on Reddit | 2 comments

• Plano 0.4.3 ⭐️ Filter Chains via MCP and OpenRouter Integration

  Posted by AdditionalWeb107@reddit | LocalLLaMA | View on Reddit | 0 comments

• Just launched Plano v0.4 - a unified data plane supporting polyglot AI development

  Posted by AdditionalWeb107@reddit | Python | View on Reddit | 1 comments

• Is it one big agent, or sub-agents?

  Posted by AdditionalWeb107@reddit | LocalLLaMA | View on Reddit | 8 comments

• I built Plano(A3B): most efficient LLMs for agent orchestration that exceed frontier model perf

  Posted by AdditionalWeb107@reddit | LocalLLaMA | View on Reddit | 35 comments

• archgw 0.3.20 - gutted out 500Mbs worth of python dependenices in the req path.

  Posted by AdditionalWeb107@reddit | LocalLLaMA | View on Reddit | 0 comments

• archgw 0.3.20 - 500MBs of python dependencies gutted out - faster, leaner proxy server for agents.

  Posted by AdditionalWeb107@reddit | Python | View on Reddit | 0 comments

• From OSS to $1M in contract value from one customer. How? Forward deployed engineers.

  Posted by AdditionalWeb107@reddit | ExperiencedDevs | View on Reddit | 7 comments

• Arch-Router: The first (and fastest) LLM router that can align to your usage preferences.

  Posted by AdditionalWeb107@reddit | LocalLLaMA | View on Reddit | 25 comments

• 🚀 HuggingFaceChat Omni: Dynamic policy-baed routing to 115+ LLMs

  Posted by AdditionalWeb107@reddit | LocalLLaMA | View on Reddit | 5 comments

• Preference-aware routing to local LLMs for Claude Code 2.0

  Posted by AdditionalWeb107@reddit | LocalLLaMA | View on Reddit | 0 comments

• Claude Code 2.0 Router - Access Ollama-based LLMs and align automatic routing to preferences, not benchmarks.

  Posted by AdditionalWeb107@reddit | LocalLLaMA | View on Reddit | 10 comments

• ArchGW 🚀 - Use Ollama-based LLMs with Anthropic client (release 0.3.13)

  Posted by AdditionalWeb107@reddit | LocalLLaMA | View on Reddit | 4 comments

• Examining the 72988 character long Claude Code Prompt

  Posted by AdditionalWeb107@reddit | LocalLLaMA | View on Reddit | 1 comments

• ArchGW 0.3.12 🚀 Model aliases: allow clients to use friendly, semantic names and swap out underlying models without changing application code.

  Posted by AdditionalWeb107@reddit | LocalLLaMA | View on Reddit | 0 comments

• real engineering work in AI gets paid. We went from a demo to a $500k in total contract value building "networking" for agents.

  Posted by AdditionalWeb107@reddit | ExperiencedDevs | View on Reddit | 19 comments

• The outer loop vs. the inner loop of agents. A simple mental model to evolve the stack quickly

  Posted by AdditionalWeb107@reddit | Python | View on Reddit | 7 comments

• The outer loop vs. the inner loop of agents. A mental model to evolve the agent stack quickly and push to production faster.

  Posted by AdditionalWeb107@reddit | LocalLLaMA | View on Reddit | 6 comments

• Detecting Hallucinations in LLM Function Calling with Entropy (Part 2)

  Posted by AdditionalWeb107@reddit | LocalLLaMA | View on Reddit | 3 comments

• GPT-5 Style Router, but for any LLM including local.

  Posted by AdditionalWeb107@reddit | LocalLLaMA | View on Reddit | 64 comments

• My team has to stop this "let me grab this AI framework" mentality and think about overall system design.

  Posted by AdditionalWeb107@reddit | ExperiencedDevs | View on Reddit | 17 comments

• Connect 3rd party SaaS tools to your agentic apps - ArchGW 0.2.1 🚀 adds support for bearer authorization for upstream APIs for function calling scenarios.

  Posted by AdditionalWeb107@reddit | LocalLLaMA | View on Reddit | 4 comments

• [Research] Align LLM routing to task preferences, not benchmarks - with a fast 1.5B model

  Posted by AdditionalWeb107@reddit | ExperiencedDevs | View on Reddit | 2 comments

• Handle follow-up or clarifying questions in RAG scenarios (with ease)

  Posted by AdditionalWeb107@reddit | LocalLLaMA | View on Reddit | 3 comments

• Strategies for handling transient Server-Sent Events (SSE) from LLM responses

  Posted by AdditionalWeb107@reddit | ExperiencedDevs | View on Reddit | 12 comments

• Strategies for handling transient Server-Sent Events (SSE) from LLM responses

  Posted by AdditionalWeb107@reddit | LocalLLaMA | View on Reddit | 3 comments

• Finally an LLM router that thinks like an engineer

  Posted by AdditionalWeb107@reddit | programming | View on Reddit | 9 comments

• Vibe coding RouteGPT - a chrome extension aligns model routing to my preferences, powered by a small but powerful LLM.

  Posted by AdditionalWeb107@reddit | LocalLLaMA | View on Reddit | 0 comments

• Finally, an LLM Router That Thinks Like an Engineer - And Its Local

  Posted by AdditionalWeb107@reddit | LocalLLaMA | View on Reddit | 3 comments

• An alternative to semantic or benchmark-based routing: A preference-aligned router model

  Posted by AdditionalWeb107@reddit | LocalLLaMA | View on Reddit | 6 comments

• Arch-Agent Family of LLMs - Designed for fast, multi-step agent orchestration.

  Posted by AdditionalWeb107@reddit | LocalLLaMA | View on Reddit | 7 comments

• From Arch-Function to Arch-Agent. Designed for fast multi-step, multi-turn workflow orchestration in agents.

  Posted by AdditionalWeb107@reddit | LocalLLaMA | View on Reddit | 16 comments

• Semantic routing and caching doesn't work - task specific LLMs (TLMs) ftw!

  Posted by AdditionalWeb107@reddit | LocalLLaMA | View on Reddit | 9 comments

• Core infrastructure patterns implemented in AI coding frameworks - will come home to roost

  Posted by AdditionalWeb107@reddit | ExperiencedDevs | View on Reddit | 28 comments

• ArchGW 0.2.8 is out 🚀 - unifying repeated "low-level" functionality in building LLM apps via a local proxy.

  Posted by AdditionalWeb107@reddit | LocalLLaMA | View on Reddit | 16 comments

• How to load a 4-bit quantized 1.5B parameter LLM in the browser?

  Posted by AdditionalWeb107@reddit | LocalLLaMA | View on Reddit | 8 comments

• Using a local runtime to run models for an open source project vs. HF transformers library

  Posted by AdditionalWeb107@reddit | LocalLLaMA | View on Reddit | 2 comments

• I think triage agents should run "out-of-process". Here's why.

  Posted by AdditionalWeb107@reddit | LocalLLaMA | View on Reddit | 16 comments

• Why are people rushing to programming frameworks for agents?

  Posted by AdditionalWeb107@reddit | LocalLLaMA | View on Reddit | 17 comments

• I built an open-source AI-native proxy for LLM agents

  Posted by AdditionalWeb107@reddit | programming | View on Reddit | 4 comments

• Arch. The AI-native proxy server that handles the low-level application logic for agents

  Posted by AdditionalWeb107@reddit | Python | View on Reddit | 0 comments

• Arch-Function-Chat Trending #1 on HuggingFace!

  Posted by AdditionalWeb107@reddit | LocalLLaMA | View on Reddit | 10 comments

• When vibe coding no longer vibes back

  Posted by AdditionalWeb107@reddit | LocalLLaMA | View on Reddit | 75 comments

• Not GPT-4, but a 3B Function Calling LLM that can chat to clarify tools calls

  Posted by AdditionalWeb107@reddit | LocalLLaMA | View on Reddit | 13 comments

• Arch-Function-Chat (1B/3B/7B) - Device friendly, family of fast LLMs for function calling scenarios now trained to chat.

  Posted by AdditionalWeb107@reddit | LocalLLaMA | View on Reddit | 7 comments

• Who is building MCP servers - and how are you thinking about exposure risks?

  Posted by AdditionalWeb107@reddit | LocalLLaMA | View on Reddit | 13 comments

• How I adapted a 1B function calling LLM for fast routing and agent hand -off scenarios in a framework agnostic way.

  Posted by AdditionalWeb107@reddit | LocalLLaMA | View on Reddit | 11 comments

• I built a small (function calling) LLM that packs a big punch; integrated in an open source gateway for agentic apps

  Posted by AdditionalWeb107@reddit | LocalLLaMA | View on Reddit | 75 comments