Open-source LLM observability tool for lazy devs
Hi HN, we’re Dylan and Matthew, building sublingual (https://github.com/sublingual-ai/sublingual), an open-source LLM observability tool you can use with zero code changes. As developers focused on iterating and building features as fast as possible, we felt observability would’ve been a helpful tool to have, but we found existing solutions had too much overhead to set up. So we gave ourselves the challenge of building an observability tool that you can integrate without changing a single line of code in your project. How it works Run your python application as usual with…
What it does
In the maker’s words, at launch
Hi HN, we’re Dylan and Matthew, building sublingual (https://github.com/sublingual-ai/sublingual), an open-source LLM observability tool you can use with zero code changes. As developers focused on iterating and building features as fast as possible, we felt observability would’ve been a helpful tool to have, but we found existing solutions had too much overhead to set up. So we gave ourselves the challenge of building an observability tool that you can integrate without changing a single line of code in your project. How it works Run your python application as usual with the same entry point, just with `subl` in front (e.g. `subl python xyz.py` or `subl sh xyz.sh` ). That allows all python subprocesses spawned to automatically patch the clients for OpenAI, Anthropic, etc., to log the LLM input/output traces locally. We provide a local dashboard as well for you to view these traces and perform some simple evals. Designing with this paradigm of being maximally unintrusive has surprisingly resulted in some uniquely enabled features. - Automatically detecting prompt templates. If you construct your prompt with a string format like f “What is {a} + {b}?”, we can use a combination of AST and reading runtime stack frames to log the prompt template strings as well. - Tracking server requests. If you’re serving your application with Flask, Django, or FastAPI, we can keep track of all the LLM calls made in a single request as well by integrating into those server libraries. - Disentangling serving and inference logic. Logging server failures will not cause your LLM inference call to fail as well. Excited to share sublingual with everyone, and there’s still a lot more work to be done, but would love to hear whether this ease-of-integration first approach is helpful.
Does the same job
all alternatives →
OpenLIT's Zero-code LLM ObservabilityOct 2025 · ▲136Trace LLM requests + costs with OpenTelemetry monitoring
More ai this month
the category →
I trained a 125M-parameter transformer to autocomplete piano performances in real time (~108 notes/sec on an iPhone 15). The idea is basically GitHub Copilot or Tabnine, except instead of prompting it with code, you prompt it by playing a few notes on a MIDI piano. The model then continues what you played, entirely on-device. The app is free if anyone wants to try it. Happy to answer questions about the model, training, Core ML, or the many things that didn't work.
AI · 17d ago · simedw.com
Astute▲585Automate your B2B brand going viral, with new media creators
AI · 18d ago · company-app.joinastute.com
Hey HN, Henry from Cactus here! We previously released Cactus Needle, a 14MB agentic LLM for tool call, device use, and structured extraction for phones, wearables, smart homes, small robots and microcontrollers. We got really great feedback here, and have now incorporated the suggestions to release Needle 2. The whole model is a single 14MB binary that runs a full session in 28MB of RAM; 45m parameters at 2bit compression. Needle hits 500 tokens/sec decode speed on a Raspberry Pi 5, sits between 400-1,500 tokens/sec on VR devices like Meta Quest 3S and Apple Vision Pro, and ranges…
AI · 26d ago · cactuscompute.com
Launched alongside, March 2025
the whole month →
Mimic Human Research & Save Findings in AI Knowledge Base
AI · 2025 · sider.ai
