Capability · Framework — observability
Log10
Log10 is a commercial LLM observability platform founded by ex-Dropbox engineers, focused on debugging production LLM apps. It offers automated feedback (their AutoFeedback model scores calls without human labels), a Python/TS SDK that one-lines on top of OpenAI / Anthropic clients, and self-hosted deployments for regulated customers. Often compared to Helicone and Langfuse.
Framework facts
- Category
- observability
- Language
- Python / TypeScript
- License
- Proprietary SaaS / Self-hosted
- Repository
- https://github.com/log10-io/log10
Install
pip install log10-ioQuickstart
from log10.load import log10
import openai
log10(openai)
resp = openai.chat.completions.create(
model='gpt-4o',
messages=[{'role':'user','content':'hi'}]
)Alternatives
- Helicone — OSS proxy
- Langfuse — OSS observability
- Arize Phoenix — OSS
Frequently asked questions
What is AutoFeedback?
A Log10-trained judge that grades completions for hallucination, relevance, and tone without human labels. Useful for flagging regressions on live traffic without having humans read every log.
Is Log10 open source?
The Python SDK is Apache-2.0; the server is proprietary SaaS with a self-hosted enterprise tier.
Sources
- Log10 docs — accessed 2026-04-20
- Log10 GitHub — accessed 2026-04-20