`loupe-ai` — Python SDK

Forensic observability + circuit attribution for Python LLM agents. Zero-code capture of every LLM call to append-only JSONL traces, with a local dashboard, TUI, and SAE-based interpretability on top.

Install

pip install loupe-ai                  # CLI + dashboard + httpx capture, batteries included
pip install 'loupe-ai[interp]'        # adds the real SAE attribution backend (~150MB)
pip install 'loupe-ai[langgraph]'     # framework SDK extras: langgraph, anthropic, openai,
                                      # pydantic-ai, llama-index, dspy, crewai, autogen, openhands

From a clone of the repo: pip install -e '.[dev]'.

Quickstart

from loupe import trace, record_step
from loupe.integrations import patch_all

patch_all()                                  # auto-capture any installed LLM SDK

@trace(framework="anthropic")
async def my_agent(query: str):
    record_step("plan", "compose request")
    # ...your real agent code; LLM calls captured automatically
    return result

await my_agent("refactor auth.py")
# trace saved to ~/.loupe/traces/{run_id}.jsonl

View traces locally:

loupe ui              # opens http://localhost:7860 — live SSE dashboard
loupe list            # terminal table of every run
loupe attribute <id>  # SAE circuit attribution per llm-call step
loupe cluster         # find features that recur across tagged failures
loupe replay <id>     # re-invoke a captured run for reproducibility testing

See SPEC.md for the wire format and ARCHITECTURE.md for the layering.

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

loupe_ai-0.0.83.tar.gz (299.2 kB view details)

Uploaded Jul 16, 2026 Source

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

The dropdown lists show the available interpreters, ABIs, and platforms. Enable javascript to be able to filter the list of wheel files.

loupe_ai-0.0.83-py3-none-any.whl (245.2 kB view details)

Uploaded Jul 16, 2026 Python 3

File details

Details for the file loupe_ai-0.0.83.tar.gz.

File metadata

Download URL: loupe_ai-0.0.83.tar.gz
Upload date: Jul 16, 2026
Size: 299.2 kB
Tags: Source
Uploaded using Trusted Publishing? Yes
Uploaded via: twine/6.1.0 CPython/3.13.12

File hashes

Hashes for loupe_ai-0.0.83.tar.gz
Algorithm	Hash digest
SHA256	`6bab109ae7b868916a31c7dfb1789588cf6b968f1e3fbb9042cb2bc193df31c0`
MD5	`0027397425a4c40d5eb57cafa1305c55`
BLAKE2b-256	`a2b4da96e0abe0e736ee206761260af747a679780805327ba95906f868ed92cf`

See more details on using hashes here.

Provenance

The following attestation bundles were made for loupe_ai-0.0.83.tar.gz:

Publisher: release.yml on YashwanthKamireddi/loupe

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

Statement:
- Statement type: https://in-toto.io/Statement/v1
- Predicate type: https://docs.pypi.org/attestations/publish/v1
- Subject name: loupe_ai-0.0.83.tar.gz
- Subject digest: 6bab109ae7b868916a31c7dfb1789588cf6b968f1e3fbb9042cb2bc193df31c0
- Sigstore transparency entry: 2186764475
- Sigstore integration time: Jul 16, 2026
Source repository:
- Permalink: YashwanthKamireddi/loupe@3c09bbc1cb3e6fdd82db90ed73cbf17fd89e9db7
- Branch / Tag: refs/tags/v0.0.83
- Owner: https://github.com/YashwanthKamireddi
- Access: public
Publication detail:
- Token Issuer: https://token.actions.githubusercontent.com
- Runner Environment: github-hosted
- Publication workflow: release.yml@3c09bbc1cb3e6fdd82db90ed73cbf17fd89e9db7
- Trigger Event: push