WildEdge

CI Python Versions Tested on Linux Tested on macOS Tested on Windows Coverage

ML inference monitoring for Python: on-device and remote models. Tracks latency, tokens, errors, and model metadata with no code changes, plus traces for agent pipelines.

uv add wildedge-sdk

Without a DSN everything is a silent no-op, so dev and CI need no configuration.

Drop wildedge run in front of your existing command. WildEdge instruments the runtime before your code starts. No SDK calls required in user code.

WILDEDGE_DSN="https://<secret>@ingest.wildedge.dev/<key>" \
wildedge run --integrations timm -- python app.py

Validate your environment before deploying. --send-test-event proves the whole pipeline end to end by sending one span event and reporting the ingest response (exit codes: 0 pass, 1 config failure, 2 connectivity failure):

wildedge doctor --integrations all --send-test-event

Useful flags:

Use the SDK when you can't wrap the process, or when you want traces and spans around your own code. wildedge run covers everything else.

import wildedge
from openai import OpenAI
from transformers import pipeline
wildedge.init(integrations=["transformers", "openai"]) # optional under `wildedge run`
local = pipeline("text-generation", model="HuggingFaceTB/SmolLM2-360M-Instruct")
remote = OpenAI()
with wildedge.trace(run_id="run-1"):
 with wildedge.span(kind="agent_step", name="draft"):
 draft = local(prompt)[0]["generated_text"]
 with wildedge.span(kind="agent_step", name="refine"):
 remote.chat.completions.create(
 model="gpt-4o-mini",
 messages=[{"role": "user", "content": draft}],
 )

See Deployment for client lifecycle and the full contract.

On-device

Remote models

Fireworks, Together, Baseten, xAI, Mistral, Groq and other OpenAI-compatible hosts work through the openai integration — pass their base_url. See Providers for every recognized endpoint.

Using httpx, requests or urllib directly instead of the openai / anthropic client libraries? Wrap the call in wildedge.llm_api() to get the same events: llm_api_example.py.

Hub tracking

Pass hubs= to track model download provenance. Hubs are framework-agnostic and can be combined with any integration.

For unsupported frameworks, see Manual tracking.

For advanced options (batching, queue tuning, dead-letter storage, attachments), see Configuration.

Using WildEdge in your project? Open a PR to add it to the list.

By default the SDK transmits only anonymized telemetry, never raw model inputs or outputs. The one exception is opt-in attachments (attachments_enabled), which upload raw bytes you explicitly pass in.

Report security and privacy issues to: support@wildedge.dev

• Deployment guide
• Manual tracking
• LLM API tracking
• Compatibility Matrix
• Changelog
• License

Each GitHub release ships llms.txt and llms-full.txt: the full documentation for that exact version in one file, built for AI assistants.