> ## Documentation Index
> Fetch the complete documentation index at: https://docs.impact.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Python troubleshooting

> Diagnose Python capture and delivery and verify the matching Trace.

## Confirm the matching Trace

Initialize before client construction, run a representative application operation, and flush or shut down. In
the intended Impact Product and Environment, find the Trace for your service and match its `impact.runtime.id`
to `runtime.status().runtime_id`. Its `impact.sdk.version` identifies the SDK. Check the actual input, output,
model, tool effect and usage that the operation exposes. The runtime ID is generated by the SDK, appears on
SDK-owned evidence and Impact-routed copies, and is read-only.

A successful `flush()` establishes local exporter completion. The route setup report distinguishes a running
SDK awaiting traffic from local delivery failures. Keep supplied route and Environment revisions current and
use the representative operation to verify readable data.

## Diagnose missing telemetry

1. Use `runtime.status()` to confirm the installed version, trace and log provider ownership, integration
   activation, limitations and delivery counts. Check the [integration catalogue](https://docs.impact.ai/sdk/python/integrations.md) to
   match the exact client, operation and capture owner. Fix a disabled or missing integration, incompatible
   client, late `init()` or customer-routed destination before retrying. Keep existing application
   instrumentation in place.
2. If only one field is missing, check its [capture boundary and content settings](https://docs.impact.ai/sdk/python/capture.md).
   Consumed stream output, native usage and omitted or partial facets depend on the actual provider operation
   and stream end.
3. If startup or export fails, inspect the content-free `runtime.status().cause`, `.action` and trace/log
   counters. On direct OTLP, confirm the exact `/v1/traces` URL and an active ingest key for that receiver.
   Strict startup checks remote Product-key discovery, while direct OTLP checks receiver authorization at
   export. Restart or redeploy after changing the key or endpoint of an active runtime. For a discovery route,
   confirm its application API origin and required grant. After an unavailable startup settles, correct the
   configuration and call init again.
4. Run the same kind of operation with a fresh Interaction ID. Confirm local export and look up the matching
   Trace.

## Read shutdown and delivery status

* `shutdown.completed` tracks cleanup. Retry incomplete cleanup to join the pending work.
* `shutdown.delivery="unknown"` can indicate an uncertain receiver acknowledgement. `"incomplete"` indicates
  lost, rejected or unfinished delivery.
* `shutdown.delivery="exported-not-readback"` describes local delivery. Check actual exported trace counts.
  Zero exported traces leaves capture unverified. A positive count still needs matching authenticated Product
  readback.
* A receiver rejection can point to authorization, the OTLP route, or request compatibility. After a
  correction, a successful export clears the current failure hint and cumulative counters retain prior failed
  attempts.

## Contact support

Email `biz@impact.ai` with the SDK and Python versions, integration, operation, diagnosis code and a
content-free summary of where the Trace stopped. Keep keys, prompt text and raw telemetry private.

## Security reporting

Use the [private security reporting process](https://docs.impact.ai/security.md).


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.