fix(weave): WEAVE_DEBUG_HTTP exception on server_info check

[jamie-rasmussen]

Description

I was doing a WEAVE_DEBUG_HTTP=1 python3 test.py where test.py just imported weave and init'ed a project, and it was showing retries before throwing a RuntimeError:

...
weave: retry_attempt
---------------------
Time: 21:56:57.074538
Thread ID: 8366911616
Method: GET
URL: https://trace.wandb.ai/server_info
Headers:
  Host: trace.wandb.ai
  Accept: */*
  Accept-Encoding: gzip, deflate
  Connection: keep-alive
  User-Agent: python-httpx/0.28.1
  X-Weave-Retry-Id: 019c5a4b-54b2-7550-a7d9-8b6d82051d87
  Authorization: <Redacted>
Body:
  None
----- Response below -----
Elapsed Time: 0.07 seconds
Status Code: 200
Reason: OK
Headers:
  date: Sat, 14 Feb 2026 03:56:56 GMT
  server: uvicorn
  content-length: 77
  content-type: application/json
  via: 1.1 google
  alt-svc: h3=":443"; ma=2592000,h3-29=":443"; ma=2592000
Body:
weave: retry_failed
weave: Unexpected error when checking if Weave is available on the server.  Please contact support.
Traceback (most recent call last):
  File "/Users/jrasmussen/source/jcr/suprot_20260213/debug.py", line 3, in <module>
    weave.init("jamie-rasmussen/2026-02-13_test")
    ~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Library/Frameworks/Python.framework/Versions/3.14/lib/python3.14/site-packages/weave/trace/api.py", line 109, in init
    return weave_init.init_weave(
           ~~~~~~~~~~~~~~~~~~~~~^
        project_name,
        ^^^^^^^^^^^^^
    )
    ^
  File "/Library/Frameworks/Python.framework/Versions/3.14/lib/python3.14/site-packages/weave/trace/weave_init.py", line 144, in init_weave
    raise RuntimeError(
        "Weave is not available on the server.  Please contact support."
    )
RuntimeError: Weave is not available on the server.  Please contact support.

This might have been broken by the switch to httpx in #5502 ?

Claude's analysis and fix is in this PR.

The problem: HTTPTransport.handle_request returns a response whose stream hasn't been consumed — that's normally the client layer's job. But pprint_response accesses response.text (which requires the body to be read), causing the
  ResponseNotRead exception. This exception then gets caught by tenacity's retry logic, which retries the request (and fails again the same way).
                                                                                                                                                                                                                                               
  The response.read() call eagerly consumes the response body at the transport level so the debug logging can inspect it. The response is still returned normally to the client afterwards since httpx caches the read content.

[codecov]

Codecov Report

❌ Patch coverage is 0% with 1 line in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
weave/utils/http_requests.py	0.00%	1 Missing ⚠️

📢 Thoughts on this report? Let us know!

[wandbot-3000]

Preview this PR with FeatureBee: https://beta.wandb.ai/?betaVersion=7c2157711b03ae46946e46c3a5410e1c5e6bf431