feat:add compaction and log token cost

[Kavya Sree Kaitepalli (KavyaSree2610)]

Summary

• Token usage logging — OpenAIApi and AzureOpenAIApi now log input/output/total token counts returned by the Responses API after every call, warning when usage is absent.
• Server-side compaction — Both clients pass context_management=[{"type": "compaction", ...}] with a configurable compact_threshold (default 200k tokens). When the server compacts, the client rebuilds its local history around the returned compaction item and logs the event. Requires openai==3.17.0; adds a --debug-http CLI flag for inspecting the underlying httpx2 client.
• Cherry-pick — "Add Per-Instance Feedback Generation" commit for auto_memory/eval/swebenchverified.py.
  (ec4ffae)
• Fix twine check — The release failed with ImportError: cannot import name 'errors' from 'packaging', because pip install -e . downgraded twine 7.0's deps (packaging>=26.1, rich>=14.3.3). Bumped packaging to 26.3 and rich to 15.0.0, and moved the build/twine install after the project install.

[Codecov Comments Bot (codecov-commenter)]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 71.52%. Comparing base (713b551) to head (f886f48).
⚠️ Report is 1 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #191      +/-   ##
==========================================
+ Coverage   68.52%   71.52%   +2.99%     
==========================================
  Files          44       44              
  Lines        2821     2946     +125     
==========================================
+ Hits         1933     2107     +174     
+ Misses        888      839      -49     

Flag	Coverage Δ
ghcp	30.65% <13.72%> (-0.51%)	⬇️
integration	40.90% <56.86%> (+0.38%)	⬆️
ollama_local	32.51% <13.72%> (-0.28%)	⬇️
slow-browser	17.31% <10.78%> (-0.20%)	⬇️
slow-other	37.98% <31.37%> (-0.70%)	⬇️
unit	67.75% <100.00%> (+4.12%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

Files with missing lines	Coverage Δ
src/microbots/auto_memory/cli.py	100.00% <100.00%> (ø)
src/microbots/auto_memory/eval/swebenchverified.py	85.20% <100.00%> (+2.24%)	⬆️
src/microbots/llm/azure_openai_api.py	100.00% <100.00%> (ø)
src/microbots/llm/openai_api.py	100.00% <100.00%> (ø)

... and 4 files with indirect coverage changes

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[Kavya Sree Kaitepalli (KavyaSree2610)]

Codecov Report

❌ Patch coverage is 86.27451% with 14 lines in your changes missing coverage. Please review. ✅ Project coverage is 71.04%. Comparing base (713b551) to head (e69707a). ⚠️ Report is 1 commits behind head on main.

Files with missing lines Patch % Lines
src/microbots/auto_memory/eval/swebenchverified.py 81.57% 7 Missing ⚠️
src/microbots/auto_memory/cli.py 40.00% 3 Missing ⚠️
src/microbots/llm/azure_openai_api.py 93.10% 2 Missing ⚠️
src/microbots/llm/openai_api.py 93.33% 2 Missing ⚠️
❌ Your patch status has failed because the patch coverage (86.27%) is below the target coverage (100.00%). You can increase the patch coverage or adjust the target coverage.

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #191      +/-   ##
==========================================
+ Coverage   68.52%   71.04%   +2.52%     
==========================================
  Files          44       44              
  Lines        2821     2946     +125     
==========================================
+ Hits         1933     2093     +160     
+ Misses        888      853      -35     

Flag Coverage Δ
ghcp 30.65% <13.72%> (-0.51%) ⬇️
integration 40.90% <56.86%> (+0.38%) ⬆️
ollama_local 32.51% <13.72%> (-0.28%) ⬇️
slow-browser 17.31% <10.78%> (-0.20%) ⬇️
slow-other 38.28% <31.37%> (-0.39%) ⬇️
unit 67.27% <86.27%> (+3.64%) ⬆️
Flags with carried forward coverage won't be shown. Click here to find out more.

Files with missing lines Coverage Δ
src/microbots/llm/azure_openai_api.py 97.43% <93.10%> (-2.57%) ⬇️
src/microbots/llm/openai_api.py 96.66% <93.33%> (-3.34%) ⬇️
src/microbots/auto_memory/cli.py 91.42% <40.00%> (-8.58%) ⬇️
src/microbots/auto_memory/eval/swebenchverified.py 81.63% <81.57%> (-1.33%) ⬇️
... and 4 files with indirect coverage changes

🚀 New features to boost your workflow:

Copilot add tests to make the patch coverage 100%

[Copilot]

[Codecov](https://app.codecov.io/gh/microsoft/microbots/pull/191?dropdown=coverage&src=pr&el=h1&utm_medium=referral&utm_source=github&utm_content=comment&utm_campaign=pr+comments&utm_term=microso...

Added tests covering the uncovered patch lines in a34a9da — patch coverage is now 100% locally:

src/microbots/auto_memory/cli.py                        35      0 100.00%
src/microbots/auto_memory/eval/swebenchverified.py     196     29  85.20%   86-93, 117-128, 152-161, 194-217, 384-390
src/microbots/llm/azure_openai_api.py                   78      0 100.00%
src/microbots/llm/openai_api.py                         60      0 100.00%

(the remaining swebenchverified.py misses are pre-existing lines outside this patch)

New tests:

• test_ask_warns_when_token_usage_is_missing for both OpenAIApi and AzureOpenAIApi
• test_parse_args_debug_http_defaults_to_false / test_main_enables_http_debug_logging_when_requested for the --debug-http flag
• _generate_instance_feedback: success, bot-unavailable (exception) and bot-failure paths
• _combine_result_feedback: empty-input short circuit and the bot-returned-failure fallback