fix(workflow-operator): read chat-provider responses for the image question-answering tasks

[PG1204]

What changes were proposed in this PR?

When the operator falls back from hf-inference to a third-party chat-completions provider, the reply comes back as {"choices": [{"message": {"content": ...}}]}. Three image tasks in ImageTaskCodegen.parsePython could not read that shape, so a correct answer was written to the result column as a raw JSON envelope:

• visual-question-answering and document-question-answering returned body.get("answer", json.dumps(body)), and a chat response has no answer key.
• zero-shot-image-classification shared the image-only branch, which always returnsjson.dumps(body).

Both now read choices[0]["message"]["content"] when the body carries choices, keeping the native hf-inference shape as the primary path. zero-shot-image-classification gets its own branch, placed ahead of the image-only tasks because the generated if/elif chain is first-match-wins. This is the same idiom image-to-text and image-text-to-text already use in this file, and the one applied to the text tasks in #7798.

image-classification, object-detection and image-segmentation are left as they are: they have no question to answer, so a free-text chat reply is not meaningful structured output for them.

This is Part A of #7906 and covers the response side only. The request side which is carrying candidate_labels into the chat message for zero-shot-image-classification, follows in Part B.

Any related issues?

Addresses #7906

How was this PR tested?

133 tests pass in the WorkflowOperator Hugging Face suites, PythonCodeRawInvalidTextSpec py-compiles the generated Python for all 117 operators, and scalafmtCheck is clean for main and test sources. Two tests were added to ImageTaskCodegenSpec: one asserts the visual/document question-answering branch reads choices ahead of the native answer lookup, the other asserts the new zero-shot-image-classification branch exists and precedes the image-only branch.

The emitted Python was also exercised directly: the three fixed tasks return the chat content, native hf-inference responses parse exactly as before, non-dict and answer-less bodies still fall through to json.dumps, and the untouched branches (image-classification, object-detection, image-segmentation, image-to-text, image-text-to-text) are unchanged.

Was this PR authored or co-authored using generative AI tooling?

Yes, this PR was co-authored with Claude in compliance with ASF policy.

[github-actions]

Backport auto-label report

This fix: PR was checked against each actively-supported release branch. release/* labels drive the post-merge backport, so add or remove one to change where this fix lands.

Release branch

Analysis

⚠️ release/v1.2

Not labeled automatically — none of the files this PR modifies exist on this branch (common/workflow-operator/src/main/scala/org/apache/texera/amber/operator/huggingFace/codegen/ImageTaskCodegen.scala, common/workflow-operator/src/test/scala/org/apache/texera/amber/operator/huggingFace/codegen/ImageTaskCodegenSpec.scala). The fix may target code that isn't on this release, or the files were moved/renamed after the branch was cut. Please check and add release/v1.2 by hand if this fix should be backported here.

Auto-label run.

[github-actions]

Automated Reviewer Suggestions

Based on the git blame history of the changed files, we recommend the following reviewers:

• No candidates found from git blame history.

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 92.46%. Comparing base (7738b6c) to head (0122bdc).
⚠️ Report is 7 commits behind head on main.

Additional details and impacted files

@@             Coverage Diff              @@
##               main    #7920      +/-   ##
============================================
+ Coverage     92.43%   92.46%   +0.03%     
- Complexity     4590     4605      +15     
============================================
  Files          1173     1173              
  Lines         47347    47355       +8     
  Branches       5306     5302       -4     
============================================
+ Hits          43764    43787      +23     
+ Misses         1995     1992       -3     
+ Partials       1588     1576      -12     

Flag	Coverage Δ		*Carryforward flag
access-control-service	81.00% <ø> (ø)
agent-service	98.62% <ø> (ø)		Carriedforward from 7738b6c
amber	89.12% <ø> (+0.09%)	⬆️
computing-unit-managing-service	73.67% <ø> (ø)
config-service	86.73% <ø> (ø)
file-service	82.59% <ø> (ø)
frontend	94.30% <ø> (ø)		Carriedforward from 7738b6c
notebook-migration-service	79.13% <ø> (ø)
pyamber	97.57% <ø> (ø)		Carriedforward from 7738b6c
workflow-compiling-service	77.19% <ø> (ø)

*This pull request uses carry forward flags. Click here to find out more.

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[github-actions]

⚠️ Benchmark changes need a look

🟢 2 better · 🔴 1 worse · ⚪ 12 noise (<±5%) · 0 without baseline

Compared against main 84e2d7d benchmarked on this same runner, so the delta is largely free of cross-runner hardware noise. The "7d avg" column still reflects the gh-pages dashboard. Treat <±5% as noise unless repeated.

Dashboard · Run

	config	throughput	MB/s	latency	max Δ latest / 7d
🔴	bs=10 sw=10 sl=64	428	0.261	22,685/29,605/29,605 us	🟢 -12.7% / 🔴 +110.5%
⚪	bs=100 sw=10 sl=64	957	0.584	104,898/122,463/122,463 us	⚪ within ±5% / 🔴 +26.9%
⚪	bs=1000 sw=10 sl=64	1,106	0.675	899,427/1,002,962/1,002,962 us	⚪ within ±5% / 🔴 +7.8%

Baseline details

Latest main 84e2d7d from same runner

config	metric	PR	latest main	7d avg	Δ latest	Δ 7d
bs=10 sw=10 sl=64	throughput	428 tuples/sec	443 tuples/sec	863.81 tuples/sec	-3.4%	-50.5%
bs=10 sw=10 sl=64	MB/s	0.261 MB/s	0.27 MB/s	0.527 MB/s	-3.3%	-50.5%
bs=10 sw=10 sl=64	p50	22,685 us	21,558 us	11,488 us	+5.2%	+97.5%
bs=10 sw=10 sl=64	p95	29,605 us	33,931 us	14,067 us	-12.7%	+110.5%
bs=10 sw=10 sl=64	p99	29,605 us	33,931 us	17,591 us	-12.7%	+68.3%
bs=100 sw=10 sl=64	throughput	957 tuples/sec	948 tuples/sec	1,127 tuples/sec	+0.9%	-15.1%
bs=100 sw=10 sl=64	MB/s	0.584 MB/s	0.578 MB/s	0.688 MB/s	+1.0%	-15.1%
bs=100 sw=10 sl=64	p50	104,898 us	103,445 us	89,893 us	+1.4%	+16.7%
bs=100 sw=10 sl=64	p95	122,463 us	127,292 us	96,506 us	-3.8%	+26.9%
bs=100 sw=10 sl=64	p99	122,463 us	127,292 us	104,915 us	-3.8%	+16.7%
bs=1000 sw=10 sl=64	throughput	1,106 tuples/sec	1,102 tuples/sec	1,156 tuples/sec	+0.4%	-4.3%
bs=1000 sw=10 sl=64	MB/s	0.675 MB/s	0.673 MB/s	0.705 MB/s	+0.3%	-4.3%
bs=1000 sw=10 sl=64	p50	899,427 us	907,952 us	886,448 us	-0.9%	+1.5%
bs=1000 sw=10 sl=64	p95	1,002,962 us	964,714 us	930,093 us	+4.0%	+7.8%
bs=1000 sw=10 sl=64	p99	1,002,962 us	964,714 us	963,104 us	+4.0%	+4.1%

Raw CSV

config_idx,batch_size,schema_width,string_len,num_batches,total_ms,total_tuples,total_bytes,tuples_per_sec,mb_per_sec,lat_p50_us,lat_p95_us,lat_p99_us
0,10,10,64,20,467.22,200,128000,428,0.261,22684.89,29605.43,29605.43
1,100,10,64,20,2089.66,2000,1280000,957,0.584,104898.01,122463.19,122463.19
2,1000,10,64,20,18090.76,20000,12800000,1106,0.675,899427.35,1002962.15,1002962.15