feat: add prompt-caching examples and migrate active text tests to gpt-6-luna by ranadeepsingh · Pull Request #2781 · microsoft/SynapseML

Rana Singh (ranadeepsingh) · 2026-10-21T19:47:11Z

Why this change

DataFrame workloads often resend the same instructions and reference material for every row. SynapseML currently lacks a dedicated prompt_cache_options parameter and rejects documented input-content cache breakpoints, preventing callers from expressing OpenAI's explicit caching requests.

This PR forwards those options and breakpoints without a client-side model, mode, or lifetime whitelist, and adds a complete notebook walkthrough. It also updates active text-generation examples and live-test selections to gpt-6-luna, while preserving caller-selected deployments, embedding models, and deliberate legacy compatibility coverage. Generic transformers do not acquire an automatic model default.

The existing Fabric model-availability preflight retains its failure behavior and adds conditional guidance about the OpenAI-as-a-Microsoft-Subprocessor setting.

What the notebook demonstrates

A stable, substantial developer-message prefix followed by changing questions.
Explicit request-level options and prompt_cache_breakpoint inside actual input content—not merely an options column.
Scalar options, row-bound options, and an implicit-caching OpenAIPrompt example.
Native OpenAI Python SDK and SynapseML requests, provider errors, and cached-token usage.

OpenAI documentation

Python SDK versus SynapseML

Run these snippets in order with a Spark session, a SynapseML build containing this change, and the OpenAI SDK installed. Set OPENAI_BASE_URL to your Azure OpenAI v1 URL ending in /openai/v1/, OPENAI_API_KEY to your key, and OPENAI_DEPLOYMENT to your chosen deployment. No deployment is created or changed.

Shared input and options

import os

endpoint = os.environ["OPENAI_BASE_URL"]
key = os.environ["OPENAI_API_KEY"]
deployment = os.environ["OPENAI_DEPLOYMENT"]
options = {"mode": "explicit", "ttl": "30m"}
prefix = "Answer using this product catalog.\n" + "\n".join(
    f"Product {n}: warranty {1 + n % 3} years; returns within 30 days with a receipt. "
    "Report shipping damage within seven days; repairs require a serial number."
    for n in range(1, 81)
)
messages = [
    {
        "role": "developer",
        "content": [{
            "type": "input_text",
            "text": prefix,
            "prompt_cache_breakpoint": {"mode": "explicit"},
        }],
    },
    {
        "role": "user",
        "content": [{"type": "input_text", "text": "What is Product 17's warranty?"}],
    },
]

OpenAI Python SDK

from openai import OpenAI

with OpenAI(api_key=key, base_url=endpoint) as client:
    response = client.responses.create(
        model=deployment,
        input=messages,
        prompt_cache_options=options,
        reasoning={"effort": "low"},
        text={"verbosity": "low"},
        store=False,
    )
print(response.output_text)
print(response.usage.input_tokens_details.cached_tokens)

SynapseML

from pyspark.sql import SparkSession
from synapse.ml.services.openai import OpenAIResponses

spark = SparkSession.builder.getOrCreate()
schema = (
    "messages array<struct<role:string,content:array<struct<type:string,text:string,"
    "prompt_cache_breakpoint:map<string,string>>>>>"
)
rows = spark.createDataFrame([(messages,)], schema)
transformer = (
    OpenAIResponses()
    .setUrl(endpoint)
    .setSubscriptionKey(key)
    .setDeploymentName(deployment)
    .setMessagesCol("messages")
    .setPromptCacheOptions(options)
    .setReasoningEffort("low")
    .setVerbosity("low")
    .setOutputCol("response")
    .setErrorCol("error")
)
transformer.transform(rows).select(
    "response.output", "response.usage.input_tokens_details.cached_tokens", "error"
).show(truncate=False)

Mapping and behavior

OpenAI field	SynapseML surface
Top-level `prompt_cache_options`	`setPromptCacheOptions(dict)` or `setPromptCacheOptionsCol(column)`
Responses `input`	`OpenAIResponses.setMessagesCol(column)`
`input[].content[].prompt_cache_breakpoint`	The same nested DataFrame field on existing `input_text`, `input_image`, or `input_file` content
Chat `messages[].content[].prompt_cache_breakpoint`	The same field on supported Chat content; text uses `type="text"`
Responses cached tokens	`response.usage.input_tokens_details.cached_tokens`
Chat cached tokens	`response.usage.prompt_tokens_details.cached_tokens`
`OpenAIPrompt` normalized usage column	`usage.input_token_details["cached_tokens"]` for either API; this is a map, unlike the raw response structs

An unset parameter or null options-column value omits the field, preserving service defaults. Selecting the column binding replaces the scalar binding. JSON values, including booleans, retain their types. Use a Spark struct column for mixed string/Boolean objects; string-valued maps are sufficient for mode and ttl. The inherited Foundry serialization path also forwards the option rather than silently dropping it.

Explicit mode needs input breakpoints; options alone do not enable explicit cache reads or writes. Top-level Responses instructions cannot hold a breakpoint, which is why the example uses developer-message content. Ordinary OpenAIPrompt templates do not insert markers; its notebook example therefore uses implicit mode.

Keep the reusable prefix stable and place variable input after it. The service decides eligibility and support; cache hits, latency improvements, and savings are not guaranteed. The examples do not assert positive cached-token counts.

The examples use 30m and a mode-only breakpoint. The guide documents additional Responses lifetime behavior, while some generated references and provider deployments differ. SynapseML forwards serializable values and exposes provider errors through the error column; it does not enforce a 30m-only rule. prompt_cache_options.ttl is not translated into the older prompt_cache_retention policy. Existing usage schemas expose cached_tokens, but neither the cache-write counters nor their TTL breakdown.

Validation

Affected Scala production/test compilation, all four Scala style scopes, and pinned Black checks passed.
69 focused Scala tests passed, including real Spark HTTP serialization and provider-error propagation.
Freshly generated and isolated installed wrappers: 25 Python tests plus 12 subtests passed; all 251 installed Python members matched their wheels.
Repository-pinned OpenAI SDK 2.47.0: exact extracted notebook request, nested markers, omission, Chat mapping, and provider-error propagation verified with mocked transport.
Exact native notebook example passed against the existing deployment; a deliberately invalid lifetime returned the provider's HTTP 400 invalid_request_error. No cache-hit assertion.
All five executable cache tutorial cells passed with seven mocked requests and seven live requests. The mocked test compares native SDK and SynapseML scalar request bodies exactly and accepts zero cache hits.
Mixed-type struct columns were verified through both direct API transformers and both OpenAIPrompt routes: eight HTTP requests preserved string/Boolean values. The same four routes failed before the fix.
All three runnable snippets above executed successfully against a loopback service with real Spark; native SDK and SynapseML request bodies matched exactly.
Local runtime: JDK 11 and Spark 3.5.0. Python 3.12.3 differs from the target's 3.11.8; the local PySpark Scala library is 2.12.18 versus the build's 2.12.17.
Six independent full-patch precommit review lenses passed on the same frozen 29-path source.
SHA-pinned manual Azure build 242404552 finished failed: Black requires line wrapping in two new OpenAI.ipynb cells. The selected local Python Black checks did not cover notebook formatting. This correction remains outstanding.
Published CI results include 270 OpenAI passes plus two intentional custom-endpoint ignores, 167 cognitive Python passes, 43 CPU and 4 GPU notebook passes, and 111 website sample passes. The new cache serialization, provider-error and Python persistence tests passed.
All 54 coverage-publication tasks were skipped by the manual-run coverage condition. The Azure Pipelines app posted the failed aggregate on the verified synthetic merge commit, not the source-head check rollup. Neither coverage nor full CI is qualified.
Automated review identified one stale review-index entry. Its artifact-only correction links the current finding/resolution and complete precommit pass without changing source. The distinct final CI-qualified review remains pending.
Successful predecessor CI does not qualify this amendment; no package has been released.

azure-pipelines · 2026-10-21T19:47:21Z

Azure Pipelines: There may be pipelines that require an authorized user to comment /azp run to run.

github-actions · 2026-10-21T19:47:22Z

Hey Rana Singh (@ranadeepsingh) 👋!
Thank you so much for contributing to our repository 🙌.
Someone from SynapseML Team will be reviewing this pull request soon.

We use semantic commit messages to streamline the release process.
Before your pull request can be merged, you should make sure your first commit and PR title start with a semantic prefix.
This helps us to create release messages and credit you for your hard work!

Examples of commit messages with semantic prefixes:

fix: Fix LightGBM crashes with empty partitions
feat: Make HTTP on Spark back-offs configurable
docs: Update Spark Serving usage
build: Add codecov support
perf: improve LightGBM memory usage
refactor: make python code generation rely on classes
style: Remove nulls from CNTKModel
test: Add test coverage for CNTKModel

To test your commit locally, please follow our guild on building from source.
Check out the developer guide for additional guidance on testing your change.

## Summary Move the six qualifying pre-commit review artifacts into reviews/pr-2781 without changing their bytes, and add an old-to-new path/hash index. ## Prompting Intent Publish the reviewed change with durable, public-safe evidence in the actual numbered PR directory. Preserve historical feedback and privacy exclusions without recursively reviewing review text. ## Linked Sources - microsoft#2781 - reviews/pr-2781/index.md ## Rationale The initial public commit retains the original artifact paths. The migration index makes their provenance and unchanged bytes explicit, while recording current validation and the still-pending remote CI/final review gates. Only allocated review evidence changes; the reviewed product manifest is unchanged. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> Copilot-Session: 4f9a58fc-ce1f-4a73-951d-4794857a150a

Rana Singh (ranadeepsingh) · 2026-10-21T19:50:05Z

/azp run

azure-pipelines · 2026-10-21T19:50:19Z

Azure Pipelines: Successfully started running 1 pipeline(s).

Copilot

🔵 Needs a closer look

Current-head service-backed CI, branch-pinned Python validation, and final CI-qualified review remain pending.

No open issues 🎉

What changed in this PR

Adds opt-in prompt-cache controls to SynapseML’s OpenAI transformers while preserving provider defaults and caller-selected deployments.

Changes:

Adds validated scalar/per-row cache options, serializer forwarding, and persistence coverage.
Updates current examples and test selections to gpt-6-luna.
Adds conditional Fabric administrator guidance and records validation evidence.

File	Description
reviews/pre-pr-luna-defaults-cache/task-unknown-attempt-2-stage-attempt-1-review-6-gpt-6.1-sol.md	Records hardening review and validation limitations.
reviews/pre-pr-luna-defaults-cache/task-unknown-attempt-2-stage-attempt-1-review-5-gpt-6.1-sol.md	Records test-coverage review.
reviews/pre-pr-luna-defaults-cache/task-unknown-attempt-2-stage-attempt-1-review-4-gpt-6.1-sol.md	Records correctness review.
reviews/pre-pr-luna-defaults-cache/task-unknown-attempt-2-stage-attempt-1-review-3-gpt-6.1-sol.md	Records robustness review.
reviews/pre-pr-luna-defaults-cache/task-unknown-attempt-2-stage-attempt-1-review-2-gpt-6.1-sol.md	Records architecture review.
reviews/pre-pr-luna-defaults-cache/task-unknown-attempt-2-stage-attempt-1-review-1-gpt-6.1-sol.md	Records initial source review.
docs/Explore Algorithms/OpenAI/Quickstart - Understand and Search Forms.ipynb	Updates example deployment.
docs/Explore Algorithms/OpenAI/Quickstart - Multimodal OpenAI Prompter with Responses API.ipynb	Updates Responses example deployment.
docs/Explore Algorithms/OpenAI/OpenAI.ipynb	Documents caching and updates reasoning examples.
docs/Explore Algorithms/OpenAI/Langchain.ipynb	Updates LangChain example deployment.
core/src/test/scala/com/microsoft/azure/synapse/ml/fabric/OpenAIFabricSettingSuite.scala	Tests conditional guidance and existing status gates.
core/src/main/scala/com/microsoft/azure/synapse/ml/fabric/OpenAIFabricSetting.scala	Extends missing-model administrator guidance.
cognitive/src/test/scala/com/microsoft/azure/synapse/ml/services/openai/OpenAIV1EndpointSuite.scala	Updates endpoint-test model fixtures.
cognitive/src/test/scala/com/microsoft/azure/synapse/ml/services/openai/OpenAIResponsesSuite.scala	Updates Responses fixtures.
cognitive/src/test/scala/com/microsoft/azure/synapse/ml/services/openai/OpenAIResponsesMultimodalSuite.scala	Updates multimodal Responses fixtures.
cognitive/src/test/scala/com/microsoft/azure/synapse/ml/services/openai/OpenAIPromptSuite.scala	Updates prompt model selections and assertions.
cognitive/src/test/scala/com/microsoft/azure/synapse/ml/services/openai/OpenAIPromptResponsesSuite.scala	Updates live Responses selection and reasoning settings.
cognitive/src/test/scala/com/microsoft/azure/synapse/ml/services/openai/OpenAIPromptMultimodalRequestSuite.scala	Tests Foundry cache forwarding; updates fixtures.
cognitive/src/test/scala/com/microsoft/azure/synapse/ml/services/openai/OpenAIPromptCacheSuite.scala	Tests cache validation, omission, and serialization.
cognitive/src/test/scala/com/microsoft/azure/synapse/ml/services/openai/OpenAIEmbeddingsSuite.scala	Updates non-embedding default fixture.
cognitive/src/test/scala/com/microsoft/azure/synapse/ml/services/openai/OpenAIChatCompletionSuite.scala	Updates chat test selections and fixtures.
cognitive/src/test/scala/com/microsoft/azure/synapse/ml/services/openai/OpenAIChatCompletionMultimodalSuite.scala	Updates multimodal chat fixtures.
cognitive/src/test/scala/com/microsoft/azure/synapse/ml/services/openai/OpenAIAPIKey.scala	Updates shared live-test deployment.
cognitive/src/test/python/synapsemltest/services/test_ServiceParamPythonBridge.py	Tests generated cache accessors, copying, and persistence.
cognitive/src/test/python/synapsemltest/services/openai/test_StructuredOutput.py	Updates structured-output deployment.
cognitive/src/test/python/synapsemltest/services/openai/test_OpenAIResponsesMultimodal.py	Updates Responses model fixture.
cognitive/src/test/python/synapsemltest/services/openai/test_OpenAIMultimodalRequest.py	Updates request and response fixtures.
cognitive/src/test/python/synapsemltest/services/openai/test_OpenAIDefaults.py	Updates defaults-test deployment.
cognitive/src/test/python/synapsemltest/services/openai/test_OpenAIChatCompletionMultimodal.py	Updates chat model fixture.
cognitive/src/test/python/synapsemltest/services/langchain/test_LangchainTransform.py	Updates LangChain model fixtures.
cognitive/src/main/scala/com/microsoft/azure/synapse/ml/services/openai/OpenAIResponses.scala	Includes cache options in Responses payloads.
cognitive/src/main/scala/com/microsoft/azure/synapse/ml/services/openai/OpenAIChatCompletion.scala	Includes cache options in chat payloads.
cognitive/src/main/scala/com/microsoft/azure/synapse/ml/services/openai/OpenAI.scala	Defines cache parameters and shared validation.
cognitive/src/main/scala/com/microsoft/azure/synapse/ml/services/aifoundry/AIFoundryChatCompletion.scala	Includes cache options in Foundry payloads.

_{Review effort: Balanced}

Copilot

🔵 Needs a closer look

Supplied evidence leaves current-head service-backed CI and branch-pinned Python validation pending for the cross-language API changes.

No open issues 🎉

_{Review effort: Balanced}

## Summary Ask for full-sentence favorite-color answers in three live-test fixtures. Preserve the existing minimum-length assertions and all production behavior. Include the complete fresh six-lens pre-commit evidence and integrate the upstream website lockfile update without rewriting published history. ## Prompting Intent Diagnose PR microsoft#2781's failed OpenAI tests, preserve completed E2E evidence, fix the demonstrated cause without weakening tests, and require fresh engineering validation before final CI-qualified review. ## Linked Sources - PR and current checks: microsoft#2781 - Target update: microsoft#2780 - Review evidence: reviews/pr-2781/index.md - Trusted workflow: .github/skills/synapseml-pr-loop/ ## Rationale Luna correctly returned "Red." to a question that permits a short answer. The tests instead required more than ten characters. Requesting a full sentence aligns the fixture with that assertion rather than weakening the assertion, changing deployments, or modifying serializers. All 55 affected live tests passed locally with one unchanged intentional ignore. The complete frozen patch passed six sequential independent gpt-6.1-sol/max reviews. Fresh PR-head CI and final CI-qualified review remain required. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> Copilot-Session: 4f9a58fc-ce1f-4a73-951d-4794857a150a

Rana Singh (ranadeepsingh) · 2026-10-21T21:51:21Z

/azp run

azure-pipelines · 2026-10-21T21:51:45Z

Azure Pipelines: Successfully started running 1 pipeline(s).

Copilot

🔵 Needs a closer look

The supplied evidence leaves current-head CI, branch-pinned Python validation, and the final CI-qualified review outstanding for these cross-language, provider-facing changes.

No open issues 🎉

_{Review effort: Balanced}

## Summary Move the six qualifying pre-commit review artifacts into reviews/pr-2781 without changing their bytes, and add an old-to-new path/hash index. ## Prompting Intent Publish the reviewed change with durable, public-safe evidence in the actual numbered PR directory. Preserve historical feedback and privacy exclusions without recursively reviewing review text. ## Linked Sources - microsoft#2781 - reviews/pr-2781/index.md ## Rationale The initial public commit retains the original artifact paths. The migration index makes their provenance and unchanged bytes explicit, while recording current validation and the still-pending remote CI/final review gates. Only allocated review evidence changes; the reviewed product manifest is unchanged. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> Copilot-Session: 4f9a58fc-ce1f-4a73-951d-4794857a150a

## Summary Ask for full-sentence favorite-color answers in three live-test fixtures. Preserve the existing minimum-length assertions and all production behavior. Include the complete fresh six-lens pre-commit evidence and integrate the upstream website lockfile update without rewriting published history. ## Prompting Intent Diagnose PR microsoft#2781's failed OpenAI tests, preserve completed E2E evidence, fix the demonstrated cause without weakening tests, and require fresh engineering validation before final CI-qualified review. ## Linked Sources - PR and current checks: microsoft#2781 - Target update: microsoft#2780 - Review evidence: reviews/pr-2781/index.md - Trusted workflow: .github/skills/synapseml-pr-loop/ ## Rationale Luna correctly returned "Red." to a question that permits a short answer. The tests instead required more than ten characters. Requesting a full sentence aligns the fixture with that assertion rather than weakening the assertion, changing deployments, or modifying serializers. All 55 affected live tests passed locally with one unchanged intentional ignore. The complete frozen patch passed six sequential independent gpt-6.1-sol/max reviews. Fresh PR-head CI and final CI-qualified review remain required. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> Copilot-Session: 4f9a58fc-ce1f-4a73-951d-4794857a150a

## Summary Repair the review archive's commit references after rebasing onto current master and distinguish successful historical tests from failed coverage publication. All twelve original review files and the complete reviewed source patch remain unchanged. ## Prompting Intent The engineer requested rebasing the personal feature branch onto the latest target and rerunning CI, while preserving review history, exact source evidence, privacy boundaries, and the original remote-head lease. ## Linked Sources - Pull request: microsoft#2781 - Target: microsoft@71346cf - Review evidence and migration hashes: reviews/pr-2781/index.md - Trusted loop and evidence retention: .github/skills/synapseml-pr-loop/references/loop-control.md ## Rationale Keep the original feedback bytes and repaired reachable commit links instead of rewriting reviews or treating stale CI as current. The historical R tests passed; their coverage upload failed TLS certificate-name validation. Fresh configured CI must validate the new head without bypassing coverage, TLS, tests, or review gates. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> Copilot-Session: 4f9a58fc-ce1f-4a73-951d-4794857a150a

Rana Singh (ranadeepsingh) · 2026-10-23T07:52:10Z

/azp run

azure-pipelines · 2026-10-23T07:52:23Z

Azure Pipelines: Successfully started running 1 pipeline(s).

Copilot

🔵 Needs a closer look

Supplied evidence does not establish completed current-head CI and final qualification for the cross-service Scala/Python API changes.

No open issues 🎉

_{Review effort: Balanced}

## Summary Add optional prompt cache options to OpenAI Chat Completions, Responses and OpenAIPrompt, including its Foundry delegate. Update active text-generation examples and test selections to gpt-6-luna, and make the existing Fabric missing-model error offer conditional administrator guidance. ## Prompting Intent Expose strict optional mode/TTL controls through Scala and generated Python map/column bindings. Preserve caller-selected deployments, embedding models, historical documentation and deliberate offline compatibility tests. Do not introduce a generic transformer model default or automatic cache breakpoints. ## Linked Sources - User-requested cache controls, active Luna selections, and model-availability guidance. - https://developers.openai.com/api/docs/guides/prompt-caching - https://learn.microsoft.com/en-us/fabric/admin/service-admin-portal-copilot - reviews/pre-pr-luna-defaults-cache/: six independent pre-commit review artifacts. ## Rationale Reuse ServiceParam serialization and generated wrappers rather than adding a parallel Python API. Validate both scalar and row-bound options before request serialization. Preserve the existing Fabric availability gate and add only conditional advice. A Foundry forwarding regression found during review is covered by request-level and public-transform tests. Local compilation/style checks, targeted Scala tests, notebook/Black checks, and fresh generated-wrapper tests passed. Exact branch Python and current-head remote CI remain pending. Public-safe review artifacts use the explicitly authorized temporary pre-PR directory until the real PR number is assigned. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> Copilot-Session: 4f9a58fc-ce1f-4a73-951d-4794857a150a

## Summary Move the six qualifying pre-commit review artifacts into reviews/pr-2781 without changing their bytes, and add an old-to-new path/hash index. ## Prompting Intent Publish the reviewed change with durable, public-safe evidence in the actual numbered PR directory. Preserve historical feedback and privacy exclusions without recursively reviewing review text. ## Linked Sources - microsoft#2781 - reviews/pr-2781/index.md ## Rationale The initial public commit retains the original artifact paths. The migration index makes their provenance and unchanged bytes explicit, while recording current validation and the still-pending remote CI/final review gates. Only allocated review evidence changes; the reviewed product manifest is unchanged. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> Copilot-Session: 4f9a58fc-ce1f-4a73-951d-4794857a150a

## Summary Ask for full-sentence favorite-color answers in three live-test fixtures. Preserve the existing minimum-length assertions and all production behavior. Include the complete fresh six-lens pre-commit evidence and integrate the upstream website lockfile update without rewriting published history. ## Prompting Intent Diagnose PR microsoft#2781's failed OpenAI tests, preserve completed E2E evidence, fix the demonstrated cause without weakening tests, and require fresh engineering validation before final CI-qualified review. ## Linked Sources - PR and current checks: microsoft#2781 - Target update: microsoft#2780 - Review evidence: reviews/pr-2781/index.md - Trusted workflow: .github/skills/synapseml-pr-loop/ ## Rationale Luna correctly returned "Red." to a question that permits a short answer. The tests instead required more than ten characters. Requesting a full sentence aligns the fixture with that assertion rather than weakening the assertion, changing deployments, or modifying serializers. All 55 affected live tests passed locally with one unchanged intentional ignore. The complete frozen patch passed six sequential independent gpt-6.1-sol/max reviews. Fresh PR-head CI and final CI-qualified review remain required. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> Copilot-Session: 4f9a58fc-ce1f-4a73-951d-4794857a150a

## Summary Repair the review archive's commit references after rebasing onto current master and distinguish successful historical tests from failed coverage publication. All twelve original review files and the complete reviewed source patch remain unchanged. ## Prompting Intent The engineer requested rebasing the personal feature branch onto the latest target and rerunning CI, while preserving review history, exact source evidence, privacy boundaries, and the original remote-head lease. ## Linked Sources - Pull request: microsoft#2781 - Target: microsoft@71346cf - Review evidence and migration hashes: reviews/pr-2781/index.md - Trusted loop and evidence retention: .github/skills/synapseml-pr-loop/references/loop-control.md ## Rationale Keep the original feedback bytes and repaired reachable commit links instead of rewriting reviews or treating stale CI as current. The historical R tests passed; their coverage upload failed TLS certificate-name validation. Fresh configured CI must validate the new head without bypassing coverage, TLS, tests, or review gates. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> Copilot-Session: 4f9a58fc-ce1f-4a73-951d-4794857a150a

## Summary Keep the initial migration focused on explicit gpt-6-luna selections in current examples and live test fixtures, necessary fixture compatibility, and conditional Fabric model-availability advice. Restore the new cache-related production and bridge changes to the target baseline, remove the added cache-only tests and documentation, and retain all prior public review evidence. ## Prompting Intent The engineer requested a lean initial Luna migration with all new prompt-cache controls deferred to a later follow-up. Preserve existing caching behavior, caller-selected deployments, embeddings, legacy compatibility cases, historical documentation and generic transformer defaults. Preserve the exact deferred implementation and prior artifacts; do not invent a release or alter policies, deployments, tenant settings, approvals or merge state. ## Linked Sources - Pull request: microsoft#2781 - Current source review: reviews/pr-2781/task-unknown-attempt-4-stage-attempt-1-review-6-gpt-6.1-sol.md - Retained review history and migration accounting: reviews/pr-2781/index.md - Public repository workflow: .github/skills/synapseml-pr-loop/SKILL.md ## Rationale The model migration does not require a new caching API or dependency release. Restoring the existing cognitive API preserves its established behavior and avoids coupling initial migration delivery to the deferred feature. Prior implementation bytes and candidate artifacts remain preserved separately, and all earlier public reviewer responses remain intact. ## Validation Four compilation scopes, four Scala style scopes, 60 offline Scala tests, 55 live Scala tests, 23 installed Python bridge tests plus six subtests, Black on six changed Python files, four notebook structural checks and both code-generation tasks passed. One existing Custom EndPoint test remains intentionally ignored. Six sequential independent gpt-6.1-sol/max pre-commit reviews are clean. Local Python3.12.3 and bundled Scala2.12.18 are not exact matches for target Python3.11.8 and build Scala2.12.17. The frozen23-path patch is independently byte-identical when applied to the latest website-only target update. Branch integration, fresh current-head CI and automated review, and the separate final CI-qualified review remain pending. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> Copilot-Session: 4f9a58fc-ce1f-4a73-951d-4794857a150a

## Summary Repair archival review links after rebasing the model-only migration onto latest master, and distinguish passed historical tests from failed coverage publication and still-required fresh narrowed-head validation. ## Prompting Intent Deliver only the Luna model migration and necessary compatibility guidance, defer new prompt-cache controls, preserve review originals, integrate latest master under the explicit ownership guard, and rerun current-head CI. ## Linked Sources - Public PR: microsoft#2781 - Review evidence: reviews/pr-2781/index.md - Trusted workflow: .github/skills/synapseml-pr-loop/SKILL.md - Trusted retention and CI contracts: .github/skills/synapseml-pr-loop/references/loop-control.md and ci-evidence.md - Historical pipeline: https://dev.azure.com/microsoft/SynapseML/_build/results?buildId=242356862 ## Rationale The latest target changes only website files. The full 23-path source patch, canonical manifest, and every earlier review artifact are unchanged. Keep content review evidence, but require fresh SHA-bound CI and the separate final review. Preserve all history rather than squash or rewrite original feedback. This documentation-only provenance update deliberately excludes private tracking identifiers under the publication confidentiality requirement. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> Copilot-Session: 4f9a58fc-ce1f-4a73-951d-4794857a150a

Rana Singh (ranadeepsingh) · 2026-10-23T09:56:20Z

/azp run

Copilot

🔵 Needs a closer look

Fresh CI for the narrowed current head and the separate final CI-qualified review remain outstanding.

No open issues 🎉

_{Review effort: Balanced}

azure-pipelines · 2026-10-23T09:56:32Z

Azure Pipelines: Successfully started running 1 pipeline(s).

## Summary Let DataFrame callers express OpenAI prompt caching through request-level options and actual input-content breakpoints. Add executable native Python SDK and SynapseML examples while preserving the existing Luna migration. ## Prompting Intent Demonstrate explicit options, input breakpoints and row-bound configuration in the existing notebook. Preserve serializable values, let the provider validate cache policy, and explain defaults, usage and nonguaranteed hits. Keep deployment choice, existing compatibility behavior and package pins. ## Linked Sources - Pull request: microsoft#2781 - Caching guide: https://developers.openai.com/api/docs/guides/prompt-caching - Responses API: https://developers.openai.com/api/reference/resources/responses/methods/create - Chat API: https://developers.openai.com/api/reference/resources/chat - Python SDK: https://developers.openai.com/api/docs/libraries - Verified request types: https://github.com/openai/openai-python/blob/81792ff040a2ca2b70348163049d3d69d5ba95cf/src/openai/types/responses/response_create_params.py - Review evidence and finding resolution: reviews/pr-2781/ ## Rationale Use the existing ServiceParam scalar/column abstraction rather than adding a separate configuration hierarchy or a model/mode/TTL whitelist. Normalize nested Spark structs only for cache options so mixed JSON types survive without changing unrelated serializers. Retain focused before/after HTTP, wrapper and persistence tests and all review evidence. Local gates passed: 69 Scala tests, 25 Python tests plus 12 subtests, styles, Black, generated artifacts and exact notebook mock/live examples. Six independent precommit lenses are clean. Current-head CI and final review remain required; no release or merge is performed. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> Copilot-Session: 4f9a58fc-ce1f-4a73-951d-4794857a150a

Copilot

🔵 Needs a closer look

Current-head CI and final qualification remain pending for the cross-language serialization and live-deployment changes.

1 open issue

Update index to reflect cache passes and current validation scope · New

_{Review effort: Balanced}

## Summary Correct the review index so readers see the cache-refined source, pass-5 finding and resolution, and the complete clean pass-6 evidence first. Record the actual failed notebook-style gate and unqualified manual-run coverage without implying that passing functional tests make CI green. ## Prompting Intent Address the automated review's stale-index finding, retain earlier feedback, and distinguish current evidence from historical model-only qualification. Publish only allocated review metadata under the artifact-only exception. Do not modify source, reset review budgets, waive checks, or claim readiness. ## Linked Sources - PR: microsoft#2781 - Finding: microsoft#2781 (comment) - Reviewed source: microsoft@51343aa - Current review aggregate: reviews/pr-2781/task-unknown-attempt-6-stage-attempt-1-review-6-gpt-6.1-sol.md - Finding resolution: reviews/pr-2781/task-unknown-attempt-5-stage-attempt-1-review-5-gpt-6.1-sol.md - Artifact-only policy: .github/skills/synapseml-pr-loop/references/loop-control.md ## Rationale Keep the historical sections intact but clearly subordinate to current scope and evidence. The complete 29-path source manifest is unchanged. Named CI test proof is separate from the failed style, skipped coverage, required-check binding and final-review gates. The exhausted source-review budget remains a blocker, not permission to silently extend the process. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> Copilot-Session: 4f9a58fc-ce1f-4a73-951d-4794857a150a

Copilot

🟡 Changes recommended

The unresolved notebook-formatting failure blocks the required CI gate, and successful current-source qualification remains outstanding.

1 open issue

Format cache-example cells to pass the pinned Black check · New

1 resolved since last review

Update index to reflect cache passes and current validation scope

_{Review effort: Balanced}

+    "    [\n",
+    "        StructField(\"type\", StringType(), False),\n",
+    "        StructField(\"text\", StringType(), True),\n",
+    "        StructField(\"prompt_cache_breakpoint\", MapType(StringType(), StringType()), True),\n",


Rana Singh (ranadeepsingh) requested a review from Brendan Walsh (BrendanWalsh) as a code owner October 21, 2026 19:47

Copilot AI balanced review requested due to automatic review settings October 21, 2026 19:47

Copilot started reviewing on behalf of Rana Singh (ranadeepsingh) October 21, 2026 19:47

View session

Copilot AI reviewed Oct 21, 2026

View reviewed changes

Copilot AI balanced review requested due to automatic review settings October 21, 2026 19:53

Copilot started reviewing on behalf of Rana Singh (ranadeepsingh) October 21, 2026 19:53

View session

Copilot AI reviewed Oct 21, 2026

View reviewed changes

Copilot AI balanced review requested due to automatic review settings October 21, 2026 21:50

Copilot started reviewing on behalf of Rana Singh (ranadeepsingh) October 21, 2026 21:51

View session

Copilot AI reviewed Oct 21, 2026

View reviewed changes

Rana Singh (ranadeepsingh) force-pushed the copilot/openai-luna-defaults branch from 32395b9 to d68a73a Compare October 23, 2026 07:50

Copilot AI balanced review requested due to automatic review settings October 23, 2026 07:50

Copilot started reviewing on behalf of Rana Singh (ranadeepsingh) October 23, 2026 07:51

View session

Copilot AI reviewed Oct 23, 2026

View reviewed changes

Rana Singh (ranadeepsingh) and others added 3 commits October 23, 2026 04:44

Rana Singh (ranadeepsingh) force-pushed the copilot/openai-luna-defaults branch from d68a73a to 9cd4ce1 Compare October 23, 2026 09:52

Rana Singh (ranadeepsingh) changed the title ~~feat: expose prompt cache controls and update OpenAI examples~~ fix: migrate active OpenAI examples and tests to gpt-6-luna Oct 23, 2026

Copilot AI balanced review requested due to automatic review settings October 23, 2026 09:52

Copilot started reviewing on behalf of Rana Singh (ranadeepsingh) October 23, 2026 09:52

View session

Copilot AI reviewed Oct 23, 2026

View reviewed changes

Rana Singh (ranadeepsingh) changed the title ~~fix: migrate active OpenAI examples and tests to gpt-6-luna~~ feat: add prompt-caching examples and migrate active text tests to gpt-6-luna Oct 23, 2026

Copilot AI balanced review requested due to automatic review settings October 23, 2026 13:50

Copilot started reviewing on behalf of Rana Singh (ranadeepsingh) October 23, 2026 13:50

View session

Copilot AI reviewed Oct 23, 2026

View reviewed changes

Comment thread reviews/pr-2781/index.md Outdated

Copilot AI balanced review requested due to automatic review settings October 23, 2026 17:55

Copilot started reviewing on behalf of Rana Singh (ranadeepsingh) October 23, 2026 17:56

View session

Copilot AI reviewed Oct 23, 2026

View reviewed changes

Comment thread docs/Explore Algorithms/OpenAI/OpenAI.ipynb

" [\n",

" StructField(\"type\", StringType(), False),\n",

" StructField(\"text\", StringType(), True),\n",

" StructField(\"prompt_cache_breakpoint\", MapType(StringType(), StringType()), True),\n",

Conversation

Rana Singh (ranadeepsingh) commented Oct 21, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Why this change

What the notebook demonstrates

OpenAI documentation

Python SDK versus SynapseML

Mapping and behavior

Validation

Uh oh!

azure-pipelines Bot commented Oct 21, 2026

Uh oh!

github-actions Bot commented Oct 21, 2026

Uh oh!

Rana Singh (ranadeepsingh) commented Oct 21, 2026

Uh oh!

azure-pipelines Bot commented Oct 21, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

🔵 Needs a closer look

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

🔵 Needs a closer look

Uh oh!

Rana Singh (ranadeepsingh) commented Oct 21, 2026

Uh oh!

azure-pipelines Bot commented Oct 21, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

🔵 Needs a closer look

Uh oh!

Rana Singh (ranadeepsingh) commented Oct 23, 2026

Uh oh!

azure-pipelines Bot commented Oct 23, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

🔵 Needs a closer look

Uh oh!

Rana Singh (ranadeepsingh) commented Oct 23, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

🔵 Needs a closer look

Uh oh!

azure-pipelines Bot commented Oct 23, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

🔵 Needs a closer look

Uh oh!

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

🟡 Changes recommended

Uh oh!

This branch has not been deployed

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Rana Singh (ranadeepsingh) commented Oct 21, 2026 •

edited

Loading